Cancer proteomics and its application to discovery of therapy response markers in human cancer



The administration of chemotherapy either alone or in combination with radiotherapy is an important factor in reducing the mortality and morbidity of cancer patients. Resistance to both chemotherapy and radiotherapy represents a major obstacle to a successful outcome. The identification of novel biomarkers that can be used to predict treatment response would allow therapy to be tailored on an individual patient basis. Although the mechanisms are unclear, it is accepted that development of therapy resistance is a multifactorial phenomenon involving alterations in several cellular pathways. Proteome analysis methods are powerful tools for identifying factors associated with resistance to anticancer therapy because they facilitate the simultaneous analysis of whole proteomes. The current review describes the plethora of existing proteomic approaches and details the studies that have identified biomarkers that may be useful in the prediction of clinical response to anticancer therapy. Cancer 2006. © 2006 American Cancer Society.

Despite recent progress in detection and treatment, cancer remains a public health challenge. If the tumor is not too advanced, initial treatment often involves surgical resection of the primary lesion and associated lymph nodes. However, some cancer cells may remain despite surgical intervention. These cells have the potential to reestablish a tumor burden postsurgery. Cytotoxic chemotherapy is used to treat malignant disease in a number of differing scenarios. In some malignancies, such as hematological diseases and testicular cancer, chemotherapy is a curative treatment. In many common solid tumors, e.g., lung, breast, and colorectal cancers, chemotherapy is used as an adjuvant to surgical removal in order to combat occult micrometastatic disease that is present at the time of resection. Such an approach has been shown to prevent recurrence and improve survival in many solid tumors. In certain cases, when a tumor is too large to be resected, or if the patient wishes to have less radical surgery, chemotherapy may be administered before surgery in order to downstage the disease (neoadjuvant chemotherapy). In such cases it would be valuable to know whether the tumor is chemosensitive beforehand so as not to jeopardize the patient's chance of a curative operation. In patients with metastatic cancer, chemotherapy is often used to palliate symptoms and to improve both quantity and quality of life. In these circumstances the benefits to the patient are often quite marginal and may be outweighed by the toxicity of the treatment. Therefore, a pretreatment predictive test would be highly informative.

Resistance to anticancer agents represents a major obstacle to a successful outcome. Conventional anticancer therapy aims to induce apoptosis by the generation of a lethal burden of DNA damage. Although the mechanisms remain unclear, it is accepted that the development of therapy resistance is a multifactorial phenomenon that involves multiple alterations in several different cellular pathways. DNA repair pathways and altered apoptotic machinery have been implicated in therapy resistance. In addition, the overexpression of drug detoxification enzymes and drug transport pumps has also been reported along with modifications to the drug target. The identification of novel biomarkers that correlate with treatment response would allow therapy to be tailored on an individual patient basis. Ultimately, those patients unlikely to respond to a particular treatment strategy would be spared from serious life-threatening side effects for no therapeutic gain. Biomarkers may also provide information on new drug targets for future therapeutic intervention. Overcoming resistance to chemotherapy and radiotherapy would represent a major advance in the effective management of cancer today.

Many studies have been conducted to elucidate the mechanisms of therapy resistance and to identify predictive biomarkers. Immunohistochemistry has been used to correlate the expression of specific known proteins with resistance to anticancer therapy.1–3 Nix et al.3 used immunohistochemical assays to analyze the expression of apoptotic proteins in 124 patients with laryngeal cancer. Sixty-two patients who failed to respond to radiotherapy were matched for T-stage, laryngeal subsite, and smoking history to a group of 62 patients who were successfully cured by radiotherapy. Resistance to radiotherapy was associated with expression of the antiapoptotic markers Bcl-2 and Bcl-XL and with the loss of Bax expression. This suggests that an inappropriate apoptotic response may manifest clinically as resistance to radiotherapy. Although immunohistochemistry is useful, existing knowledge is required in order to select the target protein(s) to analyze. The target is selected based on the function of the protein and its association with a specific biological process, e.g., apoptosis. Standard immunohistochemistry is unable to identify new targets or establish relations between targets and is not amenable to high-throughput screening. In 1998, however, Kononen et al.4 introduced the “tissue microarray” for the rapid and efficient immunohistochemical analysis of multiple tissue samples simultaneously. This method involves obtaining cylindrical cores of tissue from archival paraffin-embedded specimens and arraying these samples in a master block. Sections, which contain up to 1000 tissue samples from individual tumors, can then be analyzed by standard immunohistochemistry.4 Using this method, an entire cohort of cases can be analyzed by performing immunohistochemical assays on just 1 or 2 master array slides. Each tissue specimen can then be linked with information such as survival data and treatment response data. Although this has improved the throughput of standard immunohistochemistry, only a single target (antibody) per slide can be assessed.

The use of global analytical techniques to study complex phenomena, such as resistance to anticancer therapy, overcomes many of these limitations. These techniques enable the simultaneous analysis of whole genomes and/or proteomes and potentially allow all genes and/or proteins that are associated with a specific disease phenotype to be identified. The major advantage being that the regulation of previously unknown genes and/or proteins can be implicated in a particular disease state.

Genomic Approaches

Cytogenetics techniques, such as multicolor fluorescence in situ hybridization (mFISH), spectral karyotyping (SKY), and comparative genomic hybridization (CGH) exploit advanced fluorescent technology to study the whole chromosomal complement of cells.5 Cytogenetic techniques are invaluable in the study of complex disease processes and have been employed to identify the genomic aberrations associated with both chemotherapy and radiotherapy resistance.6, 7 Such techniques, however, are compromised by their limited resolution and further studies are required in order to confirm the identity of genes that are associated with a particular phenotype. Recently, CGH has been applied to the rapidly growing field of microarray technology.8 Thousands of probes are printed at a high density onto glass microscope slides and 2 differentially labeled DNA pools are hybridized simultaneously to assess DNA copy number. Array-based CGH not only offers superior resolution over chromosome-based CGH, but also improves quantitative accuracy and dynamic range. Additionally, the aberrations can be directly mapped to the genome sequence.8 Wilson et al.9 used array-based CGH to analyze cisplatin resistance in testicular germ cell tumor cell lines. This study revealed that the overexpression of genes on chromosome 16q was associated with the development of cisplatin resistance. A further understanding of the genes in this chromosomal region may offer novel therapeutic targets and increase understanding of the mechanisms of cisplatin resistance.

Transcriptomic Approaches

At present, microarray-based technology is more commonly used for gene expression profiling, which explores the relative levels of RNA expression from thousands of known genes simultaneously. The aim is to identify changes in gene expression that are associated with a disease phenotype. Gene-expression profiling has been used to analyze both chemotherapy and radiotherapy resistance in various tumor types.10–15 In the largest of these studies, Kang et al.12 exploited the Affymetrix HG-U113A microarray to analyze acquired drug resistance in gastric cancer cell lines. Four 5-fluorouracil (5FU) resistant cell lines, 3 doxorubicin-resistant cell lines, and 3 cisplatin-resistant cell lines were established from 4 different gastric cancer cell lines. The group identified 250 genes that were found to be differentially expressed in 5FU-, doxorubicin-, or cisplatin-induced chemoresistant cell lines, when compared with their drug-sensitive parent cell lines. In addition, 8 candidate multidrug resistance genes including midkine (a heparin-binding growth factor) were associated with resistance to 2 or more of the chemotherapeutic agents.

Unfortunately, there is a common drawback to the global techniques that are based on the genome and the transcriptome. The proteins within a cell are responsible for key biologic processes and also make up the bulk of pharmaceutical targets. Unfortunately, the expression levels of mRNA and the corresponding protein are often not comparable and posttranslational modifications and alternative splicing events cannot be inferred from genomic technologies. Therefore, in order to gain a more comprehensive understanding of intricate biologic systems, the information extracted from genome studies must be complemented with information on the proteins themselves. Recent technological advances have enhanced the analysis of the human proteome.

Proteomic Approaches

A number of proteome analysis methods have been utilized in cancer research. These are described below, along with details of their use in identifying factors associated with resistance to anticancer therapy.

Two-Dimensional Gel Electrophoresis and Mass Spectrometry

Two-dimensional gel electrophoresis (2DE) remains the gold standard for the separation of proteins. First-dimensional separation is achieved by isoelectric focusing, which separates proteins on the basis of their charge. This is coupled with second-dimensional separation, which exploits polyacrylamide gel electrophoresis (PAGE) to separate proteins in accordance with molecular weight. Under optimal conditions, the expression pattern of several thousands of individual protein species can be defined simultaneously on a single 2D gel. The protein profiles can be compared using sophisticated software packages to identify those proteins that are differentially expressed between samples. By itself, 2DE is purely a descriptive technique and therefore must be coupled with analytical methods, such as mass spectrometry (MS), to identify those proteins of interest. MS provides structural information, such as peptide mass and amino acid sequence, which is then used to identify the protein by searching against nucleotide and protein databases.16 Current MS-based approaches involve the digestion and extraction of proteins from 2D gels using a sequence-specific protease such as trypsin. These specific proteolytically derived peptides are then characterized for protein identification. Mass spectrometers are capable of forming, separating, and detecting molecular ions on the basis of the mass-to-charge ratio (m/z). This requires that the peptides are ionized. Ionization techniques transfer biomolecules from the solid or liquid phase to the gas phase, making them amenable to MS measurement. Two common methods of peptide ionization for MS analysis are matrix-assisted laser desorption/ionization (MALDI) and electrospray ionization (ESI). Ionization in both of these methods ensues as a result of the addition of 1 or more protons. Thus, a peptide of molecular weight 1000 Da will have an m/z value of 1001 after ionization by the addition of a single proton and 501 with the addition of 2 protons (M+2H/2).

Matrix-Assisted Laser Desorption/Ionization (MALDI)

The sample to be analyzed is mixed with a small energy-absorbing matrix molecule, such as 2,5 dihydroxybenzoic acid or α-cyano-4-hydroxycinnamic acid, which absorbs light at a specific wavelength suited to the laser source (e.g., 337 nm for the N2 laser). The sample/matrix mixture is spotted onto a metal multiwell microtiter plate and allowed to air-dry to form a crystal lattice, in which the peptide sample is integrated. The plate is then systematically irradiated with a laser to convert the solid crystalline form to gas phase. The matrix chemicals absorb energy, which is subsequently passed to the sample peptides. After excitation, peptide ions are ejected from the target surface and are directed into a mass analyzer for detection. MALDI produces predominantly singly charged ions, which allows the m/z value to be determined.

Electrospray Ionization (ESI)

ESI-based MS is generally regarded as the most sensitive technique (subfemtomole level). The liquid phase sample enters the system through a microcapillary tube (usually from a high-performance liquid chromatography, HPLC, system) held at high voltage. As the flowstream exits the tube it sprays out in a fine mist of droplets, which contain peptide ions along with other components from the HPLC mobile phase (e.g., water and acetonitrile). The peptide ions are separated from the solvent components either by passing the droplets through a heated capillary or by passing a curtain of nitrogen across the spray. Subsequently, the desolvated ions are drawn into a mass analyzer. ESI produces primarily doubly charged ions, which allows for the simple calculation of the m/z value.

Mass Spectrometry Instrumentation and Protein Identification

MS instruments consist of 3 basic elements: an ionization source, a mass analyzer, and a detector. The mass analyzer uses a physical property, such as an electric or magnetic field or time-of-flight (TOF), to separate ions of a particular m/z value.17 Ions pass through the mass analyzer and are detected by an instrument, such as an electron multiplier, and the magnitude of current produced at the detector is used to determine the m/z value of the ion. MS data are recorded as “spectra,” which display ion intensity versus their m/z value. A robust instrument frequently employed in proteome studies combines a MALDI ionization source with a TOF mass analyzer (Fig. 1). In this instrument, gas phase ions from the MALDI ionization source are directed into a flight tube. The m/z value is measured by determining the time required for the ions to traverse the length of the flight tube and strike a detector. Mass resolution can be increased by including an ion reflector at the end of the flight tube. This is essentially an ion mirror that serves to increase the length of the tube and correct for small energy differences among ions. In many cases MALDI-TOF is the method of choice for proteome studies because it is a simple and reasonably sensitive technique (to the femtomole level). In addition, it offers the full automation of analysis and database searching. MS can generate 2 types of data, which can be used for protein identification. A characteristic mass spectrum is known as a peptide mass fingerprint (PMF), which is a list of masses for the peptides in a sample. The PMF is compared with the predicted masses of peptides from the theoretical digestion of all proteins in a database. If enough peptides from the real mass spectrum and the theoretical spectrum match in mass, the protein can be identified. Unfortunately, a single peptide is rarely unique to 1 protein, thus several peptides (>3) that are derived from the same protein are typically required for identification. PMF is currently the method of choice for identification because it combines a conceptually simplistic approach with robust high-throughput instrumentation (usually MALDI-TOF MS). Unfortunately, the ambiguity of protein identification limits the use of PMF. This is the result of peptide mass redundancy. A peptide of 5 amino acids can have the same mass as a different peptide by a simple rearrangement of the constitutive amino acids, e.g., GPLSV will have the same mass as PSGVL and VGPLS, and so on.16 Posttranslationally modified proteins also reduce the success of PMF. The peptides from a modified protein will not match the masses of the peptides from the unmodified protein in the database. Also, PMF is not effective with protein mixtures and protein identification is hindered when 2 or 3 proteins are present in a sample. The presence of contaminants such as keratin and peptides from the autolysis of trypsin may also be problematic. Moreover, not all proteins are amenable to identification by PMF alone. The full lengths of a large percentage of human proteins are not represented in databases. In addition, small proteins may not yield a sufficient number of peptides from the tryptic digest for unambiguous identification. In such cases, it is preferable to subject selected ions to further fragmentation, which can provide the amino acid sequence of the peptide. The amino acid sequence for a specific peptide can be deduced by a process called tandem mass spectrometry (MS/MS). A higher level of confidence can be assigned to protein identification when searching databases with MS/MS data. MS/MS combines 2 mass analyzers and a collision cell to collect structural information for individual peptides. In this method, the first mass analyzer acts to scan all the precursor ions from the ionization source. The spectrum is used to select those ions of a particular m/z value. Such ions can be isolated and are dissociated by a process known as collision-induced dissociation (CID). CID energetically activates ions to dissociate. The selected ions enter a collision cell and are subjected to low-energy collisions with neutral gas atoms, such as argon or nitrogen. As the ions become excited, covalent bonds fragment in a predictable manner. The fragmentation process predominates primarily along the peptide backbone (at or around the amide bond). If the charge is retained on the N-terminus, the fragments are designated b-ions. In contrast, if the charge is retained on the C-terminus, the fragments are designated y-ions. This results in a collection of b- and y-ions, whose m/z values are determined by the second mass analyzer.17 The molecular weight differences between adjacent b- and y-ions correspond to a specific amino acid and thus spell a partial amino acid sequence. This sequence can be combined with the mass of a peptide to search databases. The sequences in a database can be used to predict an expected pattern of fragmentation. This can be compared with the pattern observed in the spectrum and matched for identification.

Figure 1.

Matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) mass spectrometer. Ions are guided into the TOF mass analyzer, which measures the time taken for the ions to traverse through the flight tube and strike the detector.

The amino acid sequence of a peptide may also be determined using a classical MALDI-TOF MS equipped with an ion reflector by a process known as postsource decay (PSD). After leaving the ion source, gas-phase precursor ions enter a field-free region. In this field-free region the precursor ions decay to produce PSD ions. These PSD ions have the same velocity as their precursors; however, they have lower kinetic energies owing to their lower masses.18 In linear instruments, PSD ions strike the detector at the same time as the precursor ions, and therefore their mass cannot be analyzed. In contrast, an ion reflector acts as a mass analyzer for the PSD ions and separates ions based on differences in their kinetic energies. Unfortunately, the peptide fragmentation patterns are much less predictable than competing sequencing methods and other MS instruments (such as those combining an ESI source with triple-quadrupole and ion-trap mass spectrometers) are more accurate in obtaining the amino acid sequence of a peptide for protein identification.

Study of Resistance to Anticancer Therapy Using 2DE and MS (2DE-MS)

Historically, the use of 2DE to study resistance to anticancer therapy dates back to 1986, when Shen et al.19 employed this technique to determine the mechanisms of multidrug resistance in human cancer cells. The group analyzed 4 cell lines derived from the human KB epidermoid carcinoma cell line selected for resistance to high levels of colchicine, adriamycin, or vinblastine. All cell lines displayed cross-resistance to each of the other agents and various other unrelated drugs. This study was conducted before the technological advances that have allowed MS to become a routine tool in proteome studies, and therefore simply reports the descriptive changes associated with resistance to chemotherapeutic agents. Nevertheless, several specific changes were documented, some of which were common to all cell lines and others that were unique to specific cell lines. One such change detected by 2DE was the loss of a family of proteins in the molecular weight range of 70 to 80 kDa, with pI values of 4.8 to 5.0. Not only were these proteins lost in all resistant cell lines analyzed in comparison with parental cell lines, but also reappeared in a revertant derived from colchicine-resistant cells. This observation strengthened the association of the loss of this family of proteins with the development of multidrug resistance in KB cell lines.

More recently, however, in a time whereby such proteins can be routinely identified by MS, further studies have been conducted into resistance to anticancer therapy using cell lines (Table 1). In the latest of these studies, Shin et al.24 used 2DE-based comparative proteomics to screen for proteins responsible for 5FU resistance in human colon cancer cell lines. The data obtained from this study showed that the expression of the α-subunit of the mitochondrial F1F0-ATP synthase was lower in 5FU-resistant cells when compared with parent cells. Further analysis demonstrated decreased expression of other ATP synthase complex subunits, reduced ATP synthase activity, and reduced intracellular ATP content in 5FU resistant cells. This indicated that down-regulation of ATP synthase may lead to cellular events responsible for 5FU resistance. One possible explanation is that a reduction in ATP synthase results in a lower intracellular ATP level. ATP is required for the execution of apoptosis, and therefore reduced ATP levels may decrease the apoptotic potential of a cell.

Table 1. Summary of Studies Utilizing 2DE and MS for Analysis of Chemotherapy Resistance in Cell Lines
TumorDrugProteins Altered in Resistant CellsReference
Gastric cancerDaunorubicinAnnexin I 20
Gastric cancerMitoxantroneAnnexin I  
Pancreatic cancerDaunorubicinCofilin 21
Pancreatic cancerMitoxantroneCofilin  
  Epidermal Fatty Acid Binding Protein  
  Stratifin (14-3-3-σ)  
FibrosarcomaMitoxantroneRho-Guanine Dinucleotide Phosphate (Rho-GDP) Dissociation Inhibitor 22
Colon cancerMitoxantroneAdenine Phosphoribosyl Transferase 22
  Breast Cancer Specific Gene 1  
Colon cancer5-fluorouracilMetabotropic Glutamate Receptor 4 23
Colon cancer5-fluorouracil F1F0-ATP Synthase24
MelanomaVindesineTranslationally Controlled Tumor Protein 25
 CisplatinHuman Elongation Factor 1-δ  
 FotemustineTetratricopeptide Repeat Protein  
NeuroblastomaEtoposidePeroxiredoxin IdUTP Pyrophosphatase26
  β-Galactoside Soluble Lectin Binding Protein  
  Heat Shock Protein 27  
  Heterogeneous Nuclear Ribonucleoprotein K  
Gastric cancerCisplatin Pyruvate Kinase M227
Breast cancerMelphalanRetinoic Acid Binding Protein IICalreticulin28
  Macrophage Migration Inhibition FactorCyclophin A
   Heat Shock Protein 27 
Breast cancerDoxorubicinAnnexin ICatechol-O-Methyltransferase29
  Neuronal Ubiquitin Carboxyl Hydrolase  
  Isoenzyme L1  
  Glutathione-S-Transferase pi  
  Nicotinamide N-Methyltransferase  
  Interleukin-18 Precursor  

The aforementioned 2DE studies for analyzing chemoresistance exploited cell lines as a model system. In a recent study, Allal et al.30 used this comparative proteomics approach with clinical samples to identify proteins associated with radiotherapy resistance in rectal cancer. Biopsies were taken from 17 patients before radiation therapy. All patients received preoperative radiotherapy. Surgery was performed after 6 weeks and clinical response was assessed. Protein extracts from radiotherapy-sensitive and radiotherapy-resistant tumors were compared by 2DE and differential spots were identified using MALDI-TOF MS. Their results indicated that the expression of tropomodulin, heat shock protein 42, β-tubulin, annexin V, and calsenilin were statistically associated with radiotherapy resistance and keratin type I, notch 2 protein homolog, and DNA repair protein RAD51L3 were associated with radiosensitivity.

From the results of such studies, common proteins have been associated with resistance to anticancer therapy. Some of these include DNA repair proteins, molecular chaperones, signal transduction proteins, and drug detoxification proteins. Other proteins that appear to play a role in resistance to therapy include structural proteins, proteins with metabolic functions, and proteins exhibiting calcium-binding properties. However, without functional studies it remains to be seen whether these alterations are the cause or the result of resistance to anticancer therapy.

Difference In-Gel Electrophoresis (DIGE)

Unfortunately, 2DE remains difficult to perform due to the inherent reproducibility of the technique. Each gel runs slightly differently, which makes gel-to-gel comparison an arduous task. Recently, 2D difference in-gel electrophoresis (2D-DIGE) has been introduced, which is a technique designed to minimize gel-to-gel variations. Protein extracts from 2 different samples are covalently differentially labeled with a fluorescent dye. The labeled extracts are mixed and separated by 2DE on a single gel. Scanning the gel at different wavelengths allows differences in expression levels to be observed. This reduces the problems associated with gel matching because the 2 samples are run under the same conditions. Still in its infancy, to the best of our knowledge 2D-DIGE has not yet been exploited in the analysis of chemotherapy or radiotherapy resistance.

Quantitative Proteomics Using Labeling Techniques

Direct protein quantitation based on MS signals is challenging because of the nonlinear correlation between protein abundance and MS signal intensity.31 Techniques have since been developed to overcome this problem and involve the labeling of proteins before MS analysis. In this approach, a protein is labeled with an isotopic tag such that the protein from 1 cell state is labeled with the light variant of the tag and the protein from a second cell state is labeled with the heavy variant. The samples are combined and digested with a sequence-specific protease, such as trypsin. MS assessment of the heavy:light ratio enables a comparison of expression. One example of this mass tagging approach was developed in 1999 by Gygi et al.32 This technique is called isotope-coded affinity tagging (ICAT) and is an MS/MS-based approach designed to simultaneously quantitate and identify differentially expressed proteins. An ICAT consists of 3 functional elements: a thiol-specific group, which binds to cysteine residues of the protein, a light or heavy deuterated linker, and a biotin tag. Proteins from 2 different samples are labeled with a light or heavy ICAT reagent. After digestion the labeled peptides are isolated by their affinity for avidin and are analyzed by MS/MS. In full scan mode, ICAT pairs can be identified by looking for those peptides that differ in mass by a specific mass unit (as specified by the mass difference between stable isotopes) and the relative abundance can be determined by the heavy:light ratio. Differentially expressed proteins can be identified by searching databases with MS/MS data. Such mass tagging approaches involve the chemical labeling of proteins; however, a variant of this technique exploits the biosynthetic incorporation of stable isotopes (usually in the form of amino acids) into protein during cell growth. This is known as metabolic labeling and has previously been exploited in the analysis of chemotherapy resistance. Gehrmann et al.33 compared the protein expression patterns of MCF7 breast cancer cells resistant to adriamycin and to adriamycin and verapamil with the parental MCF7 cell line. The parental MCF7 cell line was grown in 13C6-arginine- and 13C6-lysine-enriched media resulting in the C-terminal labeling of all tryptic peptides. Results showed protein abundances to be distinctive in MCF7 cells resistant to adriamycin and those selected for resistance to both adriamycin and verapamil.

Antibody Microarrays

Antibody microarrays are a novel proteomic technique and provide a powerful technology for analyzing the expression of hundreds of proteins simultaneously. A high-precision robot is used to print hundreds of monoclonal antibodies at a high density on a glass slide in a format that is compatible with existing hardware and software tools for DNA microarrays.34 The surface of the slide is chemically modified to present functional groups for the covalent binding with antibodies allowing the antibodies to maintain their activity despite immobilization. Protein extracts from 2 different samples are differentially labeled with fluorescent dyes. The labeled extracts are then hybridized simultaneously to the antibody microarray. A fluorescence-based detection procedure, analogous to that used in gene expression profiling, is exploited, in which the immobilized antibodies are used to capture fluorescently labeled antigen.35 The relative ratios of red and green fluorescence at each spot allow the expression levels of specific proteins in each sample to be determined. Although still in its infancy, this technique has been exploited in cancer research.36, 37 In 2001, Sreekumar et al.36 successfully utilized antibody microarrays to monitor alterations of protein levels in LoVo colon cancer cells treated with ionizing radiation. In the near future, therefore, it is probable that this technique will provide novel biomarkers that are able to predict treatment response.

Surface-Enhanced Laser Desorption/Ionization-TOF MS (SELDI-TOF MS)

In recent years, surface-enhanced laser desorption/ionization time-of-flight (SELDI-TOF) MS has gained considerable attention. This technique couples array-based technology with MALDI-TOF MS. Proteins are selectively retained on a platform, which has been engineered with a chemical (e.g., anionic, cationic, hydrophobic, hydrophilic, or immobilized metal affinity) or biological (e.g., antibodies, antigen binding fragments such as scFv, or receptor) bait surface. Such varied surfaces allow proteins to be selectively retained on the SELDI platform based on the intrinsic properties of the proteins themselves and acts as a prefractionation step to allow for the detection of less abundant proteins. An energy-adsorbing molecule, such as α-cyano-4-hydroxy-cinaminic acid (CHCA), is added to the chip, which crystallizes on the chip surface. Ionization ensues by laser emission and the gas-phase ions are guided into the MS. A unique fingerprint results for each sample tested. Differential expression may be determined from these protein profiles by comparing peak intensity. Initial experiments aimed to produce a discriminatory proteomic pattern based on differential mass spectra to distinguish between 2 experimental populations. Here it was the pattern rather than protein identity that was deemed as the important factor. With the advance of technology, however, it is now possible to perform secondary processing of the proteins on the chip. SELDI-TOF technology can now be combined with tandem MS. A tryptic digestion of proteins is performed and the resulting fragments are analyzed to obtain sequence information, which can be used to identify the protein of interest. Mian et al.38 used SELDI-TOF MS to confirm that molecular profiling in conjunction with artificial neural network algorithms can be applied toward predicting the biological behavior of breast cancer cells to particular chemotherapeutic agents. Specific proteomic fingerprints that were indicative of chemotherapy-resistant and chemotherapy-sensitive cellular populations were established. Isolation and sequencing could result in the identification of those proteins that are associated with chemotherapy-resistant and chemotherapy-sensitive phenotypes.


Most of the proteomic studies described in this review involved the creation of chemotherapy- and/or radiotherapy-resistant cell lines, the protein profiles of which have been compared with parental cell lines to identify potential biomarkers. Cell lines are good in vitro models but findings do not always transfer to the in vivo clinical setting. Multicellular spheroid models, which go further toward a realistic 3D in vitro tumor model, may be 1 way forward.39 The validation of data derived from proteomic studies is also imperative. Global approaches using thousands of known targets, such as microarray-based techniques, are prone to false discovery and overinterpretation.40, 41 Protein identification of unknowns derived from proteomic approaches such as 2DE-MS adds a further level of potential error and overfitting, which has resulted in guidelines to standardize the quality of future proteomics data.42 The confirmation and validation of targets identified using proteomic screening should be performed using a second method such as Western blotting or immunocytochemistry and this may be combined with a complementary RNA-based transcriptomic screening approach.43 Once the identification and significant expression change of a target protein has been confirmed, then further detailed studies can ensue. Functional studies may include the in vitro manipulation of gene expression using specific pharmacological inhibitors, antisense RNA, RNA interference, or gene knockout experiments. The analysis of the functional pathway that has been implicated may reveal further insights into the mechanism of action, cross-talk between cellular pathways, and may also identify novel therapeutic targets. The confirmation of the clinical importance of any potential target using samples from cancer patients requires carefully designed experiments with adequate sample numbers with relatively homogeneous characteristics.44

Most of the studies described in this review have attempted to identify markers of therapy resistance by using cell line models rather than clinical samples. The analysis of primary tumor tissue or serum samples provides an alternative approach but this produces further technical problems. Tumors are heterogeneous and therefore microdissection may be required and more replicates must be performed in order to draw conclusive results. Clinical tissue samples are often small, especially in the case of predictive pretreatment biopsies, resulting in limited amounts of material for analysis. MS instruments are becoming increasingly more sensitive and this will allow smaller and smaller amounts of protein to be effectively analyzed in the future. Serum samples45–47 and fresh tumor samples48, 49 have been used in the identification of potential biomarkers for cancer screening, but their use in the identification of markers of response has not been reported. In recent years, advances in MS technology have involved the hybridization of instruments whereby sections of conventional tandem mass spectrometers have been either added or replaced with devices that provide different or superior performance characteristics and enhanced sensitivities. These advances have opened up the possibility of applying MS analysis directly to tumor tissue50 and the identification of proteins from formalin-fixed, paraffin-embedded archival tumor samples.51, 52 Additional improvements in all existing protocols will also be required, such that sample handling, sample preparation, and analysis are standardized to allow data comparison between laboratories.

Proteome analysis methods provide powerful tools for the investigation of complex biological phenomena such as resistance to anticancer therapy. Many different proteomic techniques are available to the researcher, each of which have their own advantages. Proteomic techniques have identified novel biomarkers, which have the potential to predict response to anticancer therapy; however, such markers require substantial further validation before being introduced to the clinic. Continuing improvements will also increase the use of these techniques in the study of therapy resistance. Anticancer therapy resistance is probably a multifactorial phenomenon and the increased understanding of the mechanisms involved will potentially pave the way for novel therapy regimes, new drug design, and the personalization of cancer treatment.