Total Protein Methods and Their Potential Utility to Reduce the Risk of Food Protein Adulteration


  • Jeffrey C. Moore,

    1. Authors Moore, Lipp, Griffiths, and Abernethy are with the US Pharmacopeial Convention, 12601 Twinbrook Parkway, Rockville, MD 20852-1790, U.S.A. Author DeVries is a member of the US Pharmacopeial Convention's Food Ingredients Expert Committee. Author Abernethy is now with the Office of Clinical Pharmacology, Food and Drug Administration, Silver Spring, MD 20993, U.S.A. Direct inquiries to author Moore (
    Search for more papers by this author
  • Jonathan W. DeVries,

    1. Authors Moore, Lipp, Griffiths, and Abernethy are with the US Pharmacopeial Convention, 12601 Twinbrook Parkway, Rockville, MD 20852-1790, U.S.A. Author DeVries is a member of the US Pharmacopeial Convention's Food Ingredients Expert Committee. Author Abernethy is now with the Office of Clinical Pharmacology, Food and Drug Administration, Silver Spring, MD 20993, U.S.A. Direct inquiries to author Moore (
    Search for more papers by this author
  • Markus Lipp,

    1. Authors Moore, Lipp, Griffiths, and Abernethy are with the US Pharmacopeial Convention, 12601 Twinbrook Parkway, Rockville, MD 20852-1790, U.S.A. Author DeVries is a member of the US Pharmacopeial Convention's Food Ingredients Expert Committee. Author Abernethy is now with the Office of Clinical Pharmacology, Food and Drug Administration, Silver Spring, MD 20993, U.S.A. Direct inquiries to author Moore (
    Search for more papers by this author
  • James C. Griffiths,

    1. Authors Moore, Lipp, Griffiths, and Abernethy are with the US Pharmacopeial Convention, 12601 Twinbrook Parkway, Rockville, MD 20852-1790, U.S.A. Author DeVries is a member of the US Pharmacopeial Convention's Food Ingredients Expert Committee. Author Abernethy is now with the Office of Clinical Pharmacology, Food and Drug Administration, Silver Spring, MD 20993, U.S.A. Direct inquiries to author Moore (
    Search for more papers by this author
  • Darrell R. Abernethy

    1. Authors Moore, Lipp, Griffiths, and Abernethy are with the US Pharmacopeial Convention, 12601 Twinbrook Parkway, Rockville, MD 20852-1790, U.S.A. Author DeVries is a member of the US Pharmacopeial Convention's Food Ingredients Expert Committee. Author Abernethy is now with the Office of Clinical Pharmacology, Food and Drug Administration, Silver Spring, MD 20993, U.S.A. Direct inquiries to author Moore (
    Search for more papers by this author


Abstract:  Kjeldahl and combustion (Dumas) methods are widely accepted for total protein determination but lack analytical selectivity for protein because they measure protein on the basis of sample nitrogen content. Adulteration incidents exploiting this analytical vulnerability (for example, melamine) demonstrate that these methods are no longer sufficient to protect the public health. This article explores the challenges and opportunities to move beyond total nitrogen based methods for total protein measurement. First, it explores the early history of protein measurement science, complexities of current global protein measurement activities, and ideal analytical performance characteristics for new methods. Second, it comprehensively reviews the pros and cons of current and emerging approaches for protein measurement, including their selectivity for protein, ability to detect adulteration, and practicality for routine use throughout the supply chain. It concludes that some existing highly selective methods for food protein measurement have potential for routine quality control. It also concludes that their successful implementation will require matrix-specific validation and the use of supporting reference materials. These methods may be suitable only for food ingredients that have a low degree of compositional variability and are not complex finished food products.


Reliable quantitative methods to determine the total protein content of foods and food ingredients are essential not only to ensure the quality and safety of food but also to facilitate trade. The lack of such methods led directly to the adulteration of protein-based foods with melamine and related nonprotein compounds in 2007 to 2008. This adulteration resulted in injury and death of infants in the case of adulterated milk formula and the injury and death of pets in the case of adulterated pet food. A consequence of these tragedies was the implementation of trade restrictions by several countries (Ingelfinger 2008; Kennedy 2008; Xin and Stone 2008). Investigations into these incidents demonstrated that sellers of protein-based food ingredients artificially increased the apparent protein content of these ingredients and thus their economic value by adding the nonprotein nitrogen-containing compounds melamine and cyanuric acid. As this review describes, analytical methods used to determine total protein content rely on total nitrogen content and do not distinguish protein-based nitrogen from nonprotein nitrogen.

This lack of analytical selectivity for food protein and the potential for adulteration with nonprotein nitrogen have been recognized since at least the mid-20th century (Huss 1959). Selectivity is used throughout this paper to express the extent to which a method can determine an analyte in a matrix without interferences from other components with similar behavior, a quality criterion often used interchangeably with specificity (Vessman and others 2001). These weaknesses were exploited in the 1970s and 1980s when melamine and urea were fraudulently added to animal-based meals and wheat, respectively (Cattaneo and Cantoni 1979; Cattaneo and Cantoni 1982; Folkenberg and others 1990). While the recent melamine incidents thus were not the first of their kind, they certainly represent a dramatic impact on public health. Such exploitation with nonprotein nitrogen will continue to occur in the future unless protein-specific analytical procedures are adopted (Abernethy and others 2008). Approaches to mitigate future fraudulent or harmful adulteration of food protein ingredients, to protect public health, and to facilitate trade will require the development and adoption of reliable, specific quantification methods.

The 2 primary chemically based methods for protein measurement—the Kjeldahl method and the combustion (Dumas) method—rely on total nitrogen determination as a marker to estimate protein content. These methods are used in the food industry and food regulatory laboratories to perform total protein measurements (Krotz and others 2008). Other analytical methodologies for total protein determination have been developed since the 19th century, relying on methods from fields, such as biochemistry, biology, and proteomics. Most of these have been developed to address research needs and not necessarily to determine the purity and/or adulteration of food proteins. Alternative methods for protein determination in food have been reviewed extensively (Cole 1969; Weaver and others 1977; Ribadeau-Dumas and Grappin 1989; Chang 1998; Kołakowski 2001, 2005; Owusu-Apenten 2002a; Wrolstad and others 2005). In most cases, the method's selectivity for protein and its potential effectiveness in preventing adulteration have not been discussed. Few reports have addressed the reasons why total nitrogen based test procedures continue to be the standard, and few have addressed obstacles to moving past these 19th-century methods. Lastly, reported methodologies for adulteration in general emphasize the detection of specific adulterants rather than the assessment of identity and purity (Burgoon and others 1985; Cordella and others 2002; Lees 2003; Reid and others 2006; Sun 2008).

To stimulate discussion and to provide information about the development and adoption of new or alternative protein methodologies, in this overview the authors review:

  • • the early history of food protein methodology
  • • analytical strategies to prevent intentional adulteration of foods and food ingredients
  • • challenges of developing or adopting new or alternative protein quantification methods and associated reference materials
  • • criteria against which new methodologies can be evaluated, and
  • • emerging methodologies for total food protein measurement, including pros and cons.

It should be noted that the scope of this review does not necessarily include methods to detect specific nonprotein nitrogen compounds in food and feed ingredients unless such methods are coincidental to the 5 points above.

Early History of Food Protein Research, Analysis, and Valuation

Early history of food protein research and analytical methodology

Scientists in the 18th century were curious about how vegetable foods are converted into animal substances, which led them to investigate commonalities between these substances and eventually to discover protein (Rosenfeld 2003). The characteristics of animal-sourced substances, such as glue, gelatin, and egg white have been described for thousands of years. Analogous physical properties for a plant-based glutinous substance (now known as gluten) and for certain animal substances were first reported by Beccari in 1746 (Osborne 1908). Animal substances that coagulated upon heating, such as egg white, were defined as early as 1777 by the term albuminous (McCollum 1957a). Fourcroy in 1789 recognized that similar albuminous substances in plants and animals were a distinct class of biological molecules (Fourcroy 1789; McCollum and others 1939a). Mulder is credited with coining the term protein in 1838 (McCollum and others 1939b), following a suggestion by Berzelius (Vickery 1950; Brouwer 1952; Rosenfeld 2003).

During the 18th century, developments in the analysis of albuminous proteins correspond with progress in analytical chemistry, when Lavoisier developed combustion methods to analyze the organic chemical composition of substances (Szabadváry 1966). In 1786 Berthollet, using a distillation procedure, first reported that nitrogen was a constant constituent of animal extracts (Berthollet 1786; McCollum and others 1939a; Szabadváry 1966). Based on Lavoisier's procedures for combustion analysis, Gay-Lussac and Thénard developed the 1st combustion method for nitrogen determination and determined the nitrogen (azote) content of egg white protein in 1810 (Davy 1815; Szabadváry 1966).

From the early to middle 19th century, 2 types of food analysis methods involving protein were developed by those interested in the composition of food and the role of nitrogenous compounds in nutrition. Mulder and his contemporaries agreed that nitrogenous protein substances in the diet played an important role in nutrition (McCollum and others 1939a; Brouwer 1952; McCollum 1957b). Led by Liebig and Boussingault in the 1830s, this thinking led to food analyses based solely on nitrogen content. Liebig in particular asserted that the nutritional value of food was based solely on its “plastic”“body-forming” protein content, which could be assessed by measuring the total nitrogen content of a food, ignoring other constituents and analytical methods (McCollum and others 1939a; McCollum 1957b, 1957c; McCosh 1984; Rosenfeld 2003). Liebig measured the nitrogen content of foods using combustion methods based on his improvements on methods discovered by Lavoisier, Gay-Lussac, and Thénard's, and Boussingault used improved combustion methods developed by Dumas (McCollum 1957b; Szabadváry 1966; Rosenfeld 2003).

A 2nd approach for the analysis of food composition was pioneered by Einhof, Vogel, Gorham, Hermbstaedt, and others beginning around 1800. This approach incorporated methods to separate and quantify constituents, such as proteins, fat, starch, and others from cereal grains (Osborne 1908; McCollum and others 1939a; McCollum 1957c). On the basis of solubility properties, individual types or fractions of proteins were separated from other plant constituents (Osborne 1908; McCollum and others 1939a; McCollum 1957c). This approach is consistent with current understanding of nutrition and food compositional analyses but did not gain acceptance until later due to the authority then given to Liebig's theories. McCollum, in his review of nutrition history, suggests that the dominance of Liebig's theory, even though it turned out to be incorrect, promoted the nutritional analysis of food based solely on nitrogen content (1st approach) and stifled further development of methods initiated by Einhof and others until the late 19th century (McCollum 1957c).

In the middle to late 19th century, a number of significant developments in total nitrogen content methods for food protein analysis included early versions of the methods still used today. The 1st quantitative combustion method using total nitrogen to measure proteins in foods, albeit unreliably, is credited to Dumas (Dumas 1831; Szabadváry 1966; Rosenfeld 2003). More dependable combustion procedures for nitrogen were improvements on the Dumas method and were reported by Shiff, and Varrentrapp and Will (the Soda-lime Process) from 1841 to 1868. These never enjoyed widespread popularity because a much simpler and more reliable wet chemistry method for total nitrogen determination was developed shortly thereafter by Kjeldahl (Szabadváry 1966).

Although it is not clear who first reported the use of a total nitrogen-to-protein conversion factor to quantify the total (crude) protein of food, the analytical methods developed by Henneberg and Stohmann in 1864 at the Weende Experimental Station in Germany used a factor of 6.25 (Henneberg 1865; Atwater and Woods 1896; McCollum and Simmonds 1929). This factor was based on the assumptions that protein consistently contained 16% nitrogen and that all nitrogen in food was from protein (McCollum and Simmonds 1929; Jones 1931/41; Koivistoinen 1996; Salo-Väänänen and Koivistoinen 1996). Early 19th-century protein nitrogen studies such as those conducted by Mulder and Gay-Lussac supported this factor, but the combustion methods used at that time were not necessarily reliable or accurate (Brouwer 1952; Szabadváry 1966).

McCollum reports that in the late 19th century analysts recognized that nonprotein nitrogen substances were present in food and that the use of total nitrogen measurements to quantify total protein were therefore inadequate (McCollum 1957c). To overcome this analytical issue, Wagner, Sestini, Kellner, Dehmel, and Schulze from 1878 to 1879 reported methods that putatively precipitated “true protein” and then quantified this fraction by total nitrogen measurements (Sestini 1878; Wagner 1878; Dehmel 1880; Kellner 1879; Schulze 1879; McCollum 1957c). These methods were based on the hypothesis that measuring this “true protein” fraction would provide a more accurate estimate of the nutritional value of nitrogenous substances in food (McCollum 1957c). This notion was contested by Weiske and others who in 1879, demonstrated in rabbit studies that a nonprotein, nitrogen-containing substance, asparagine, functioned as a nutritional substitute for protein (McCollum 1957c; Weiske and others 1879). This resulted in decreased interest in efforts to advance the chemical analysis of food by separating protein from nonprotein components (McCollum 1957c). Indeed, only one Official AOAC Method of Analysis using a similar method could be found from this period until the late 20th century. This Official Method is based on one reported by Van Slyke in 1893 to measure the casein content of milk by precipitating the casein with acetic acid and separating it from the supernatant before performing total nitrogen analysis (Van Slyke 1893; Wiley 1893; AOAC 1927).

In 1883 Kjeldahl developed a wet chemistry method for nitrogen analysis. It was faster, simpler, and more accurate than the previously established combustion procedures (Kjeldahl 1883; Dyer 1895; Szabadváry 1966). Kjeldahl's nitrogen determination method became pivotal in food and agricultural chemistry and became the authoritative reference method for total (crude) protein quantification in foods (Lynch and Barbano 1999). At least one authority suggests that the Kjeldahl method has been the subject of more studies than any other in analytical chemistry (Chen and others 1988).

Extensive work in the early 20th century at USDA's Protein and Nutrition Research Division improved assessment methods for protein quality and quantity (Jones 1931/41). These studies included isolation of individual proteins and amino acid analysis for different food ingredients, but such approaches were not then considered practical for quantitative protein analysis (Jones 1931/41). A major contribution of this group was improved measurement accuracy based on the development of nitrogen-to-protein conversion factors for specific food categories and food ingredients (Jones 1931/41).

In summary, the nutritional theories of the 19th century and analytical technological capabilities of the 20th century had a significant role in the evolution of food protein measurement. These conditions inhibited the development of analytical strategies to separate protein from other food matrix substances before analysis and promoted determinations based on total nitrogen to calculate total protein content. This assertion is supported by the failure of 3 attempts from the scientific community to develop and advance the former type of methods. The 1st was led by Einhof in the early 19th century to analyze the protein composition of grains, but this was stifled by Liebig's dominating nutrition theory. The 2nd effort was led by Wagner and others in 1879 to develop methods to precipitate and quantify the “true protein” contents of foods, but these approaches did not advance because of widespread acceptance of Weiske's nutritional theory. Finally, early 20th-century USDA scientists concluded that analytical technology of that time was not capable of efficiently separating and analyzing individual proteins. Instead, they promoted protein measurement by specific nitrogen-to-protein conversion factors. Unfortunately, protein-specific measurement methods were not accepted. Because of advances made in protein nutrition and analytical sciences since the 19th century, coupled with the aforementioned incidents of adulteration, a 4th attempt is now warranted.

Protein measurement for food valuation

For purposes of trade, the first occurrences of basing the market value of protein-based food ingredients on total protein content measurement is unknown, but the need for reliable methods dates to the late 19th century. In 1883 Kjeldahl justified his new method based on total nitrogen determination by noting the need for reliable tools to evaluate the protein contents of incoming barley ingredients used for brewing (Kjeldahl 1883). This challenge—developing reliable analytical methods that facilitated trade by providing reproducible and accurate analytical results for both buyers and sellers—was one of the principal reasons for the formation of the Association of Official Agricultural Chemists (AOAC, now known as AOAC International) in the late 19th century (Helrich 1984).

Analytical Strategies to Prevent Economic Adulteration

Economic adulteration of food is the fraudulent substitution of an authentic component of a food with a cheaper and nonauthentic component for economic gain (Smith and Bonwick 2007). At least 2 general analytical strategies to detect food adulteration have evolved. The 1st approach uses analytical tests to identify one or more suspected adulterants, for example, analytical methods to detect hard wheat flour in durum wheat flour (Alary and others 2002). In this case an absence of result indicates that the test material is not adulterated with a specific material, but it requires a priori knowledge about the adulterant and is not useful for detecting unknown adulterants. This 1st approach typically is used for monitoring specific suspected adulterants, especially at low levels. In the case of food protein, reliable methods exist to detect specific adulterants such as melamine in different matrices, but this approach cannot prevent future adulteration with unknown protein substitutes, either those containing nitrogen or those not.

The 2nd approach is based on compendial identification and purity tests that substantiate the ingredient's identity and quantify its purity (Abernethy and others 2008). An example is the Food Chemicals Codex monograph for Sucralose, which specifies a procedure based on infrared spectroscopy with an associated chemical reference standard to substantiate identity and a procedure based on high-performance liquid chromatography (HPLC) and reference standard to quantify the ingredient's purity and to substantiate identity (US Phamacopeial Convention 2008). This approach is effective when either a known or unknown adulterant is substituted for the original material at concentrations sufficiently high to be recognized by perturbations in the test results. When adulterants are present in low concentrations such assessments may be less useful. For example, using purity assessment to detect the presence of an adulterant at 900 ppm, analysts must be able to reliably detect the difference between 99.91 and 100.00% purity, a goal that may be beyond the sensitivity and specificity of the procedure. However, from a practical perspective, in many cases counterfeiters must adulterate at high relative concentration levels to realize economic gain, suggesting that this approach remains a useful tool for reducing the risk of adulteration. In the case of food protein, no current compendial methods are sufficiently selective to differentiate protein from other nitrogen-containing compounds, a limitation that fails to protect public health.

The Complexity of Food Protein Measurement and Its Consequences

The complexity of food protein measurement

Proteins in foods are an important component of their nutritional value and functional suitability. Protein quantification is therefore an important tool that is used throughout the global food supply chain, a complicated and interwoven web of valuation, quality and safety assurance, and regulatory activities that depend on total nitrogen based protein measurements.

Some reasons to measure proteins in food and food ingredients include: (1) establishing the value of protein-based commodities for trading purposes, (2) assessing the quality of protein-based food ingredients, (3) assessing the conformity of food ingredients to buyer/seller or regulatory specifications, (4) establishing the authenticity of protein-based food and food ingredients, and (5) analyzing the nutritional content of foods for labeling and human or animal diet formulation (Ribadeau-Dumas and Grappin 1989; Chang 1998; Owusu-Apenten 2002a; Lees 2003; DuPont and others 2005; Bonfatti and others 2008).

Several analytical solutions have been developed and adopted for total protein measurements depending on the locus in the supply chain where the measurement is made. Table 1 summarizes methods for total protein analysis for food and feed adopted by AOAC International, AACC International, American Oil Chemists’ Society (AOCS), the International Organization for Standardization (ISO), and the International Dairy Federation (IDF). Protein measurements for commodities that are traded in large volumes (such as cereal grains, their flours, fluid milk and its derivatives) are often made with rapid and, if possible, inline infrared spectroscopy methods that are not direct measurements of nitrogen or protein but are calibrated against nitrogen-based methods (Watson 1977; Biggs and others 1987; Barbano and Lynch 1989; Williams and Sobering 1993; American Association of Cereal Chemists 2000; AOAC International 2005; Barbano and Lynch 2006; U.S. Code of Federal Regulations 2009a). Other protein-based ingredients that are traded in smaller volumes are typically analyzed using traditional nitrogen-based methods such as Kjeldahl or combustion (Dumas) (Krotz and others 2008). In the United States, the protein contents of finished food products are analyzed for nutritional labeling using AOAC International total nitrogen methods (U.S. Code of Federal Regulations 2009b). For finished products, methods such as the protein digestibility-corrected amino acid score (PDCAAS) and/or amino acids analyses are used in conjunction with total nitrogen based methods (U.S. Code of Federal Regulations 2009b). Thus nitrogen-based protein methodologies currently are directly or indirectly linked to almost all protein measurement activities. Consequently, universal replacement of the nitrogen-based assay could have a widespread impact on the global protein measurement community.

Table 1–. Methods for total protein measurement in food and feed, by matrix.
Matrix categorySpecific matrixMethod typeMethod reference
  1. AOAC = AOAC International (2005), AOCS = American Oil Chemists' Society (2004), AACC = American Association of Cereal Chemists (2000), ISO = International Organization for Standardization, IDF = International Dairy Federation, NIR = near-infrared spectroscopy, Mid-IR = mid-infrared spectroscopy, TCA = trichloroacetic acid.

Animal feedsAnimal feed and pet foodCombustionAOAC 976.05; AOAC 968.06; AOAC 990.02; AOAC 990.03; AACC 46-30
Animal feed and pet foodKjeldahl and its variationsAOAC 954.01; AOAC 984.13; AOAC 988.05; ISO/IDF 5983-1:2005; ISO/IDF 5983-2:2009; AACC 46-11A; AACC 46-16
Animal feed, forage (plant tissue), grain, and oilseedsKjeldahl (block digestion)AOAC 2001.11
Feed and feedstuffsKjeldahl and its variationsAACC 46-09; AACC 46-10
Cereals, grains, flours, pulsesCereal grainsNIRAACC 39-10
Cereal grainsKjeldahl and its variationsAOAC 979.09; AACC 46-11A; AACC 46-16; AACC 46-16
Cereal grainsBiuretAACC 46-15; ISO 20483:2006
Cereal adjunctsKjeldahlAOAC 945.18; AACC 46-11A; AACC 46-16; AACC 46-16
Cereal grains and oilseedsCombustionAOAC 992.23; AACC 46-30
Wheat (whole grain)NIRAOAC 997.06; AACC 39-25
Flour (wheat)Kjeldahl and its variationsAOAC 920.87; AACC 46-11A; AACC 46-12; AACC 46-13; AACC 46-16; AACC 46-30
Flour (wheat)Dye-bindingAACC 46-14B
Flour (wheat)NIRAACC 39-11
Soy flourKjeldahlAOCS Bc 4-91
SoybeanKjeldahlAOCS Ac 4-91
SoybeanNIRAACC 39-20
Soybean mealCombustionAOCS Ba 4f-00
Soybean mealNIRAACC 39-20
PulsesKjeldahlISO 20483:2006
MeatsMeatKjeldahl (block digestion)AOAC 981.10
Meat and meat productsNIRAOAC 2007.04
Meat and meat products, including pet foodsCombustionAOAC 992.15
MilkMilk (fluid)Dye-bindingAOAC 967.12; AOAC 975.17; ISO 5542:1984
Milk (fluid)Mid-IRAOAC 972.16; ISO 9622:1999
Milk (fluid)Kjeldahl and its variationsAOAC 991.20; ISO-IDF 8968-1/20-1:2001; ISO-IDF 8968-2/20-2:2001; ISO-IDF 8968-3/20-3:2004
Milk (fluid)TCA ppt Protein by KjeldahlAOAC 991.22; AOAC 991.23; ISO-IDF 8968-5/20-5:2001
Milk (dried)KjeldahlAOAC 930.29
Milk and milk productsCombustionISO-IDF 14891/185:2002
CreamKjeldahlAOAC 920.109
OtherBaked productsKjeldahlAOAC 935.39
BeerKjeldahlAOAC 920.53
Beer, wort, and brewing grainsCombustionAOAC 997.09
BreadKjeldahlAOAC 950.36
Brewing sugars and syrupsKjeldahlAOAC 945.23
CacaoKjeldahlAOAC 970.22
Caseins and caseinatesKjeldahlISO 5549:1978
CheeseKjeldahlAOAC 920.123; AOAC 2001.14; ISO/TS 17837: 2008; IDF/RM 025:2008
Evaporated milk (unsweetened)KjeldahlAOAC 945.48
ForagesNIRAOAC 989.03
Fruit productsKjeldahlAOAC 920.152
GelatinKjeldahlAOAC 935.46
Ice cream and frozen dessertsKjeldahl or Dye-bindingAOAC 930.33
Infant formula (milk-based)KjeldahlAOAC 986.25
Laboratory maltKjeldahlAOAC 950.09
Macaroni productsKjeldahlAOAC 930.25
Milk chocolateKjeldahlAOAC 939.02
Nuts and nut productsKjeldahlAOAC 950.48
Oilseed byproductsKjeldahlAOCS Ba 4d-90
Oilseed byproductsCombustionAOCS Ba 4e-93
Oilseeds and animal feedCombustionISO 16634-1:2008
PeanutKjeldahlAOCS Ab 4-91
PlantsKjeldahl and its variationsAOAC 977.02; AOAC 978.04
Starch dessert powdersKjeldahlAOAC 945.56
Sunflower seedKjeldahlAOCS Ai 4-91
Sweetened condensed milkKjeldahlAOAC 920.115
TeaKjeldahlAOAC 920.103
WinesKjeldahlAOAC 920.70
YeastKjeldahlAOAC 962.10; AACC 46-11A; AACC 46-16

The value of food protein

Food proteins provide nitrogen primarily in the form of essential amino acids. Besides providing nutritional value, these proteins also contribute to the functional properties of food by imparting texture and flavor, which are organoleptic properties that influence human interest in purchasing and consuming foods.

Economics of food protein measurement

The market value of protein-based food ingredients such as cereal grains and dairy ingredients is determined, in part, by their total protein content (Barbano and Lynch 1990; Wrigley 1994; Owusu-Apenten 2002b). In 1988 an estimated 400 million metric tons of food protein were produced globally (Owusu-Apenten 2002b), so analytical measurements of food protein determine the allocation of large sums of money (Barbano and Lynch 2006). Koletzko and Shamir (2006) estimated that, based on Codex Alimentarius discussions, changing the nitrogen-to-protein conversion factor from 6.38 to 6.25 for all dairy products would result in a $94 million loss in Europe. The nutritional formulation of animal diets, including that of animals used for human food, is another component of protein economics. Overuse of protein in animal diets may be not only harmful to animals but also costly for feed producers, including concurrent environmental effects following increased nitrogen elimination in feces (Mosse 1990). Another economic consideration is the cost of making analytical measurements of food protein. The cost and frequency of protein measurement eventually are factored into the price of the food and are passed on to consumers (Kennedy 2008).

All of these factors point to the need for reliable, rapid, and inexpensive quantitative protein methods capable of producing comparable results for both buyers and sellers of food ingredients.

Protein measurement and public health

Protein measurement affects public health in terms of both nutrition and food safety. Severe protein deficiency in infants and young children can lead to kwashiorkor and marasmus and in less severe forms may hinder physical and mental development. In many parts of the world, protein measurements assess and ensure the nutritional value of food. With respect to safety, the recent incidents of melamine adulteration clearly demonstrated how the lack of appropriate protein measurement methods combined with misguided economic incentives can have devastating consequences (Ingelfinger 2008). As long as protein determination remains nonselective, consumers and industry will run the risk of further food adulteration, as well a risk for confusion in valuation.

Need for new analytical approaches

Both the economic and public health consequences of protein measurement demonstrate the need for high-level analytical performance. Current total nitrogen based methods do not achieve the level of selectivity and specificity needed for total protein measurement, and alternative analytical methods with improved selectivity must be developed and adopted. These new methods must also maintain unusually high levels of accuracy (trueness) and precision necessary for the analysis of food ingredients and commodities (Barbano and Lynch 2006).

Analytical Challenges of Developing New, More Selective Protein Quantification Methods

Defining total protein

The definition of total protein highlights an analytical challenge: Proteins are difficult to define biochemically because they can contain a wide variety of amino acids that form the primary structures of proteins. Secondary and tertiary structure can differ depending on the protein's environment or the methods of purification used. Posttranslation modification (such as glycosylation) can further change the characteristics of the protein. In addition, free amino acids and small peptides perhaps could be included in the definition of a protein. Several definitions for protein have been developed depending on the scientific field and reason for measurement (Mariotti and others 2008).

In food sciences, measurements are made to assess ingredient quality and value, and for nutrition labeling, so total protein is not defined biochemically but rather by the method used to make the measurements. For dairy products, total protein is defined empirically as Kjeldahl nitrogen times the conversion factor of 6.38, and in most cases Kjeldahl is the reference method against which other methods are calibrated (Lynch and Barbano 1999). Researchers have tried to determine the appropriate conversion factors for various foods, but the accuracy of measurements made using this approach continues to be questioned. (Jones 1931/41; Merrill and Watt 1973; Salo-Väänänen and Koivistoinen 1996; Lynch and Barbano 1999; International Dairy Federation 2006; Mariotti and others 2008). Variations or inaccuracies associated with total nitrogen measurements for protein quantification arise from observations that: (1) variability of nitrogen recovery occurs during analysis (Simonne and others 1997), (2) foods contain both proteinaceous nitrogen as well as other inorganic and organic nitrogen-containing compounds, and (3) both the nitrogen content and the nitrogen-to-protein conversion factor may vary even within the same protein source (Salo-Väänänen and Koivistoinen 1996; Lynch and Barbano 1999; Mariotti and others 2008).

For the nutritional sciences, proteins are most commonly defined by their amino acid composition (Mariotti and others 2008). Methods such as PDCAAS (a measure of the bioavailability of nutritionally essential amino acids from a given protein source in comparison to a reference protein) are used for nutrition labeling (Schaafsma 2000). This calculation, too, includes the total protein contents measured by the Kjeldahl method and is subject to the same limitations as other nonspecific approaches.

One potential revised definition for total protein would be the summation of all the peptide-bound amino acids in a food after the protein has been sufficiently isolated from the food to accurately measure the amino acid content, in a fashion similar to the methodology applied for total fat in foods (AOAC International 2005).

The ambiguity associated with the total protein assay complicates comparisons of the analytical accuracy of alternative methods. One option is the use of spike-recovery results based on purified proteins taken from the same source as those that will be tested. Still, in order to evaluate accuracy, the results of most newly developed quantitative protein analysis methods are compared to those from the Kjeldahl method. Several authors have pointed out that total nitrogen based methods for total protein assessment have limited accuracy. Thus, the usefulness of such comparisons is questionable (Salo-Väänänen and Koivistoinen 1996; Mariotti and others 2008). The Kjeldahl method is considered authoritative from regulatory and commercial perspectives. If newly developed and/or adopted methods do not produce the same results as Kjeldahl, a substantial effort would be required to once more achieve uniformity in measurement, unless the new methods were universally adopted or at least procedures using the Kjeldahl and new methods were performed simultaneously.

Variability of protein in food ingredients

Many total protein assays have a variable analytical response per unit of protein, depending on the type of protein. For example, in the Bradford dye binding assay the relative color response per unit of protein for bovine α-chymotrypsinogen is about one-half that of bovine albumin (Krohn 2005). These differences are attributed to the compositional and physicochemical variations of different proteins.

Proteins vary in biochemical and physicochemical properties because of both genetic variations across species and metabolic response to the environment within species. For example, a recent review of cow milk protein nomenclature describes variations in the molecular weight, amino acid sequence, and varying degrees of glycosylation or phosphorylation (Farrell and others 2004). All these variations, as well as the differences in secondary structure and isoelectric point, may influence a protein method's detection sensitivity (Krohn 2005). Another example includes the altered color response for the biuret protein assay associated with variation in the imino acid (proline and hydroxyproline) composition of gelatin (Zhou and Regenstein 2006). Further, protein glycosylation interferes with measurement by many commonly used total protein methods (Fountoulakis and others 1992; Noble and others 2007).

Protein processing may influence protein measurement. For example, heat processing of milk and soy proteins alters their solubility and thus their measurement (Gotham and others 1988; Koppelman and others 2004). Other common processing-induced changes include variations in molecular weight distribution following hydrolysis, racemization and/or oxidation of amino acids, protein cross-linking, amine–carbonyl reactions such as the Maillard reaction, and sulfitolysis (Damodaran 1996; Horneffer and others 2007).

These sources of variability must be understood and taken into account when analysts develop and validate total protein quantification methods that use reference materials. In some cases, analysts must develop and validate process-specific methods and reference materials. In other cases (finished food products, ingredients with highly variable protein composition), it may not be practical to develop reliable selective and specific total protein methods.

Food matrix

Food matrix components can influence both the accuracy and the precision of protein measurement. As a result, most reliable measurements rely on the validation of matrix-specific adaptations (Lovet 1997; Ellis and others 1998). Despite their lack of selectivity for protein, total nitrogen protein quantification methods have minimal matrix effects due to the fact that as the nitrogen is analyzed, and the food matrix is almost completely oxidized during analysis. This allows nitrogen-based protein methods to be applied to almost any food matrix with minimal matrix-specific adaptation. The use of more selective methods for protein measurements introduces the possibility of more matrix interferences.

Biological and process-induced variability also can introduce variability in the composition of the food matrix. These variations must be defined when analysts develop new protein test methods and associated reference materials, and new methods must be validated in each matrix for which their use is proposed. Food ingredients and products that are highly variable may preclude the development of matrix-specific methods. Efforts should focus on ingredient matrices that have a relatively high concentration of protein and low degree of matrix variability. Examples of standardized ingredients in which protein content plays an important role in determining market value are shown in Table 2. Development of reliable total protein quantification methods for these ingredients may require the development and validation of matrix-specific methods and may be most useful with associated reference materials.

Table 2–. Standardized ingredients for which protein content is a factor for determining market value.
Food ingredientFood ingredient (continued)
  1. aIgoe and Hui 2001.

  2. bU.S. Phamacopeial Convention 2008.

  3. cU.S. Food and Drug Administration 2001.

  4. dRodger 2001.

Fluid milkaWheat glutenb
Dry milkaWheat protein isolateb
CaseinbSoy protein concentrate and isolateb
Sodium caseinatebFlour (various grains)
Calcium caseinatebHydrolyzed vegetable proteinb
WheybPartially hydrolyzed proteinsb
Whey protein concentratebAlbumin (egg-derived)a
Whey protein isolatebDried eggs
Demineralized wheybCollagenb
Delactosed wheybGelatinb
Milk-derived lactoferrincFish protein concentrate and isolatea
LactalbuminaYeast extractb
Corn glutenaAutolyzed yeastb
ZeinbDried yeastsb

Characteristics of Improved Methods for Total Protein Measurement

Desirable characteristics for methods to replace total nitrogen based approaches must be identified. Sections “Definition of analyte” to “Safety” present desirable characteristics with discussion.

Definition of analyte

The essential value of food protein is its ability to provide dietary amino acids (Mariotti and others 2008). Secondary concerns are the physicochemical properties of protein-bound amino acids that provide quality attributes. Posttranslational modifications of proteins may be essential for human cellular physiology, but these attributes typically are not factors in human nutrition. An ideal definition of protein for a quantification method should therefore be peptide-bound amino acids.


A method that shows a high degree of selectivity for peptide-bound amino acids can help prevent adulteration of protein. The need for selectivity must be balanced against the potential downside of excessive selectivity. Procedures that rely on methods that yield overly selective results could differentiate proteins of equal nutritional value and functional quality that differ because of glycosylation, methylation, or other such modification, genetic heterogeneity, and chemical reactions during processing.

Accuracy (trueness)

A suitable method should have sufficient accuracy that can be measured using spike-recovery experiments with highly purified protein samples. When sample extraction or precipitation procedures are used for sample preparation, protein recovery must be established. Another approach is comparison of standard to sample for quantitative total amino acid content.


A suitable method should have equal or better precision compared to current total nitrogen based approaches. Table 3 summarizes the reported precision for current Kjeldahl- and combustion (Dumas)-based total nitrogen methods.

Table 3–. Precision of total nitrogen based methods for protein quantification, by matrix.
MatrixMethodRSDr (%)RSDR (%)0AOAC method referenceaOther references
  1. aAOAC International 2005.

  2. RSDr= intralaboratory relative standard deviation, RSDR= interlaboratory relative standard deviation.

Animal FeedCombustion0.591.10990.03Sweeney 1989
BarleyCombustion2.133.15992.23Bicsak 1993
BeerCombustion1.6 to 8.34.6 to 8.2997.09 
CanolaCombustion0.871.79992.23Bicsak 1993
Casein and caseinateKjeldahl0.34 to 0.591.16 to 1.33 Wiles and others 1998
Casein and caseinateCombustion0.30 to 0.350.62 to 0.64 Wiles and others 1998
Cheese (various)Kjeldahl0.21 to 0.830.39 to 1.142001.14 
Chinchilla foodKjeldahl0.890.992000.11 
CornCombustion1.152.88992.23Bicsak 1993
Corn silageKjeldahl1.642.162000.11 
Dog foodKjeldahl0.870.912000.11 
Fish mealKjeldahl0.73 to 1.620.98 to 2.37 Miller and others 2007
Fish mealCombustion1.482.01 Miller and others 2007
Grass hayKjeldahl1.941.942000.11 
Infant formulaKjeldahl1.74.1NAWiles and others 1998
Infant formulaCombustion2.07.3NAWiles and others 1998
Legume hayKjeldahl1.451.452000.11 
MeatCombustion0.60 to 2.231.32 to 3.35992.15King-Brink and Sebranek 1993
MeatKjeldahl0.82 to 2.411.59 to 2.84 King-Brink and Sebranek 1993
Meat and bone mealKjeldahl1.901.902000.11 
Milk (fluid)Kjeldahl0.384 to 0.510.420 to 1.02991.20Grappin and Horwitz 1988
Milk (fluid)Kjeldahl0.651.9 Wiles and others 1998
Milk (fluid)Combustion2.07.3 Wiles and others 1998
Milk (fluid)Kjeldahl-based TCA precipitated protein nitrogen content0.2850.702991.22 
Milk powder (whole and skim)Kjeldahl0.74 to 1.72.29 to 2.52 Wiles and others 1998
Milk powder (whole and skim)Combustion0.33 to 0.490.82 to 0.87 Wiles and others 1998
Milk replacerKjeldahl1.391.392000.11 
Protein blockKjeldahl0.450.762000.11 
SorghumCombustion2.572.84992.23Bicsak 1993
SoybeansCombustion0.771.24992.23Bicsak 1993
Spent grainCombustion5.56.7997.09 
SunflowerCombustion2.002.94992.23Bicsak 1993
Sunflower seedsKjeldahl2.382.382000.11 
Swine pelletKjeldahl0.470.602000.11 
WheatCombustion0.991.74992.23Bicsak 1993
Whey protein concentrateKjeldahl0.591.56 Wiles and others 1998
Whey protein concentrateCombustion0.310.74 Wiles and others 1998
WortCombustion1.4 to 1.64.1 to 4.3997.09 

Limit of detection

The limit of detection is not a critical attribute because these methods will be used at a macrolevel to quantify the purity of high protein content ingredients.

Reference materials

A suitable method should use matrix-specific purified protein reference standards or method calibrants that are suitable for the method and matrix of the protein-containing food.


A suitable method should be simpler than the Kjeldahl method, should involve minimal sample preparation, and should use equipment that is generally available in food control laboratories.


A suitable method should be rapid and adaptable to high-throughput analysis with average time per sample of less than 3 to 6 min (Simonne and others 1997; Chang 1998; Owusu-Apenten 2002b).


The analytical cost per sample should be less than or equal to that of nitrogen-based methods ($0.37 to $1.00 per sample) with minimal associated equipment costs (Owusu-Apenten 2002b).

Compatibility with food matrices

The method should be validated for individual food ingredient matrices that may contain a mixture of proteins, but matrices should display a suitable degree of variability in terms of both protein and matrix.


The analytical method should not introduce safety hazards such as concentrated acids or heavy metal catalysts.

Review of Currently Available Methods for Total Protein Quantification

Quantitative test methods for food protein have been extensively reviewed but generally without detailed consideration of the method's characteristics that are important for preventing food protein adulteration (Sections “Definition of analyte”to“Safety”). The following sections address these issues and are focused on methods that have been developed and applied for food protein quantification.

Nitrogen-based methods

Current Kjeldahl method and its variants

Background Kjeldahl reported his wet chemistry method in 1883 as a new means of indirectly quantifying total (crude) food protein by measuring the total organic nitrogen content (Kjeldahl 1883). The method rapidly gained acceptance because it overcame many of the limitations—notably accuracy, simplicity, and speed—of nitrogen determination by combustion analysis (Dyer 1895; Szabadváry 1966). Since its inception the Kjeldahl method has been the most widely accepted method for total protein determinations in feeds, foods, and food ingredients. It first appeared as an official AOAC method in 1887 for milk protein determination (Richardson 1887). It has also been subject to many modifications and improvements (Bradstreet 1965).


The Kjeldahl method indirectly quantifies the total protein content of foods by direct nitrogen measurement and subsequent multiplication by a conversion factor. During manual Kjeldahl analyses, total organic nitrogen content is analyzed in 4 basic steps: wet digestion of the sample to convert nitrogen to ammonium sulfate, neutralization with alkali to convert the ammonium sulfate to free ammonia, distillation of the ammonia into boric acid, and back titration of excess boric acid with standardized alkali. More details on the chemistry of these steps can be found elsewhere (Bradstreet 1965; Chang 1998; Owusu-Apenten 2002b; Rhee 2005). Reported adaptations of this method include replacement of the original mercury catalysts, a microversion that incorporates reduced sample size and reduced amounts of reagents, semiautomated and automated versions, and colorimetric methods for quantifying the distilled ammonia (Simonne and others 1997; Owusu-Apenten 2002b). The generic nitrogen-to-protein conversion factor has been 6.25 since the 19th century, and additional conversion factors have since been developed for specific protein matrices, as reviewed by Mariotti and others (2008). The Kjeldahl method has been validated and standardized for total (crude) protein estimation for a wide variety of food matrices and has been adopted by AOAC International, AACC International, ISO, and AOCS (see Table 1).

Simplicity, cost, throughput, applicability

The manual Kjeldahl method is not a simple method because of the number of wet chemistry steps involved in the analysis. The use of concentrated sulfuric acid in the sample digestion step presents safety concerns. The elapsed analysis time for the manual method is approximately 3 h, but automated versions reduce this to 12 min with throughput of one sample every 3 min (Owusu-Apenten 2002b). Reagent cost per sample has been estimated to be $1.00 with a capital equipment cost of approximately $6000 (Owusu-Apenten 2002b).

Specificity, interferences, and susceptibility to adulteration

No major interferences have been reported for the Kjeldahl method when it is applied to the intended, unadulterated matrix because most organic matrix components are oxidized during digestion.

In terms of selectivity, the Kjeldahl method generally is specific for total organic nitrogen (Chang 1998). Organic nitrogen compounds in food include protein-bound amino nitrogen, free amino nitrogen, nucleic acids, phospholipids, and nitrogenous glycosides (Mariotti and others 2008). The specificity of the Kjeldahl method for total organic nitrogen has been challenged. Although the recovery of amino nitrogen is thought to be quantitative, the recovery of nitrogen from the other sources may not be quantitative (Jung and others 2003; Miller and others 2007; Mariotti and others 2008). In addition, Simonne and others (1997) reported that some sources of inorganic nitrogen may be partially or fully recovered during analysis. The Kjeldahl method, therefore, is not specific for protein nitrogen and also lacks a high degree of selectivity for organic nitrogen. Therefore, this method produces unacceptable results for samples adulterated with nonprotein nitrogen.


In addition to the method's lack of specificity for total organic nitrogen, other factors can introduce error into Kjeldahl measurements, in part because of the number of steps involved. These factors include the accuracy of sample weighing, contaminated reagents or glassware, incorrect reagent blank compensation, lack of recovery of nitrogen during digestion, the accuracy of the normality of the boric acid, and the alkali used for the final titration (Chen and others 1988; Barbano and Lynch 1990; Lynch and Barbano 1999). The use of primary standards, such as ammonium p-toluenesulfonate (PTSA), glycine PTSA, and nicotinic acid PTSA to evaluate nitrogen recovery during digestion and the standardization of the titrants have been used to control for these sources of error (Tellefson 1980; Chen and others 1988; Ribadeau-Dumas and Grappin 1989; Lynch and Barbano 1999). The use of reference materials with certified nitrogen content for milk protein analysis presents another approach to improve accuracy (Ribadeau-Dumas and Grappin 1989; Wiles and Gray 1996; Staniszewska and others 2008).

The accuracy of the Kjeldahl method also directly depends on the accuracy of the nitrogen-to-protein conversion factor applied, which is a subject of considerable debate (Ribadeau-Dumas and Grappin 1989; Mosse 1990; Salo-Väänänen and Koivistoinen 1996; Lynch and Barbano 1999; Mariotti and others 2008). The presence of nonprotein nitrogen-containing compounds and the variability of the nitrogen content within the same protein source may influence the analytical result unless accurate conversion factors are determined (Salo-Väänänen and Koivistoinen 1996; Lynch and Barbano 1999; Mariotti and others 2008). For example, mushrooms have high and variable nonprotein nitrogen content from chitin and chitosan compounds, and use of Kjeldahl analysis with the 6.25 conversation factor has been shown to overestimate mushroom protein content by as much as 23% compared to amino acid analysis (Weaver and others 1977).


Table 3 provides details of reported precision for nitrogen-based protein methods, including Kjeldahl, by food matrix. Although the Kjeldahl method has been standardized and published for a number of matrices, many of the AOAC Offical Methods were published many years ago before inter- and intralaboratory relative standard deviation calculations were routinely reported following method validation.


The primary advantages of the Kjeldahl method for food protein analysis are its compatibility with nearly all food matrices, its lack of food matrix interferences for unadulterated samples, its high level of precision, its wide acceptance in the food analytical and regulatory communities, and its options for automation. The primary disadvantages are its lack of selectivity for protein and issues with accuracy attributed to nitrogen-to-protein conversion factors and nitrogen recovery.

True proteinusing Kjeldahl nitrogen analysis

Background The Kjeldahl method measures not only protein amino nitrogen but also other forms of organic and inorganic nitrogen, as has been recognized since the late 19th century as a weakness (McCollum 1957c). Consequently, in 1879 some analysts reported methods to measure true protein by total nitrogen after the protein was precipitated and separated from nonprotein nitrogen substances, but this approach did not gain widespread use (see the section on “Early history of food protein research and analytical methodology”).

Further developments were not reported until over 100 ys later in the 1990s. At this time, the dairy industry was interested in changing the milk pricing system from a weight-and-fat-content system to a multicomponent system that included protein (Barbano and Lynch 1992). This was designed to be a better system for both milk producers and buyers because it would be capable of aligning the value of milk components in consumer dairy products with identical milk components in the raw milk supply—and deterring adulteration of milk with water (Brog 1969, 1970). Milk analyses using total nitrogen based methods were not adequate for use in this system because the variable content of nonprotein nitrogen in milk introduced significant error (Barbano and Lynch 1991). This led to the development of a true protein method for fluid milk (Barbano and Lynch 1991; Kuzyk and others 1994). The method was adopted as final in 1994 as an AOAC International method (991.22) and also IDF-ISO-AOAC method (Barbano and Lynch 1991; AOAC International 2005).


Protein is precipitated from the matrix by addition of trichloroacetic acid (TCA). The precipitated protein is isolated by filtration and is analyzed for total organic nitrogen content using the Kjeldahl method. The nitrogen content of the precipitate is multiplied by a 6.38 nitrogen-to-protein conversion factor to yield total TCA-precipitated protein.

Simplicity, cost, throughput, and applicability

The precipitation and filtration in this analysis require several manual steps that add to the time and skill required. The use of TCA and filter paper is the only additional reagent cost compared to the standard Kjeldahl method. This method has been validated only for fluid milk applications in AOAC International (2005).

Selectivity, interferences, and susceptibility to adulteration

AOAC International 991.22 was designed to be specific for TCA-precipitable protein nitrogen from milk. No reports are available on the evaluation of the selectivity of this method, but it should be more selective for protein nitrogen than the basic Kjeldahl method because of the separation of the protein from other components of the milk that may contain nonprotein nitrogen. With respect to potential for adulteration, this method is vulnerable to the addition of nitrogen-containing compounds that precipitate with TCA or complex with the precipitated protein itself. It is also vulnerable to adulteration by replacement of authentic protein with a protein from another source (nonauthentic protein) that is cheaper or higher in nitrogen content.


Intra- and interlaboratory relative standard deviations of 0.285% to 0.421% and 0.702% to 1.223%, respectively, have been reported for fluid milk analyses using this method (Kuzyk and others 1994; AOAC International 2005). These results are similar to those reported for normal Kjeldahl analysis for this matrix (see Table 3). An indirect variation of this method measures the total nitrogen and nonprotein nitrogen contents of milk separately using the Kjeldahl method but shows lower precision (Barbano and Lynch 1991).


The TCA precipitation method for fluid milk is a simple adaptation of the existing Kjeldahl method that improves selectivity for protein nitrogen. It has been validated only for the fluid milk matrix, so application to other matrices will require development and validation.

Combustion (Dumas) method

Background Combustion procedures dating back to the 18th century were the first developed to measure the nitrogen content of foods. Dumas is credited with developing the most reliable combustion procedure of the time for estimating protein content by total nitrogen. His method relied on difficult volumetric measurements of nitrogen gas that, along with other challenges of combustion procedures, lost favor to wet chemistry procedures (Szabadváry 1966). Combustion procedures recaptured interest with the advent of reliable combustion nitrogen analyzers (CNAs) in the 1980s. Methods were subsequently developed and validated for numerous matrices using CNA-based methods.


Current combustion methods for nitrogen analysis are based on the following principle (Owusu-Apenten 2002b): Samples are combusted in an induction furnace at 950 to 1050 °C in the presence of oxygen to form water, CO2, SO2, NOx, and N2. CO2 and SO2 are removed, and NOx is reduced to N2. Total N2 is then measured using a thermal conductivity detector. Nitrogen-to-protein conversion factors are then used to calculate the total (crude) protein content of the sample.

Simplicity, cost, throughput, and applicability

Modern CNAs provide total nitrogen analysis that is simpler, faster, cheaper, and safer than Kjeldahl-based procedures. Sample preparation and technician involvement during analysis are minimal, and analysis does not involve wet chemistry or hazardous chemicals (Foster 1989). Automated equipment for this method is available, and the analysis time per sample has been reported to be as short as 3 min (Chang 1998; Owusu-Apenten 2002b) at a cost of $0.37 to $0.50 per sample (Owusu-Apenten 2002b). Owusu-Apenten (2002b) also reviewed applications of this method for different matrices. See Table 1 for combustion-based methods adopted by standards-setting organizations for different matrices.

Selectivity, interferences, and susceptibility to adulteration

This method is reported to quantitatively recover all forms of nitrogen, organic and inorganic, during analysis (Simonne and others 1997) and is specific for total nitrogen, but it lacks any degree of selectivity for protein. Therefore, this method is susceptible to adulteration by all organic and inorganic compounds that contain nitrogen.


The widespread practice of using Kjeldahl-based nitrogen-to-protein conversion factors for combustion-based protein calculation results in an overestimation of protein content (Simonne and others 1997; Thompson and others 2002). The combustion procedure can provide protein estimates equivalent to those of Kjeldahl-based measurements if appropriate coefficients are applied to correct for the difference in nitrogen recovery (Simonne and others 1997).


Intra- and interlaboratory relative standard deviations reported for different matrices range from 0.30% to 8.3% and 0.62% to 8.2%, respectively (see Table 3).


The combustion (Dumas) method for estimating the total (crude) protein content of foods is an improvement on the Kjeldahl method because of the former's faster analysis time, simplicity, and safety. Combustion methods provide similar or better precision compared to Kjeldahl, and with correction factors by food matrix can be calibrated to provide similar results. This method provides no improvements in selectivity for protein or susceptibility to adulteration when compared to Kjeldahl.

Modified Lassaigne method

Principle The original Lassaigne method is a qualitative test for organic nitrogen and is performed by fusing nitrogen-containing carbonaceous compounds with metallic sodium to form sodium cyanide that is then detected with Prussian blue. An adaptation of this method reported by Demirata and others (2002) allows quantification of organic nitrogen for food matrices in 3 steps: fusion of organic nitrogen with sodium to form sodium cyanide, conversion of sodium cyanide to thiocyanate in the presence of ammonium polysulfide, and complexation of thiocyante with Fe(III) that is quantified spectrophotometrically at 480 nm.

Simplicity, cost, throughput, and applicability

Compared to the Kjeldahl method, the modified Lassaigne method has a shorter digestion time and involves a simpler spectrophotometric quantification step than the titrimetry required by Kjeldahl. Its disadvantages include the handling and use of elemental sodium and the requirement to completely dry food samples before exposure to elemental sodium. Demirata and others (2002) reported the application of this method to meat products and baby food.

Performance characteristics and susceptibility to adulteration

Like Kjeldahl, this method measures organic nitrogen but not specifically protein-bound nitrogen (Demirata and others 2002) and is therefore susceptible to adulteration with nonprotein nitrogen. Comparison of this method to the Kjeldahl and Dumas methods gave similar results when applied to baby food and meat (Demirata and others 2002). In some circumstances, the repeatability of this method has been reported to be superior to that of the Kjeldahl and Dumas methods (Demirata and others 2002). A limitation is interference by a number of metals including Ag, Cu, Co, and Zn (Demirata and others 2002).

Copper binding-based methods

Biuret method

Background Originally described by Berthollet in 1873, the biuret method involves the formation of colored complexes between copper ions and protein. Its name comes from the reaction of biuret (NH2CONHCONH2) with cupric ions to form a colored complex, although the current method does not use biuret (Owusu-Apenten 2002c). The first applications of this method were for clinical estimations of protein in urine and serum in the early 20th century (Cole 1969). The currently used version of this method was first described by Gornall and others 1949. It became more useful to clinical chemists when improvements to the speed and reliability of the method were developed by Layne (1957). Adaptations of this method for food protein analysis began in the 1940s (Cole 1969; Owusu-Apenten 2002c).


Cupric ions are thought to react with peptides and proteins with 3 or more amino acid residues under alkaline conditions to form a colored chelation complex involving 4 to 6 peptide nitrogen atoms that absorbs light at 540 nm (Kołakowski 2001; Krohn 2005). The method has been extensively reviewed (Kołakowski 2001; Owusu-Apenten 2002c; Hortina and Meilinger 2005; Krohn 2005).

Accuracy, selectivity, interferences, and susceptibility to adulteration

Reactivity in this method requires the presence of proteins or peptides with 3 or more amino acids, along with any other compounds with 2 interconnected –CO–NH– groups, such as biruet, malonic diamide, and succinic diamide (Chang 1998; Johnson and others 2001; Kołakowski 2001; Hortina and Meilinger 2005; Krohn 2005). This specificity may not be absolute because Hortina and Meilinger (2005) reported that a variety of compounds are reactive in this method, including some amino acids (histidine, asparagine, threonine, and serine), dipeptides, and other compounds. They concluded that the absorbance measured at 540 nm in this method results from interactions of copper with both the peptide backbone and side-chain amino acids (Hortina and Meilinger 2005). Other compounds that may be reactive include EDTA, thiols, DNA, glutamate, aspartate, plant colors and dyes, lignins, glucose, glucosamine, starch, ammonium salts, and polysaccharides. Nonprotein and nonnitrogenous substances that have been reported to interfere include lipids, starch suspensions that cause light scattering, and dextran, which causes precipitation (Kołakowski 2001; Owusu-Apenten 2002c; Krohn 2005). Any nonprotein-interfering substances limit the utility of this method to rule out adulteration of food ingredients. Two modifications of the biuret method may enhance its specificity. To minimize interferences from ammonia, amino acids, DNA, guanidine, and urea, the protein–copper complexes can be measured at 226 nm at a pH >12 (Drochioiu and others 2002). TCA protein precipitation and separation by filtration before analysis have also been used to more specifically measure protein (Kołakowski 2001).

High correlations between Kjeldahl-based results and the biuret method have been reported for food matrices (Torten and Whitaker 1964; Pomeranz 1965; Cole 1969; Pomeranz and others 1977; Keller and Neville 1986; Owusu-Apenten 2002c). Accuracy as characterized by spike-recovery experiments using food proteins has not been reported. The color intensity per unit protein (often called the response factor) depends on the protein's amino acid content: Proline, cysteine, and hydroxyproline amino acids increase the response factor (Kołakowski 2001; Owusu-Apenten 2002c; Zhou and Regenstein 2006). This may be because of the effect of these amino acids on protein secondary structure, causing altered copper ion interaction with peptide bonds (Zhou and Regenstein 2006). The dependence of response factor on protein structure necessitates the use of reference materials to ensure accurate results (Chang 1998; Krohn 2005). This also suggests potential for adulteration by substitution of the authentic protein for a nonauthentic one with an equal or greater response factor.

Simplicity, cost, throughput, and applicability

The basic method for analysis of relatively pure protein samples without interfering substances is simple, fast, and inexpensive. It involves solubilization and appropriate dilution of protein, addition of a premixed alkaline copper reagent, incubation for 15 to 30 min, and absorbance measurement at 540 nm (Chang 1998; Krohn 2005). For increased throughput, the method has also been adapted for 96-well microplate analysis (Krohn 2005).

Use of this method for more complex matrices requires the use of more extensive sample preparation procedures in order to minimize interfering matrix substances. Such techniques include isolation of protein or extraction of interfering substances such as lipids by means of appropriate solvents (Owusu-Apenten 2002c). The biuret method has been adapted for food matrices of cereal grains and their flours, meat and dairy products, yeast proteins, and wheat and other grains (American Association of Cereal Chemists 2000; Owusu-Apenten 2002c).


No interlaboratory studies to evaluate the precision of this method for food matrices have been reported.

Other performance characteristics

The lower sensitivity of the biuret method compared to other colorimetric methods may be a disadvantage for some applications (Chang 1998; Krohn 2005) but may not be an issue during analysis of protein-based ingredients that typically have a high concentration of protein.


The biuret method is simple and inexpensive, and it provides a higher degree of selectivity for protein than total nitrogen based methods. Its results can be influenced by interfering matrix substances, intentionally added biuret-positive substances, and the variability of the response factor depending on protein primary and secondary structure. Effective use of this method for protein quantification and prevention of adulteration would require matrix-by-matrix development and validation, including the use of reference protein standards. A sample preparation step to isolate protein from the remaining matrix and potential adulterants would be necessary.

Lowry method

Background This protein method was developed by Lowry in the 1940s to rapidly measure small concentrations of protein in blood and was published in 1951 (Lowry and others 1951; Kresge and others 2005). It is one of the most cited methods in the life sciences. The method has wide utility in the clinical and biochemical fields because of its simplicity and sensitivity. Modifications to the original procedure have improved the linearity of color response and sensitivity to protein and are the basis of the currently used modified Lowry method (Miller 1959; Hartree 1972). The modified version of the basic procedure is sold commercially by Bio-Rad (as the DC Protein Assay) and Pierce (as the Modified Protein Assay Reagent). The method is simplified by use of preformulated and stabilized reagents and is compatible with sodium dodecyl sulfate (SDS) for samples that contain detergents (Krohn 2005). The history and chemistry of the Lowry method and its improvements have been reviewed (Peterson 1979; Owusu-Apenten 2002d; Krohn 2005).

Principle, selectivity, interferences, accuracy, and susceptibility to adulteration

This method is an enhancement of the biuret method by addition of the Folin-Ciocalteau (tungstate–molybdate) reagent to improve its sensitivity. Although the chemical mechanisms that result in color formation in this method are not fully understood, they occur in 2 steps: First, an alkaline mixture of cupric ions and sodium–potassium tartrate (Lowry reagent) react with proteins to form a Cu2+ coordinate complex (the biuret reaction). Because of the similar chemistry of this 1st step and that of the biuret method, biuret-positive compounds that cause selectivity and adulteration susceptibility in the biuret method also pose issues for the Lowry method (see the section on“Biuret method”). The 2nd step is thought to be a redox reaction during which electrons are transferred to tungstate–molybdate in the Folin-Ciocalteau phenol reagent, leading to a colored product that absorbs light at 750 nm. The electrons transferred in this redox reaction can originate in the Cu2+ coordinate complex, tyrosine and tryptophan residues, or other reducing agents present in the reaction mixture (Owusu-Apenten 2002d). In this way both the peptides and tyrosine and tryptophan residues contribute to color formation. The redox chemistry of the 2nd step of the Lowry method decreases the selectivity of the method for protein itself compared to the biuret method. Reducing agents and sulfhydryl agents, such as cysteine, tyrosine, and phenolic acids can reduce the tungstate–molybdate complex, leading to Lowry color formation and could therefore present a vulnerability to adulteration (Owusu-Apenten 2002d).

In total, more than 180 compounds are known to interfere with the Lowry method (Peterson 1979; Owusu-Apenten 2002d; Krohn 2005). These include free amino acids, lipids and fatty acids, plant pigments, reducing agents including reducing sugars, free thiols, polysaccharides, nucleic acids, phenolics, Fe2+, sulfur dioxide, and sulfites. The number of possible interfering and Lowry-reactive compounds suggests that the method would be vulnerable to adulteration and to a greater extent than the biuret method.

Sample pretreatment with deoxycholate and TCA (DOC-TCA) to precipitate the protein improves the method's selectivity for protein (Bensadoun and Weinstein 1976; Peterson 1977; Peterson 1983). Protein precipitation procedures may not eliminate the interference of protein-bound phenolic compounds (Lindeboom and Wanasundara 2007). In addition, glycosylation of proteins may increase the response factor (Fountoulakis and others 1992).

Limited data describe the accuracy of the Lowry method for food matrices. One report showed that Lowry had a significant bias versus Kjeldahl for protein analysis of seeds from Cruciferae species (Lindeboom and Wanasundara 2007), but in other cases a good correlation to Kjeldahl analysis has been noted for wheat and human milk (Keller and Neville 1986; Owusu-Apenten 2002d). Like those of the biuret method, Lowry protein response factors are influenced by the imino acid (proline and hydroxyproline) composition of different gelatins (Zhou and Regenstein 2006). This indicates the need for reference protein standards to ensure accuracy (Krohn 2005). It also indicates that the Lowry method has some degree of selectivity between different proteins and could be susceptible to adulteration by substitution of the authentic protein for a nonauthentic one with an equal or greater response factor.

Simplicity, cost, throughput, and applicability

The Lowry method is neither as simple nor as fast as the biuret method. For relatively high-purity protein samples, it involves solubilization and dilution of the protein with an initial 10-min incubation with the Lowry reagent that must be precisely timed and followed by simultaneous addition and mixing of the Folin-Ciocalteau reagent (Owusu-Apenten 2002d; Krohn 2005). The cost of this method is low. A microplate version of the modified methods is available to increase throughput and decrease reagent usage (Krohn 2005).

Use of this method for more complex matrices, including many food ingredients, requires more extensive sample preparation procedures like those described for the biuret method (see the section on “Biuret method”). Adaptations of the method for food matrices have been reviewed for cereal grains and flours, dairy proteins, yeasts, and collagen (Owusu-Apenten 2002d).


This method is neither well suited nor practical for the selective measurement of protein in food matrices in which there is potential for adulteration. The basic Lowry method itself is more complicated and time-consuming and more susceptible to adulteration compared to other methods such as biuret. One of the principal reasons for the widespread use of this method in clinical sample analyses is its sensitivity, which may not be as important for food matrices when large sample sizes can be used.

Bicinchoninic acid (BCA) method

Background Of the copper-binding methods, the bicinchoninic acid (BCA) method is the most recent, as first described by Smith and others (1985). BCA was developed to overcome the weaknesses of the Lowry method, including the 2 incubation steps, the precise timing requirement for these steps, and the incompatibility with nonionic detergents and buffer salts. The BCA reagent used in this method is available preformulated from Pierce (BCA Protein Assay Reagent) and Sigma (Bicinchoninic Acid Kit for Protein Determination).

Principle, selectivity, interferences, accuracy, and susceptibility to adulteration

The BCA method couples the reduction of Cu2+ to Cu1+ by protein (the biuret reaction) with the formation of a colored complex between BCA and Cu1+. Two possible reactions cause the reduction of Cu2+ to Cu1+: (1) a temperature-insensitive reaction between oxidizable protein side chains (cysteine, tryptophan, tyrosine, and phenylalanine) and (2) a temperature-sensitive reaction involving copper ions complexing with the peptide backbone of the protein (Wiechelman and others 1988; Owusu-Apenten 2002e). The colored complex formed between Cu1+ and BCA is thought to be a chelation complex involving 2 molecules of BCA with 1 cuprous ion (Smith and others 1985). The reaction times used in this method do not result in an end-point determination, indicating that reaction time and temperature can influence the response factor (Chen and others 1989; Krohn 2005). This can be used to limit detection of interfering substances.

The redox chemistry for the Cu2+ to Cu1+ oxidation step in this method decreases the selectivity of the method for protein itself compared to the biuret method. Any compound capable of reducing Cu2+ to Cu1+ during the procedure has the potential to be a BCA-positive substance, including biogenic amines, free amino acids, and di- or tripeptides, reducing sugars, phenolic acids (Kamath and Pattabiraman 1988; Slocum and Deupree 1991; Owusu-Apenten 2002e), nonprotein compounds that give a positive result in the biuret reaction, or phospholipids (Sapan and others 1999). BCA-positive phenols have a higher response factor than does protein (Gates 1991). For example, the response factor per unit substance was 106 times greater for pyrocatechol compared to protein (Gates 1991). The number of potential BCA-positive substances for this method suggests that it is more susceptible to adulteration than the biuret method.

Several variations to improve the selectivity of the BCA method and minimize interference from BCA-positive substances have been reported (Brown and others 1989; Gates 1991), but none have been evaluated for food matrices. Sonication improves BCA reagent interactions with insoluble cereal protein substances, and centrifugation removes interfering substances (Chan and Wasserman 1993). Many phenolic compounds have been reported to bind to proteins, which may minimize the effectiveness of precipitation procedures to separate protein from these potential adulterants before analysis (Lindeboom and Wanasundara 2007). Other interfering compounds such as EDTA chelate copper to limit its reactivity in BCA reactions (Krohn 2005), and lipids, phospholipids, H202, and Tween detergents may also interfere (Sapan and others 1999). In addition, glycosylated proteins may interfere with protein measurement by this method (Fountoulakis and others 1992; Noble and others 2007). One advantage of the BCA method over other copper-binding methods is its compatibility with nonionic detergents and denaturing agents (Chang 1998; Krohn 2005).

The limited accuracy information for this method with respect to food matrices indicates that results similar to Kjeldahl were obtained for cereal protein analysis using a solid phase modified BCA method that used bovine serum albumin as a standard (Chan and Wasserman 1993). As with the biuret and Lowry methods, protein response factors for BCA are influenced by protein types, which indicates the need for reference protein standards to ensure accuracy. This suggests another potential source for adulteration, substitution of the authentic protein for a nonauthentic one with an equal or greater response factor (Chang 1998; Sapan and others 1999; Krohn 2005).

Simplicity, cost, throughput, and applicability

The BCA method is relatively simple and involves only 1 incubation step. For relatively high-purity protein samples, solubilization and dilution of protein require approximately 30 min of incubation with a premixed BCA reagent, followed by measurement of absorbance measurement at 562 nm. This reaction time is longer than that of the biuret method, although a modification of the BCA method using a microwave oven reduces the incubation time to 20 s (Atkins and Tuan 1992). A microplate version of the method increases throughput and decreases reagent use (Atkins and Tuan 1992; Krohn 2005). The cost of this method is low.

Application of this method for food protein has been reported only for corn meal flour, soybeans, casein, zein (Owusu-Apenten 2002e), beer (Siebert 2005), and cereal (Chan and Wasserman 1993). Use of this method for more complex matrices including many food ingredients would likely require the use of extensive sample preparation.


Limited information about the precision of this method for food protein analysis shows repeatability for a solid phase modified BCA method similar to or better than Kjeldahl results for cereal protein (Chan and Wasserman 1993).


The basic BCA method may not be a suitable protein measurement procedure to prevent adulteration because it lacks selectivity for protein. Improving the selectivity of this method by introducing procedures to separate protein before analysis may be useful, but the presence of protein-bound BCA-positive substances could still limit its usefulness for food protein analysis.

Summary of copper-based methods

The biuret method appears to have the greatest potential to provide a reliable method for food protein measurement, with a relatively high degree of selectivity for protein to detect adulteration. It is a simple method with selectivity for a certain characteristic of peptides, the presence 2 interconnected –CO–NH– groups. Both the Lowry and BCA methods, because of their more complex redox chemisty, are more susceptible to producing erroneously high protein measurements for adulterated protein samples compared to biuret because of the significant number of false-positive compounds reported for these 2 methods. Use of the biuret method for analysis of relatively high-purity food proteins and more complex food ingredients likely will require a protein isolation/separation step to remove biuret-positive compounds and matrix-interfering compounds and the use of protein reference standards to deal with protein-to-protein response factor variability.

Dye-binding methods

Anionic dye-binding (also called Udy dye-binding) with Acid Orange 12, Orange G, and Amido Black 10B

Background Certain organic dyes react with proteins to form insoluble complexes. Fraenkel-Conrat and Cooper (1944) were the first to report quantitative reactions between Orange G dye and proteins. Udy (1954) was the first to report the application of this method for food protein quantification using Acid Orange 12 dye. Udy later commercialized the method for analysis of protein in food matrices. Other commercial instruments, such as the ProMilk Mark II or ProMilk PMA (Foss Food Technology Corp., Hillderod Denmark), were developed and used for milk protein analysis (in the 1960s and 1970s) using Amido Black 10B dye but are no longer commercially available. Most recently, the SPRINT rapid protein analysis system was commercialized by CEM Corporation using a dye-binding method (Urh 2008).


In acid-buffered solutions negatively charged sulfonic groups in azo dyes combine stoichiometrically with positively charged constituents in proteins to form an insoluble dye–protein complex. After the insoluble complex is separated from the reaction solution, the unbound dye concentration can be determined spectrophotometrically and used to calculate protein concentration. Protein constituents known to be positively charged under acid conditions due to their high pKa can bind with negatively charged azo dyes. This applies to proteins with free terminal amino groups and basic amino acid residues (ɛ-amino group of lysine, imidazole of histidine, and guanidine of arginine) (Fraenkel-Conrat and Cooper 1944). Three dyes have predominated for use in this method: Acid Orange 12, Orange G, and Amido Black 10B. Their use for food applications and their chemistry was reported by Owusu-Apenten (Owusu-Apenten 2002f).

Selectivity, interferences, accuracy, and susceptibility to adulteration

The selectivity of this method for protein is not well studied, but several food constituents, including calcium chloride, starch, and beta-glucans, have been reported to complex with anionic azo dyes to produce false-positive results (Udy 1954; Vanderzant and Miah 1961; Kołakowski 2001; Owusu-Apenten 2002f). Another potential source of error using this method could be the hydrolysis of a protein to increase the number of free terminal amino groups to which the anionic dyes bind, but anionic dye-binding methods do not display interference from lipids (Owusu-Apenten 2002f).

Solid phase anionic dye-binding methods involve: (1) absorption of protein onto solid phase supports, such as filter paper, glass fiber filters, or nitrocellulose, (2) optional treatment of absorbed protein with dilute TCA (to fix the protein), and (3) dyeing and excision of protein spots from solid phase supports into solubilizing solution that is then assayed. Although this procedure may be useful for isolating protein from interfering or nonprotein-reactive compounds, its complexity may limit its use.

Results from this method are well correlated with results from the Kjeldahl method (Sherbon 1967, 1968, 1974; Sherbon and Hemphill 1967; Sherbon and Luke 1969; Ashworth 1971). Protein-to-protein variability in dye-binding capacity using this method has been reported, indicating that the use of reference materials may be necessary to ensure accuracy (Kołakowski 2001; Owusu-Apenten 2002f). The extent of processing of protein-based ingredients can affect their dye-binding capacities. For example, collagen degradation increases dye-binding capacity (Veis and Cohen 1954), and processing steps that change the reactive lysine content of food affect its dye-binding capacity (Hurrel and others 1977). Like all methods that yield protein-to-protein variation in color response factor, this method could give artificially high protein values if authentic proteins in a sample were replaced by nonauthentic ones with a high response factor.

Simplicity, cost, throughput, applicability

The anionic dye-binding method is relatively fast and has been directly used with some food matrices (milk and cereal grains) without extensive sample pretreatment steps to extract and solubilize protein. It is not a simple one-step procedure because it involves a shaking step followed by a centrifugation or filtration step to separate the insoluble dye-bound protein before analysis. This manual method is inexpensive, but automated methods such as use of CEM Corporation's SPRINT can require approximately $24,000 of equipment and a reagent cost of approximately $3.00 per sample (personal communications 2008). The total analysis time for the manual method depends on the matrix, but is approximately 15 min per sample, although use of an automated method allows analysis times of 2 to 3 min per sample.

This method is widely used for protein analysis in a wide range of food matrices, and 3 official methods for food protein analysis use this procedure. Two official methods that use Acid Orange 12 and Amido Black 10B dyes have been adopted by AOAC International for milk protein analysis (AOAC International 2005). Method 46-14B approved by the American Association of Cereal Chemists (AACC) uses Acid Orange 12 dye for protein analysis in cereal grains, oilseeds, legumes, forage, animal products, and dairy products (American Association of Cereal Chemists 2000). This method is widely applicable for other food matrices (Owusu-Apenten 2002f).


The Udy method is precise: Reported intralaboratory RSD ranges from 0.00% to 1.10%, and interlaboratory RSD ranges from 0.52% to 1.13%. A similar study for other dairy ingredients reported an interlaboratory RSD range of 0.3% to 1.24% (Sherbon 1968). Two studies for milk indicated a lower intralaboratory RSD for the Orange G and Udy dye-binding methods when compared to Kjeldahl (Sherbon 1967; Sherbon and Hemphill 1967).


Although the anionic dye-binding method has few interfering substances compared to the unmodified copper-binding methods, it is not as simple as the biuret method and has potential for artificially high results with protein hydrolysis.

Bradford (Coomassie Brilliant Blue G-250 dye-binding) method

Background The use of Coomassie Brilliant Blue G-250 (CBBG) dye for the quantification of protein was first reported by Bradford (1976). Variations of this method in order to improve performance characteristics have been reported (Owusu-Apenten 2002g; Owusu-Apenten 2002h).

Principle, selectivity, interferences, accuracy, and susceptibility to adulteration

When CBBG dye binds with protein it forms a sparingly soluble complex and shifts the absorption maximum for CBBG to 595 nm. Binding is caused by electrostatic interactions between the dye, basic amino acids (arginine, lysine, histidine amino acid residues, and NH2-terminal amino group) and aromatic amino acid residues (Compton and Jones 1985; Tal and others 1985; Owusu-Apenten 2002g). Van der Waals forces and hydrophobic interactions are also thought to be involved in the color formation (Owusu-Apenten 2002g; Krohn 2005).

Dye-binding has been reported to occur only with proteins and polypeptides >3000 Da that are soluble under the acidic reaction conditions used in this method (Sedmak and Grossberg 1977; Krohn 2005). Other food polymers, such as plant polysaccharides, tannins, and glycosaminoglycans such as heparin (Godshall 1983; Khan and Newman 1990) may yield false-positive results. This indicates the method is not selective for protein and thus may be susceptible to adulteration with these Bradford-positive substances. Other compounds reported to interfere with this method include chlorophyll, pectins, ethanol, ionic and nonionic surfactants, lipids, and flavonoids (Owusu-Apenten 2002g). One advantage of this method is that many substances incompatible with other colorimetric protein methods, such as free amino acids, buffer salts, chelating agents, reducing sugars, and phenolics are compatible with the Bradford method (Owusu-Apenten 2002g).

DOC-TCA precipitation to isolate sample protein before analysis cannot be used with the Bradford method because of the formation of interfering precipitate (Chiappelli and others 1979; Pande and Murthy 1994; Owusu-Apenten 2002h). Two alternative precipitation procedures, low-temperature TCA precipitation and a protein coprecipitation procedure with calcium phosphate, may reduce the susceptibility of this method to adulteration by increasing its selectivity for protein and by reducing interferences (Owusu-Apenten 2002h). Keller and Neville (1986) reported a good correlation between the Bradford and Kjeldahl methods for human milk. Kamizake and others (2003) reported a 7.8% to 11.6%, 18% to 66%, and 6.9% to 9.4% difference for Bradford results compared to total protein nitrogen Kjeldahl for whole milk powder, whey protein powder, and buttermilk powder, respectively.

Significant protein-to-protein variation in dye-binding capacity occurs with the Bradford method and therefore necessitates the use of appropriate reference standard protein (Krohn 2005). Use of a BSA protein standard instead of casein standard for total protein estimation in whey protein powder increased the measured value by more than 40% (Kamizake and others 2003). In addition, protein glycosylation may interfere with protein measurement by this method (Khan and Newman 1990; Fountoulakis and others 1992). Glycation of human serum albumin has been associated with a 20% underestimation in Bradford protein content (Brimer and others 1995). This may be caused by the reduction in lysine and arginine residues available for dye interactions because of glycation at these sites.

The batch-to-batch variability of commercially available CBBG dyes greatly influences protein–dye interactions, necessitating standardization of the CBBG dye formulation (Owusu-Apenten 2002h). Reaction conditions, such as pH, presence of salts that alter hydrophobic interactions between dye and protein, and dye-to-protein ratio also change the sensitivity of the method (Owusu-Apenten 2002g; Owusu-Apenten 2002h).

Simplicity, cost, throughput, applicability

The basic Bradford method is simple, fast, and inexpensive. For relatively high-purity protein samples and some simple food matrices, it involves solubilization and dilution of protein, approximately 5 min of incubation with a CBBG reagent, and absorbance measurement at 595 nm. Solubilization of some proteins is difficult under the reaction conditions and variations including the use of a urea–mercaptoethanol solvent to resolubilize thermally denatured beta-lactoglobulin and egg albumin proteins have been reported (Gotham and others 1988). A microplate version of the basic Bradford method may increase throughput and reduce reagent use (Krohn 2005). Reported applications of this basic method for simple food matrices include skim milk powder, whole milk powder, beer, and wort (Lewis and others 1980; Kamizake and others 2003). Use of this method for more complex matrices, including many food ingredients, requires the use of more extensive sample preparation procedures, such as protein extraction to minimize matrix interferences for cereal products, corn, legumes, and potatoes (Snyder and Desborou 1978; Esen 1980; Owusu-Apenten 2002h).


A study organized by the American Society of Brewing Chemists found intra- and interlaboratory RSDs of 2.6% to 4.5% and 15.5% to 35.5%, respectively, using Bradford protein analysis (American Society of Brewing Chemists 1988). This relatively poor precision was attributed to the differences in individual batches of CBBG dyes used in the study (American Society of Brewing Chemists 1988; Owusu-Apenten 2002h).

Other performance characteristics

A nonlinear response is expected for the Bradford method when using the original protocol (Bradford 1976). This may limit the protein concentration range that can be measured and thus the flexibility and ease of use of the method in the field. A number of modifications have been reported to improve linear range using a ratio of absorbances at different wavelengths (Sedmak and Grossberg 1977; Bearden 1978; Zor and Selinger 1996). One commercial kit, the Coomassie Plus (Thermo Scientific Pierce Protein Research Products) is also known to have an improved linear response.


The Bradford method has not been validated or collaboratively evaluated for food matrices. Yet it has potential as a replacement for total nitrogen based protein measurements because of its simplicity, low cost, and potential for high protein selectivity and minimal susceptibility to adulteration when combined with a sample protein isolation step before analysis. One limitation of this method is its inability to detect peptides or proteins <3000 Da. Another is the potential that some proteins may not be easily measurable because of inherent lack of solubility under the acidic conditions of this method. On the other hand, these characteristics might provide better selectivity for individual protein types to reduce the risk of adulteration by nonauthentic proteins.

Ultraviolet absorption


Measuring protein using ultraviolet (UV) absorbance at 280 nm was first reported by Warburg and Christian (1942) and later by Layne (1957).

Principle, selectivity, interferences, accuracy, and susceptibility to adulteration

Most proteins exhibit a distinct UV absorption maximum at 280 nm due to the presence of the aromatic amino acids (tryptophan or tyrosine), cystine, or disulfide-bonded cysteine residues (Krohn 2005). Protein concentration is determined by solubilizing the test protein, measuring absorbance at 280 nm, and comparing the absorption with a reference standard or known absorptivity for the specific protein. This method is selective for compounds that absorb at the measurement wavelength. At 280 nm, protein-bound aromatic amino acid residues as well as free amino acids and nucleic acids will absorb. Correction for nucleic acid absorbance at 280 nm is routinely done by also measuring absorbance at other nucleic acid absorption wavelengths (235 and 265 nm) and using a ratio versus absorbance at 280 nm (Chang 1998; Kołakowski 2001). A large number of other chemical structures, such as free amino acids, nucleic acids, and phenolic acids can absorb at 280 nm, making this method highly vulnerable to adulteration.

To minimize interferences during absorbance measurement, protein solutions must be clear and colorless with no turbidity, or protein concentration can be overestimated (Chang 1998). One strategy to overcome interferences is the use of 4th-derivative analysis of the UV absorption of tryptophan and tyrosine between 270 and 300 nm to enhance the resolution of overlapping peaks and reduce matrix or background absorption interferences, but signal-to-noise ratio and therefore sensitivity are decreased (Luthi-Peng and Puhan 1999). Another approach is the combination of UV detection with HPLC separation as described in the section on “HPLC.” Like other methods, a vulnerability of this method is the replacement of authentic proteins with nonauthentic proteins that absorb at the measurement wavelength.

The UV absorbance of protein extract solutions from milk, beef, flour, bean, and egg yolk are highly correlated with results from Kjeldahl analysis (Nakai and Le 1970; Toma and Nakai 1971; Beavers and others 1973; Gabor 1979). Luthi-Peng and Puhan (1999) found good correlation between 4th-derivative UV spectrometry and Kjeldahl for total protein analysis of bovine milk and indicated that derivative absorbances follow the Beer–Lambert law that allows accurate quantification. Both traditional UV absorption and derivative spectroscopy methods must be calibrated against suitable reference protein solutions for accurate results (Luthi-Peng and Puhan 1999).


The basic UV method for protein is simple but not highly selective for protein. Minimizing the adulteration susceptibility of this method requires sample pretreatment steps even for relatively pure proteins to remove other compounds that absorb at 280 nm, and even then the method is useful only for proteins that can be solubilized and form clear solutions.


Mid-infrared transmittance

Background and principle The mid-infrared (MIR) region of the IR spectrum ranges from 2.5 μm to 25 μm (4000 to 400 cm−1). Absorption of infrared radiation in the MIR region reflects the fundamental vibrations (stretching and bending) of molecules, and it has been widely used to quantify protein concentrations in fluid samples (such as milk) by directly measuring transmitted radiation at specific wavelengths characteristic of the peptide bond. The primary absorption band for protein is 6.465 μm (1550 cm−1), and it originates from peptide bond absorption characteristics including N–H bending deformation and C–N stretching vibration (Ribadeau-Dumas and Grappin 1989). Other food constituents can absorb in this IR region and are taken into account using multivariate analysis and calibration factors from chemical reference methods to calculate protein results. Fourier transformation IR instruments are now widely used in the fluid milk industry for these measurements (Lynch and others 2006).

Simplicity, cost, throughput, applicability

After MIR equipment is calibrated the method is simple and rapid (0.5 to 2 min per sample). Calibration strategies include analysis of at least 8 samples covering the expected food matrix range that has been previously analyzed for protein using chemical reference methods (Kaylegian and others 2006). The primary reported application for this method is fluid milk analysis, and it has been adopted by AOAC International (2005) as method 972.16. Because of the high-throughput capacity, simplicity, and acceptable results for this method, it is currently the predominant method in fluid milk processing quality control labs and testing labs of the Dairy Herd Improvement Associations.

Selectivity, interferences, and susceptibility to adulteration

The selectivity of this method is difficult to characterize because advanced statistical tools such as multivariate analysis are needed to calculate results. Because of overlapping absorption spectra of many substances in mid-IR, simultaneous evaluation at multiple wavelengths or deconvolution procedures are needed to separate the absorption of proteins from other overlapping substances. The addition to analytical samples of substances that absorb at 6.465 μm (1550 cm−1) may yield false-positive results, but multivariate analysis procedures generally have the capacity to identify the presence of such substances because of a significant increase in the residuals, that is, the variances of the spectra that remain unexplained by multivariate analysis.

Because of the extensive use of this method for fluid milk analysis, only interferences for this matrix have been reported. Compounds that absorb in the protein region and could therefore give protein-positive results include carboxylic acids, such as citrate and lipolysis products (Ribadeau-Dumas and Grappin 1989). Other potentially interfering compounds include those with functional groups that absorb at 6.465 μm (1550 cm−1), such as those with N–H bending deformations at this wavelength. Melamine, for example, absorbs in this range (Sigma-Aldrich Co 1997). Because of the number of potential compounds that can absorb at 6.465 μm, this method may be susceptible to adulteration.


The accuracy of this method for milk protein analysis depends on instrument, calibration uncertainty, and sample deterioration (Kaylegian and others 2006). The AOAC method for this procedure indicates that a high level of accuracy is possible with a mean error of less than 0.05% (AOAC International 2005). Calibration is often performed with samples analyzed using Kjeldahl total nitrogen based protein methods. Because of the limited accuracy of the Kjeldahl method for milk analysis and its vulnerability to nonprotein nitrogen (see above in the section on “Current Kjeldahl method and its variants”), use of the true protein Kjeldahl-based method (AOAC International 991.22) has been recommended for calibration (Barbano and Lynch 2006). As with all methods that apply multivariate analysis, the selection of calibration samples that reflect the variances expected in the sample for analysis is critical.


The precision of mid-IR measurements depends on the selection of a suitable sample. Although manufacturer-supplied validation samples achieved a reproducibility of 0.028%, optimized validation samples improved the reproducibility to 0.01% (Kaylegian and others 2006). AOAC International (2005) method 972.16 for the mid-IR determination of protein and other constituents of milk includes reproducibility of ≤ 0.02% as an acceptance criterion.


Although practical for high-throughput analysis of protein in fluid food applications such as milk, this method lacks specificity for protein and is therefore susceptible to adulteration. Use of additional information from IR spectra may, however, improve the ability of this method to prevent adulteration.

Near-infrared reflectance

Background and principle Measurements in the near-infrared (NIR) region (0.7 to 2.5 μm or 14285 to 4000 cm−1) are generally used for qualitative, not quantitative, purposes. Absorption bands in this region are weak overtones or combinations of fundamental stretching vibrational bands that occur in the MIR region (3000 to 1700 cm−1 or 3333 to 5882 nm). They are broad and lead to complex spectra that require multivariate (multiple-wavelength) calibration to calculate the desired constituent concentration.

Both reflectance and transmission NIR have been reported for protein analysis. Diffuse reflectance happens when radiation penetrating the particle surface excites vibrational modes in the analyte and is then scattered in all directions. Karl Norris is credited with first recognizing the potential of NIR diffuse-reflectance measurements for routine protein analysis of grains from his work at USDA during the 1960s (Wetzel 1983). Bands at 2055, 2100, and 2180 nm are commonly used for such protein determinations, but this region of the NIR spectrum includes overlapping peaks for both starch and protein (Wetzel 1983). Like mid-IR methods, NIR results for protein are calculated using multivariate analysis and calibration factors from chemical reference libraries.

Simplicity, cost, throughput, applicability

As with mid-IR methods for protein measurement, analysis is simple and rapid, but extensive calibration is required for this method (Wetzel 1983; American Association of Cereal Chemists 2000). Applications of the method have primarily been grain matrices. AACC has adopted several NIR reflectance methods for protein analysis in ground samples of small grains, wheat flour, and soybeans and intact kernels of soybeans and whole grain wheat (Methods 39-10, 39-11, 39-20, 39-21, and 39-35, respectively) (American Association of Cereal Chemists 2000). AOAC International has adopted method 997.16 for NIR transmission analysis of wheat grain for protein content (AOAC International 2005). The United States Grain Standards Act established use of AOAC International method 992.23 (Crude Protein in Cereal Grains and Oilseeds Generic Combustion Method) as the reference method to calibrate NIR analyzers used by the Grain Inspection, Packers, and Stockyards Administration (U.S. Code of Federal Regulations 2009a). Use of NIR reflectance methods for protein analysis in cheese and milk powder matrices has also been reported (Ribadeau-Dumas and Grappin 1989).


As is the case with MIR methods, the degree of selectivity of this protein method is difficult to establish. The method is not specific for protein because other compounds absorb at the wavelengths used in this method. Because of the complex and overlapping absorption spectra of many compounds at these wavelengths, the simultaneous evaluation at multiple wavelengths or deconvolution procedures is needed to identify the absorption caused by proteins. Multivariate analyses are generally able to identify the presence of potentially interfering compounds by means of a significant increase in the residuals, that is, the variances of the spectra that remain unexplained by multivariate analysis.


As with all methods that apply multivariate analysis, the selection of calibration samples that reflect the variances expected in the sample is critical. Analysts must perform extensive calibration against samples that represent the range of protein and other matrix constituents that may be variably present (Wetzel 1983). AACC specifies the use of 30 to 50 samples for this purpose (American Association of Cereal Chemists 2000). NIR reflectance values are subject to the effects of particle size on scattering and sample penetration. Consistent particle size distribution in the test sample is necessary to ensure accuracy (Wetzel 1983). One study reported comparable results between this method and the AOAC International combustion method (990.03) for whole wheat grain analysis (Delwiche and others 1998).


Intra- and interlaboratory RSDs for NIR analysis of whole-grain wheat were reported to be 0.36% to 0.92% and 0.61% to 1.53%, respectively (Delwiche and others 1998).


Compounds that interfere with this method include those that absorb at the wavelengths used in analysis. For example, nitrogen-containing fertilizer compounds (ammonia, urea, ammonium sulfate, and ammonium nitrate) interfere with NIR reflectance protein measurements (McDonald and Bruns 1988). The interference for ammonia and ammonium ions may be the result of absorption peaks at the wavelengths used for protein determination (McDonald and Bruns 1988).

Susceptibility to adulteration

Like mid-IR methods, this method is susceptible to adulteration with compounds that absorb at the same wavelengths as protein (McDonald and Bruns 1988).


Although this method is practical for the analysis of large numbers of samples in applications such as establishing protein content for traded commodities, it is susceptible to yield artificially high protein values for adulterated protein samples. Analysis of additional NIR spectral data may be a potential approach to address this issue.

Amino acid-based methods

Ninhydrin-based total amino acid content

Background In 1910 Siegfried Ruhemann first reported the deep blue color produced by the reaction of triketohydrindene hydrate (ninhydrin) with primary amino groups (Friedman 2004). This chromophore is now referred to as Ruhemann's purple. All primary amines, including amines, amino acids, peptides, proteins, and ammonia, react with ninhydrin. In the presence of ninhydrin most free amino acids are oxidatively deaminated to form ammonia, and ninhydrin is reduced to hydrindantin. Ammonia and hydrindantin react to form the purple product diketohydrindylidene diketohydrindamine (Ruhemann's purple) that can be quantified at 570 nm (Kołakowski 2001). The Ruheman purple response factor of this reaction depends on amino acid type. Some amino acids such as cysteine and proline yield no Ruhemann purple color because their reactions with ninhydrin form other products (Friedman 2004). The chemistry and applications of the ninhydrin reaction have been reviewed by Friedman (Friedman 2004).


One approach for measuring total protein is to quantify total amino acids content. Hydrolysis of protein into amino acids followed by addition of the ninhydrin reagent to quantify total amino acid content was reported in 1972 and was confirmed in a number of laboratories (McGrath 1972; Horstmann 1979; Landry and Delhaye 1996; Starcher 2001). Interference from ninhydrin-positive substances, such as free amino acids, ammonium salts, amino sugars, nucleic acids, and ammonia are a limitation of the approach (Schilling and others 1963; Kołakowski 2001). Use of a blank measurement with ninhydrin before protein hydrolysis to exclude nonprotein ninhydrin-positive substances has been used to improve selectivity for protein (Drochioiu and others 2002). This blank measurement is then subtracted from the total hydrolyzed protein ninhydrin value (Drochioiu and others 2002).

Potential for food protein analysis to prevent adulteration

Use of this method for food protein measurement has not been well established, and gelatin protein determination is the only application reported (Drochioiu and others 2002). Limitations of the method include the failure of all amino acids to react with the ninhydrin reagent to form a colored product measurable at 570 nm and the potential for incomplete hydrolysis of the protein sample. Use of protein reference standards may be useful if standards can be developed and are similar in composition to the material being tested. Cations, such as Cu2+, Fe3+, Mn2+, Fe2+, and Mo2+ have been reported to interfere with the ninhydrin reaction by enhancing or inhibiting color development (Kołakowski 2001). Cations that enhance color development could be used as adulterants. Additionally, it is possible that addition of nonprotein substances that release ninhydrin-reactive substances under hydrolysis conditions could be involved in adulteration.


Total ninhydrin-reactive amino acid content for protein quantification may be a valuable tool to measure food proteins. Only one group has reported use of this method, and further studies will be necessary to evaluate the potential of this method. These studies will need to explore matrix compatibility, selectivity, accuracy, and precision of the method for food protein measurement.

Amino acid compositional analysis

Background In 1951 Moore and Stein reported an ion-exchange chromatographic method with postcolumn ninhydrin derivatization for amino acid compositional analysis (Moore and Stein 1951; Moore and others 1958; Sarwar and others 1983; Rutherford and Gilani 2009). An automated version of this method was the predominant method used for amino acid compositional analysis until the 1980s. Peace and Gilani (2005) reported that reversed-phase HPLC and GC were commonly used variations of this method, and capillary electrophoresis was considered an emerging technology for food applications. Although these methodologies have been most commonly used for determinations of the nutritional quality of food proteins, they have also been reported for quantifying total protein content (Sittampalam and others 1988; Owusu-Apenten 2002b).

Principle, selectivity, accuracy, and susceptibility to adulteration

Compositional amino acid analysis methods include hydrolysis of protein into amino acids, pre- or postcolumn derivatization, separation, quantification, and use of nonnaturally occurring amino acids as internal standards (Peace and Gilani 2005; Rutherford and Gilani 2009). Derivatization with ninhydrin, fluorescamine, or o-phthalaldehyde is often used to improve separation and/or detection. Total protein content using this method is calculated on a weight basis by summing individual amino acid concentrations times their molecular weights and using correction factors to account for amino acid recovery (Owusu-Apenten 2002b).

Accurate quantification of amino acid contents is complicated by the variable recovery of amino acids during protein hydrolysis (Peace and Gilani 2005; Rutherford and Gilani 2009). Factors that have been reported to contribute to this variability include the resistance of different proteins to hydrolysis and selective degradation of certain amino acids. Acid hydrolysis partially degrades methionine, cysteine, threonine, and serine and completely degrades tryptophan. Leucine, phenylalanine, isoleucine, and valine require more extensive hydrolysis because of bulky side chain hindrance (Blackburn 1978). Strategies to overcome these limitations include correction factors based on experimentally derived degradation kinetics, derivatization to protect labile amino acids during hydrolysis, and use of additional analyses to determine the content of amino acids known to be degraded or inadequately hydrolyzed (Blackburn 1978; Ribadeau-Dumas and Grappin 1989; Peace and Gilani 2005). Such approaches must be developed and validated for each protein type (Darragh and Moughan 2005). Use of matrix-matched protein reference materials as a form of internal standard may be an effective approach. Limitations of compositional amino acid analysis for protein quantification are that free amino acids may be added to adulterate a protein. In addition, when recovery or correction factors are relied on to account for amino acid recovery and degradation, substitution of authentic protein with nonauthentic protein with different amino acid composition could artificially inflate total protein values.

Simplicity, cost, throughput, applicability

When correction factors are employed to account for amino recovery and degradation issues, 3 steps are involved in this analysis: hydrolysis, separation, and detection. Hydrolysis and separation times from 5 min to 70 h and 6 to 180 min, respectively, have been reported, and automated analyzers are available (Peace and Gilani 2005; Rutherford and Gilani 2009). Methods that directly measure all amino acids are available but complicate and lengthen the analysis time (Peace and Gilani 2005). Applications of this method for total protein analysis in food matrices have been reviewed for a wide range of foods and food ingredients (Owusu-Apenten 2002b).


Use of this method for total protein analysis may not be practical because of the complexity of the method for accurate measurement and the potential for adulteration, especially with free amino acids.

Formol titration

BackgroundSteinegger (1905) first proposed the use of this method to determine the protein content of milk and to detect its adulteration with water. Pyne (1932) reported an adaptation of this method that eliminated the interference of phosphates by addition of oxalate. Other variations of this method have been reviewed (Taylor 1957; Cole 1969).


Free amino acids and protein-bound amino acids and peptides react with formaldehyde, producing methylene amino acid derivatives and changing the pKa of these amino groups. Two versions of this method have been developed. The direct version quantifies the carboxylic acid groups of these derivatives by adding neutralized formaldehyde to the test material and titrating to an end point with alkali (Taylor 1957). The indirect version quantifies the number of free amino groups by titrating the test material to a certain pH, adding formaldehyde, then titrating again to a specific pH (Taylor 1957). When calibrated against another total protein method such as Kjeldahl, results from these methods can be used to estimate total protein content.

Simplicity, cost, throughput, applicability

This method is relatively simple and has been applied primarily to dairy matrices including bovine fluid milk, human milk, ice milk and ice cream, and bovine skim milk powder. Results are comparable to those from Kjeldahl analysis (Pyne 1932; Castillo and others 1962; Hill and Stone 1964; Bakalor 1965; Roeper 1974; Abou Dawood and others 1977; Kumar and Seth 2004). For increased throughput it has been adapted for automated analysis by means of an autotitrator (Bakalor 1965).


This method is selective for substances and matrices that are titratable under the method conditions. Based on the chemistry for the direct and indirect methods reported by Taylor, these methods are selective for carboxyl and amino groups, respectively (Taylor 1957).

Susceptibility to adulteration

Addition of nonprotein titratable substances and substitution of nonauthentic protein for authentic protein are potential vulnerabilities for adulteration using this method.


This method may not be as suitable as other methodologies to selectively measure protein.

p-Benzoquinone method

p-Benzoquinone reacts with amino acids, including protein-bound amino acids, to form a charge-transfer complex that absorbs at 350 nm and can be used to spectrophotometrically quantify total protein. Barreto and others (1990) described the addition of p-benzoquinone to a dilute buffered milk solution followed by a 10- to 20-min incubation at 100 °C to quantify the total protein content of skim milk powder (Lichtig and others 2001). Quinones such as p-benzoquinone react with a variety of primary and secondary amines (Barreto and others 1990). This lack of specificity for neither amino acids nor protein-bound amino acids indicates that this method may not be a suitable total protein method to prevent adulteration.

Turbidimetric/nephelometric methods

Measuring the turbidity of precipitated protein suspensions in order to quantify the protein content of food and food ingredients has been reported since at least the early 20th century (Kober 1913). For high-purity proteins analysts can add a precipitant (for example, a low concentration of TCA, sulfosalicylic acid, or potassium ferricyanide in acetic acid) to a protein solution and measure the reduction in transmitted light (around 540 to 600 nm) caused by the scattering of radiation by protein particles (Layne 1957; Cole 1969). Measurement in more complex matrices such as wheat flour requires an initial step to extract protein from the matrix before addition of a precipitant (Feinstein and Hart 1959). Food matrices analyzed using this method include corn proteins, barley protein, and milk proteins (Kober 1913; Feinstein and Hart 1959; Tappan 1966; Cole 1969; Paulis and others 1974).

Limitations of this method include protein-to-protein variability, fluctuations in turbidity of the protein suspension over time, lack of reproducibility, and interference of matrix materials such as nucleic acids that precipitate under similar conditions (Layne 1957). These limitations suggest that this method may not be as suitable as other available methods for selective protein measurement.

New spectral probes

Simple spectrophotometric methods for total protein determination remain of interest for both research and routine analysis because of their simplicity and low equipment cost. New spectrophotometric and fluorometric probes are being developed for protein quantification (Qi and others 1998; Hong and others 1999; Qin and others 2006; Sözgen and others 2006; Sun and others 2008).

Several total protein methods involving protein-binding fluorescent dyes have been reviewed to assess their applications to biopharmaceuticals (Noble and others 2007). Examples include the use of fluorescamine (4-phenylspiro-[furan-2(3H),1-phthalan]-3,3′-dione), a nonfluorescent compound that reacts with terminal amino groups and the ɛ-amino of lysine to form fluorescent moieties (Udenfriend and others 1972; Noble and others 2007), and the use of CBQCA [3-(4-carboxybenzoyl)quinoline-2-carboxaldehyde], another fluorescent dye that reacts with secondary amino groups (You and others 1997; Noble and others 2007). Advantages for both of these methods compared to other colorimetric methods are their low limits of quantification and high sensitivity (Udenfriend and others 1972; You and others 1997). Both have reported protein-to-protein variation in fluorescence response factor (Noble and others 2007). Limitations include the lack of selectivity for protein and addition of free amino acids or adulteration by protein hydrolysis or protein hydolyzates that exposes more terminal amino groups, any of which could artificially increase protein values.

For nonfood applications a fluorometric total protein method, the Quant-iT (Invitrogen, Carlsbad, Calif., U.S.A.), has been commercialized, but little information about its chemistry has been reported. Noble and others (2007) reported that the dye used in this detergent/dye-based method binds to hydrophobic regions of proteins and to detergent-coated protein. Qi and others (1998) described the use of a fluorescent probe, tetrachlorotetraiodofluorescein, by itself or in combination with a surfactant for total protein determination. This simple method reportedly has sensitivity similar to that of the Bradford assay, significant protein-to-protein variation, minor interferences from sodium and potassium salts and urea, and little interference from free amino acids (Qi and others 1998). Sun and others (2008) reported the use of the 4-chloro-(2′-hydroxylophenylazo)rhodanine-Ti(IV) complex as a fluorometric probe for protein determinations in milk powder and cornmeal.

Several spectrophotometric probes have been utilized for total protein determination in food matrices. Qin and others (2006) reported the use of the o-nitrophenylfluorone-Mo(IV) complex as a colorimetric probe in a Tween 20 microemulsion for protein determination of milk powder. This method was reported to be rapid (5-min reaction time) and more sensitive than the Bradford method (Qin and others 2006). Potential interferences to this method include cross-reactivity with glucose, free amino acids, metal ions, and lipids (Qin and others 2006). Sözgen and others (2006) reported use of a copper(II)-neocuproine spectral probe for spectrophotometric determination of total protein with a reaction similar to that of the biuret method. The authors proposed an outer-sphere electron transfer redox reaction mechanism for the method and reported that Cu(II) chelators and reducing agents are the most significant interfering agents for this assay. The effects of these interfering agents can be minimized by the use of protein isolation steps before analysis (Sözgen and others 2006). The method was also adapted for complex food matrices using a protein isolation step before analysis (Sözgen and others 2006).

Although most of the methods just described have reported improved analytical characteristics, such as sensitivity compared to traditional colorimetric tests for total protein, none have demonstrated comparatively higher selectivity.

Summary and perspectives

Table 4 summarizes the strengths and weaknesses of existing methods for food protein measurements. Many methods show a higher selectivity for protein than do total nitrogen based methods. Most of these are indirect methods because the analytical target is not the protein itself but rather its constituents or their functional groups. Many of these methods have analytical weaknesses based on their chemistry, and thus they can be exploited by those trying to cheat. Most of these methods may not be able to differentiate authentic from nonauthentic proteins. Additionally, many food matrices have unique requirements that necessitate different approaches for protein measurement. Thus a combination of different protein analysis methods may be necessary to effectively prevent adulteration. Specific analytical strategies that may have potential for this purpose include the following:

Table 4–. Summary of currently available methods for food protein measurement.
Kjeldahl and Dumas (combustion) methods• Applicable to most matrices• Use of concentrated acid
• Few interferences• Cannot distinguish nonprotein nitrogen from protein nitrogen
• Automated version available• Accuracy
• Precision 
“True Protein” nitrogen using Kjeldahl• Can distinguish nonprotein nitrogen from protein nitrogen• Few evaluated matrices
Copper binding (biuret, Lowry, BCA)• Inexpensive, adaptable for high-throughput• False-positive (potential adulterant) compounds unless protein isolation step used
 • Matrix interferences unless protein isolation step used
 • Requires solubilization of protein
 • Precision not well understood
Anionic dye-binding• Inexpensive, fast, automated version available• False-positive (potential adulterant) compounds unless protein isolation step used
• May not require protein solubilization• Potential effects of protein hydrolysis on measurement
• Precision 
Bradford• Simple, inexpensive, fast• Requires solubilization of protein and some protein may not be soluble under reaction conditions
• Can be highly selective for protein if protein isolation step is used• Inability to detect proteins below 3000 Daltons
 • Some matrix interferences
UV• Simple, inexpensive, fast• Requires solubilization and clear sample solution
 • False-positive (potential adulterant) compounds unless protein isolation step used
Infrared (mid-IR and near-IR)• Fast and high-throughput• Expensive
• Precision• Requires extensive calibration
Total amino acids content• Simple adaptation available to minimize false-positives• Hydrolysis time
 • Not evaluated for many matrices
Amino acid composition• Applied to many matrices• Accuracy issues from amino acid recovery
• Minimal interferences from non-amino acid adulterants• Does not distinguish free from peptide-bound amino acids
Formol titration• Simple• False-positives from nonprotein titratable substances
p-Benzoquinone• Simple• False-positives from nonprotein primary and secondary amines
Turbidimetric / nephelometric• Simple• False-positives from matrix interferences
 • Precision
Other spectral probes• Better sensitivity compared to other probes• Selectivity not improved compared to other probes
  • 1Use of a protein isolation step before total nitrogen based protein analysis is one potential solution. An example of this analytical strategy, the true protein Kjeldahl method described in the section “True protein using Kjeldahl nitrogen analysis,” has already been developed and validated for the fluid milk matrix and has been adopted by AOAC. The benefit of using this strategy is that Kjeldahl and Dumas equipment are already widespread in food-control laboratories. Limitations are the additional time and reagent cost that increase the cost of total protein analysis. Another limitation is this strategy's inability to distinguish different types of protein, which allows opportunity for adulteration with less expensive proteins. Implementation of this strategy will require development and validation of protein-isolation procedures for individual food ingredient matrices and possibly the use of reference standard proteins to quantify and to account for potential variability in the isolation step's efficiency.
  • 2The Bradford method may have potential by itself and has the benefit of not requiring a protein isolation step. Relative to the other colorimetric methods examined in this review, this method had the fewest reported weaknesses in terms of potential for adulteration. Because this method has significant protein-to-protein variation in response factor, it includes the benefit that it may be able to distinguish different protein types. Successful use of such an approach requires further development, including further investigation of the known nonprotein Bradford-positive substances so analysts better understand the risk for adulteration with these substances. Other requirements include matrix-by-matrix validation, development of reference protein standards for accurate measurements, and standardization of the dye used in this method. Limitations include interference by some matrix substances such as lipids, the poor solubility of some proteins under acidic conditions that may limit the number food ingredients that can be analyzed, and the reported detection only of proteins >3000 Da.
  • 3A potential solution is the combination of a protein-isolation step with a colorimetric method (such as Bradford or biuret) that has relatively few adulteration vulnerabilities. Relative to the other colorimetric methods reviewed, these methods are the simplest and fastest. A benefit of this strategy is that potential false-positive or matrix-interfering compounds can be separated from the protein before analysis. A disadvantage is the additional time and reagent cost that could increase the cost of total protein analysis. Successful use of this strategy may require extensive development work, including development of isolation procedures that can effectively separate protein from interfering compounds. Further disadvantages are the requirements for matrix-by-matrix development of reference standards to address protein-to-protein variability in response factors and development of internal reference protein standards to address variability in the efficiency of protein isolation.
  • 4The ninhydrin-based total amino acid content method reported by Drochioiu and others (2002) has potential and warrants further investigation. Benefits of this method include its improved selectivity for protein compared to nitrogen-based approaches and its lack of need for a protein isolation step because false-positive compounds are accounted for by a blank measurement. Limitations are that the method may not be suitable for food ingredient matrices that include compounds that could interfere with measurement and longer analysis time necessitated by the protein hydrolysis step. Successful use of this method may require further exploration of its analytical performance characteristics, including precision. The method also will require development of internal reference standard proteins to account for potential variability in the efficiency of protein hydrolysis, as well as matrix-by-matrix method validation.

Emerging Methods for Food Protein Quantification

Antibody-based methods (ELISA, biosensor)

Enzyme-linked immunosorbent assays (ELISA) are highly selective biochemical methods that can detect and quantify specific proteins in complex food matrices. ELISA tests historically have been used for the detection of allergens and other undesirable substances in food. An ELISA is based on the specific reactions between an antibody and an antigen. The biochemistry of the ELISA method has been reviewed (Märtlebauer 2003; Catala and Puchades 2008). The protein under investigation serves as the antigen, and the antibodies are obtained from animals or cell cultures that have developed an immune response to this antigen. Antibodies recognize specific segments of a protein, called epitopes, which typically are regions of the protein that consist of 7 to 12 amino acids. An epitope can be a linear sequence of amino acids (primary epitope) or can be more complicated structures (conformational epitopes) in which the proximity of the amino acids forming the epitope is established by the folding of the protein. The immune response of a mammal exposed to an antigen triggers a number of cellular events that lead to the development of multiple antibodies specific for different epitopes of the antigen. ELISA can employ either a mixture of different antibodies, also called polyclonal antibodies, or a specific antibody from a single cell line, a monoclonal antibody. ELISAs are useful in the food industry for both qualitative and quantitative applications. The robustness of an ELISA test against changes in the protein caused by processing and matrix interferences must be well established. Antibodies against primary epitopes tend to be more useful for mild processing and matrix interactions than are antibodies directed against conformational epitopes, indicating that the selection of the appropriate antibody is a key factor to develop a robust ELISA.

Validation of an ELISA requires the use of reference materials and must be validated for each matrix that will be analyzed (Catala and Puchades 2008). Reported applications of this method include food authentication applications such as discriminating between animal species in meat and dairy products as well as detecting and quantifying allergenic or contaminating proteins in food products (Brandon and Friedman 2002; Koppelman and others 2004; Whitaker and others 2005; Asensio 2008). Prepared materials in commercial kit format are available for routine use in food control laboratories. This method has not been investigated for use in total protein measurement in food matrices, indicating that more research is necessary so analysts can understand the utility and issues associated with ELISA.

One of the limitations of ELISA is the susceptibility of the antigen to processing, especially heat, changes in pH, and other denaturing conditions. Those conditions can change the primary and/or secondary structure and physical characteristics of the protein and potentially can destroy conformational epitopes or can chemically degrade the protein, thereby destroying linear as well as conformational epitopes. Food processing can cause any or all of these effects on proteins that can make the protein unrecognizable to the antibody, leading to false-negative results. Other limitations include limited reproducibility, relative complexity of the ana-lytical method, and long analysis time. Relative standard deviations for intra- and interlaboratory reproducibility of 4.2% and 7.5%, respectively, were reported for ELISA analysis of soy proteins in model processed foods (Morishita and others 2008). This is substantially less reproducible than the value reported for total protein and colorimetric methods. ELISA methods have often been used to evaluate processed finished foods that are inherently more variable and lower in protein than the standardized food ingredients matrices used to evaluate other test methods. Most quantitative ELISA methods are more complex than other methods such as Bradford or Lowry. An ELISA method typically involves multiple washing and incubation steps that can introduce errors, but ELISAs can easily be adapted to microplate format for high-throughput analysis (Camafeita and others 1997).

Although ELISA methods have potential utility in total protein analysis to prevent adulteration, for food ingredients the complex mixture of proteins with variable conformations may require multiplexed analysis utilizing microarray methods. Their use for food protein quantification has not yet been reported but has been identified as an area for further exploration (Markoulatos and others 2004).

Another application of the antigen–antibody reaction is in biosensors, which involve a sensing layer with an immobilized biological recognition element such as an antibody and a transducer coupled to the sensing layer (Leca-Bouvier and Blum 2005). In general, biosensor methods are operationally simple, rapid, selective, and inexpensive and have been reviewed in terms of their use for protein analysis (Leca-Bouvier and Blum 2005). Applications have included the use of a surface plasmon resonance based immunoassay biosensor for quantification of immunoglobulin G in bovine milk (Indyk and Filonzi 2003) and the use of a multiplex immunoassay biosensor to detect soy, pea, and wheat protein adulteration of milk powder (Haasnoot and others 2001; Haasnoot and du Pré 2007).

Although antibody-based methods for the analysis of total protein in food show promising characteristics, the susceptibility of proteins to processing conditions make challenging the selection of antibodies that react to antigens and that are not influenced by food processing conditions. In addition, validation may be an impediment to implementation because method suitability must be established for each matrix individually to ensure that all relevant variations of processing conditions for this matrix are properly considered.


The application of HPLC to analysis of the amino acid composition of proteins has been discussed above. The separation of proteins includes reversed-phase, ion-exchange (based on net charge), hydrophobic-interaction (based on surface distribution of hydrophobic residues), and size-exclusion (based on size) chromatography. Although for protein analysis mass spectrometry is more sensitive and precise than UV absorption, applications of the latter are more economical and have been used for the authentication and detection of adulteration of dairy protein-based ingredients and products. Reviews of food protein applications for these methods include the analysis of the 4 major proteins in bovine milk (Bordin and others 2001), the analysis of 4 caseins (Bramanti and others 2002), and others (Ribadeau-Dumas and Grappin 1989; Nollet 2003; Sass-Kiss 2008).

HPLC separation procedures can quantify individual proteins in food ingredients. The limitations of this method for routine analysis of food ingredients include long analysis times, the need for protein isolation from food matrices, the high volume of sometimes toxic solvents required for mobile phases, the difficulties of identifying specific proteins, and equipment cost.

Electrophoresis and microfluidic lab-on-a-chip methods

Electrophoresis is a powerful tool for separating, with high resolution, complex mixtures of proteins and peptides. Polyacrylamide gel electrophoresis (PAGE) and its modifications are powerful qualitative tools and have traditionally been used for food protein analysis but generally are not considered to be quantitative (Luo and others 2004). Their utility in routine analysis of proteins is also limited by the time and reagent volumes involved in this labor-intensive method (Luo and others 2004; Anema 2009). Capillary electrophoresis (CE) has gained popularity for the separation and quantification of food proteins (Cifuentes and García-Canas 2008) because it uses lower volumes of reagents, is more rapid and adaptable to high-throughput analysis than PAGE, and maintains high resolution. Applications of this method generally have been in research for and quality-based compositional analysis, investigation of effects of genetics and processing on proteins in foods, and food authentication for milk proteins (Recio and others 1997). The suitability of CE is hampered by the absorbance of proteins onto the capillary wall, which impairs the reproducibility of this method (Cifuentes and García-Canas 2008), and the limited usefulness of UV detection for identifying proteins.

Microfluidic lab-on-a-chip electrophoretic methods have been developed and used for the rapid and automated separation and quantification of proteins (Goetz and others 2004; Wu and others 2008). This method has been reported for analysis of proteins ≤250 kDa and has been reported to analyze 10 samples at once within 30 min (Anema 2009) For the analysis of milk proteins, the relative standard error has been reported to be as high as 15%, indicating that this method is not yet sufficiently reproducible for routine quantitative food-protein analysis (Anema 2009).

Mass spectrometry

The analysis of proteins by mass spectrometry became possible with the advent of suitable ionization procedures such as electron-spray ionization (ESI) and matrix-assisted laser desorption ionization (MALDI). MALDI time-of-flight mass spectrometry (MALDI-TOF-MS) is useful for food protein research because of its extensive compatibility with food matrix components. In contrast to ESI, MALDI does not require solubilization of proteins (Guy and Fenaille 2006; Horneffer and others 2007). For complex food matrices, ESI-MS requires cleanup before MS analysis using an LC separation step, but MALDI-TOF-MS is more compatible with crude extracts (Vaidyanathan and Goodacre 2003).

The applications of this method for food protein include quantification, fingerprinting, adulteration detection, and authentication (Léonil and others 2000; Careri and others 2002; Guy and Fenaille 2006). One example is the use of MALDI-TOF-MS to quantify gluten gliadins (25 to 40 kDa) in food samples (Camafeita and others 1997). MALDI-TOF-MS characteristics that make it attractive for protein quantification include speed (minutes), operational simplicity, low reagent cost, and high resolution (analysts can separate peaks that differ by a few Daltons) (Horneffer and others 2007). Furthermore, this procedure is compatible with alcohol (used for protein extractions) and buffers, salts, nonionic detergents, and reducing agents and is relatively simple to perform with current automated systems (Camafeita and others 1997). Beyond the identification and quantification of proteins, MALDI-TOF-MS can also provide information about protein conformation and structure and degree of hydrolysis caused by processing. These characteristics may be useful in assessing the quality of protein-based food ingredients (Horneffer and others 2007).

Applications of MS methods are restricted by the insolubility of some proteins, the limited ionization of some plant and meat proteins because of their low content of charged amino acids (Léonil and others 2000), and decreased mass resolution and accuracy above 30 kDa (Vaidyanathan and Goodacre 2003). Equipment cost (>$100,000) also can be a barrier (Goetz and others 2004).

Summary of emerging methods

Table 5 summarizes the strengths and weaknesses of the emerging protein analysis techniques described in this section. These methods have a high degree of selectivity for individual proteins in food ingredients, which may complicate the measurement of total protein for food ingredients but is useful for identifying and authenticating food protein ingredients. Overall, most of the emerging methods described are not yet sufficiently developed or practical for use in routine protein measurements. They do have potential applications as methods and instrumentation mature.

Table 5–. Summary of emerging methods for food protein measurement.
Biochemical methods e.g., ELISA• 96-well plate (and more)• High specificity possible• Sensitive to protein degradation
• Test strip• High-throughput automation possible• Labor-intensive unless fully automated
• Biosensor• Low instrument costs• Protein needs to be solubilized
  • Currently not precise enough
Chromatographic methods e.g., HPLC• Columns: size-exclusion, hydrophobic interaction, ion exchange• Analysis of full protein• Protein needs to be solubilized
• Detectors: ultraviolet and mass spectrometer• Less sensitive to protein degradation• Instrument costs
  • High operating costs (solvents)
  • Longer analysis time
Electrophoretic methods e.g., capillary electrophoresis• Capillary electrophoresis• Analysis of full protein• Instrument costs
• Lab-on-chip• Potential for high-throughput• Protein needs to be solubilized
 • Short analysis time 
 • Less sensitive to protein degradation protein identification possible 
 • Low operating costs 
Mass spectrometric methods e.g., direct MS analysis• Electron-spray ionization• Analysis of full protein• High instrument costs
• MALDI• Less sensitive to protein degradation• EI requires solubilization
 • MALDI does not require solubilization• MALDI currently not precise enough (matrix effects)
 • High identification power 

Reference Materials for Total Protein Measurement

Reference materials help ensure the comparability of measurement results and demonstrate the reliability and utility of analytical test methods (Taverniers and others 2004). Comparability of test results is particularly important for food ingredients whose functional and nutritional quality determines their economic value, as is the case of most protein-based food ingredients. Determining a fair commercial value necessitates that buyers and sellers of the same ingredient arrive at comparable test result regardless of the test method used or the point in the supply chain where the measurement is made. Reference materials can help ensure the comparability of analytical results by traceability to common reference standards with assigned property values and measurement uncertainties (Taverniers and others 2004).

Protein-based food ingredients are often composed of a mixture of proteins and other matrix materials as shown in Table 6. Several potential challenges exist for the development of such reference materials:

Table 6–. Composition of potential protein-based ingredients for development of reference material to support total protein measurement.
SourceIngredientIngredient compositionComposition of major proteins
  1. MW = molecular weight.

Bovine milkWhey protein isolate>90% protein, <6% fat, <6% lactose, <6% water (U.S. Pharmacopeial Convention 2008)Major families of proteins include β-lactoglobulins, α-lactalbumin, proteose-peptones, serum albumin, immunoglobulins. MWa 18 to 1000 kDa (Farrell and others 2004)
Casein, caseinate salts>84% protein, <2.25% fat, <2% lactose (US Pharmacopeial Convention 2008)αs1–, β–, κ–, and αs2–are the main casein protein families (Farrell and others 2004). MW 19 to 24 kDa.
WheatWheat gluten, wheat protein isolate>71% protein, <21% starch, <2% fat, <10% water (U.S. Pharmacopeial Convention 2008)Gliadins (α, β, γ, and ω) MW 30 to 80 kDa; Glutenins MW 12 to 130 kDa (Damodaran 1996).
SoySoy protein concentrate, soy protein isolate65% to 90% protein, <4% fat, <10% water (U.S. Pharmacopeial Convention 2008)Mostly globulins, including glycinin and α–, β–, and γ–conglycinin. MW 35 to 350 kDa (García and others 1997a, 1997b).
CornZein88% to 96% protein, <8% water, <2% ash (US Pharmacopeial Convention 2008)Alcohol-soluble proteins known historically as prolamines: α–, β–, γ–, and δ–zein are the major fractions. MW 9 to 27 kDa (Zhu and others 2007)
  • • A 1st challenge is defining the measurand and unit and establishing the definitive characterization/measurement method for defining the protein value for standards. It may not be practical to define the analyte in such detail that it can be traceable to an SI unit and thus to a primary reference standard, as is the case for other analytes. For example, dietary fiber and enzymes are linked to an internationally recognized, albeit arbitrary, method or material in an effort to make measurements comparable even without traceability to SI units.
  • • A 2nd challenge is that for a number of analytical methods the measured amount of total protein depends on the nature and composition of the specific protein. This is well established for colorimetric protein methods, most of which have a variable analytical response per unit of protein depending on the type of protein (Chang 1998; Krohn 2005). For example, in the Bradford dye-binding assay to quantify total protein, the relative color response per unit of protein for bovine α-chymotrypsinogen is about one-half that of bovine albumin (Krohn 2005). These differences in response can be attributed to the compositional and physicochemical variation of different proteins. Also, the inherent biological variability of raw materials and differences in methods of manufacture have an effect on measurement of total protein content of a sample. In all these cases, carefully selected reference materials can mitigate some of these influences, and a case-by-case analysis is necessary to establish the usefulness of a specific reference material for a particular matrix.
  • • A 3rd challenge is that protein-based ingredients in foods contain, in addition to protein, other matrix components including carbohydrates, lipids, and inorganic substances as indicated in Table 6. Not only do these matrix components have the potential to interfere with protein measurement, but they also can vary depending on factors related to their biology and method of manufacture (De Noni 1997). Like the challenge of protein variability, there is a potential need for process- or biology-specific reference materials, but they should be made on a case-by-case basis.

These challenges are not unique, and several approaches have been proposed to address them. Cordeiro and others reported the use of high-purity protein for total protein measurements of milk using RP-HPLC with UV detection supported by 7 reference standards for individual high purity milk proteins (Bordin and others 2001; Cordeiro and others 2001). These researchers indicated a high level of reliability for this approach with relative standard uncertainty <4% (Cordeiro and others 2001). An advantage to this approach is that potential variations in matrix compositions may be less likely to affect protein results because the matrix is separated from the protein before measurement. Potential limitations of this approach are the need for a chromatographic separation that may not be practical for routine analysis and assurance that the standard protein is sufficiently similar to the sample protein.

Another approach that is commonly used for analysis of complex matrices is the development of matrix-matched external reference standards (standards with compositions the same as or similar to those of the sample analyzed) with an assigned value for the analyte (Morabito and others 2004). An advantage of this approach is that matrix effects are controlled during analysis.

The development of reference materials supports the comparability and certainty of results of total protein methods for food ingredients that are more selective for protein than are current nitrogen-based methods. A comprehensive understanding of the protein and matrix composition, their potential variability that may influence detection, the similarity of standard and test samples, and the traceability implications is necessary to develop such standards.


As long as the value of food ingredients is based on protein content, the incentive to adulterate these materials by measures designed to inflate protein measurement will exist. The use of 19th-century crude protein measurements that do not detect many types of adulteration can be detrimental to public health, as recently demonstrated by melamine incidents. Development and implementation of more suitable protein measurement methods—ones that cannot be as easily falsely manipulated—have the potential not only to reduce the risk to public health but also to advance protein measurement science beyond total nitrogen based methods.

A path forward is the development, validation, and implementation of new analysis methods specific for protein-based food ingredient(s). This should include both quantitative protein methods with higher discrimination power against the presence of nonprotein nitrogen sources, and complementary qualitative authentication techniques capable of detecting the presence of unexpected nonprotein compounds in food and feed samples.

As presented in this review, several relatively simple and selective quantitative protein methods have potential suitability for development into methods to replace current total nitrogen techniques for use in food control laboratories. Most of the alternative methods were developed and evaluated for research purposes and may be compatible only with food ingredients that consist mainly of protein and/or show a relatively low degree of variability in composition. Future research should further develop and validate these methods for specific food and feed ingredients considering the unique protein and matrix challenges of each one. Development of appropriate protein reference materials will also be essential for more selective and accurate methods.

Qualitative authentication methods for protein-based food ingredients were not included in the scope of this review, but should be further investigated. Together, the future development and utilization of new quantitative purity and qualitative authentication tools have the potential to significantly reduce the risk of future protein adulteration.