A real-time polymerase chain reaction assay for identification and quantification of Flavobacterium psychrophilum and application to disease resistance studies in selectively bred rainbow trout Oncorhynchus mykiss


Correspondence: Gregory D. Wiens, National Center for Cool and Cold Water Aquaculture, 11861 Leetown Rd, Kearneysville, WV 25430, USA. Tel.: (304) 724-8340; fax: (304) 725-0351; e-mail: greg.wiens@ars.usda.gov


Rapid detection and quantification of Flavobacterium psychrophilum, the causative agent of bacterial cold water disease (BCWD) in rainbow trout, are crucial to disease surveillance and encompass an essential component of BCWD research. Real-time, or quantitative polymerase chain reaction (qPCR) assays that have previously targeted the 16S rRNA gene of F. psychrophilum are complicated by polymorphisms and off-target amplification. Insignia primer and probe development software were used to identify a conserved single-copy signature sequence in the F. psychrophilum genome that codes for a hypothetical protein with unknown function. Primer and probes were used in a TaqMan qPCR assay that amplified 210 F. psychrophilum isolates with a lower limit of linear detection at 3.1 genome equivalents per reaction, with no amplification of 23 nontarget bacterial isolates. The assay was not inhibited by host spleen DNA or spleen homogenate. Methods were successfully applied to detect F. psychrophilum in rainbow trout from naturally occurring BCWD outbreaks and to quantify bacterial loads in experimentally infected rainbow trout. This assay will be applied to future studies to characterize disease pathogenesis in fish selectively bred for BCWD resistance.


Flavobacterium psychrophilum is an economically important pathogen of rainbow trout, Oncorhynchus mykiss, and is the causative agent of bacterial cold water disease (BCWD) (Nematollahi et al., 2003; Barnes & Brown, 2011). Disease prevention is difficult as there is no commercial vaccine for BCWD, and treatment of infection is confounded by a limited number of approved chemotherapeutants and demonstrated antibiotic resistance (Nematollahi et al., 2003; Barnes & Brown, 2011; Bruun et al., 2011).

Family-based fish breeding programs that select for disease resistance comprise an increasingly important element of sustainable and cost-effective aquaculture production. The National Center for Cool and Cold Water Aquaculture (NCCCWA, Kearneysville, WV) established a selective breeding program in 2005 to improve survival of rainbow trout to BCWD. Relative survival to intraperitoneal challenge with F. psychrophilum was found to be moderately heritable and resulted in development of differentially susceptible family lines designated ARS-Fp-R (resistant), ARS-Fp-S (susceptible), ARS-Fp-C (control), and ARS-Fp-C × R (a C × R cross) (Hadidi et al., 2008; Leeds et al., 2010). Current projects are underway to elucidate phenotypic differences between families that lead to differential survival to BCWD, and methods that allow rapid and accurate detection of F. psychrophilum are fundamental to BCWD research.

Previous conventional and nested PCR assays for F. psychrophilum have targeted the 16S ribosomal RNA gene (Toyama et al., 1994; Wiklund et al., 2000; Taylor & Winton, 2002), gyrase subunit genes (Izumi et al., 2005), and peptidyl-prolyl cis-trans isomerase C gene (Yoshiura et al., 2006). Two quantitative polymerase chain reaction (qPCR) assays have been described, which rapidly and quantitatively detect the 16S rRNA gene sequence (del Cerro et al., 2002; Oriuex et al., 2011). However, interpretation of assay results is complicated by six 16S rRNA gene copies (Duchaud et al., 2007) that can vary in sequence (Soule et al., 2005), and the reported off-target amplification from non-F. psychrophilum bacteria (Oriuex et al., 2011).

The objective of this study was to design and validate a qPCR assay based on a single-copy gene target common to F. psychrophilum strains.

Materials and methods

Flavobacterium psychrophilum primer and probe development

Insignia software (Phillippy et al., 2009) (http://insignia.cbcb.umd.edu/query.php) was used to select signature sequences within a 102–500 base pair size window for primer and probe development unique to F. psychrophilum strains CSF 259-93 (G.D. Wiens, unpublished data) and JIP02/86 (Duchaud et al., 2007). A total of 6 targets were suggested, ranging in size from 106 to 220 bp, and a 220-bp signature sequence located within a single-copy gene was chosen, which encodes a conserved hypothetical protein of unknown function. Nucleotide sequence uniqueness of the signature sequence was verified through a BLASTn search of the nonredundant nucleotide database (www.ncbi.nlm.nih.gov/blast/Blast.cgi). The forward primer (FpSigfwd, 5′ GGTAGCGGAACCGGAAATG 3′), reverse primer (FpSig rev, 5′ TTTCTGCCACCTAGCGAATACC 3′), and probe (FpSig probe, Tet-5′ CGCTTCCTGAGCCAGA-MGBNFQ3′) were designed using primer Express Software (Invitrogen) and are shown in Fig. 1. The melting temperatures (Tm) of the forward and reverse primers were 63.0 and 63.1 °C, respectively.

Figure 1.

Probe and primer locations are underlined within Flavobacterium psychrophilum strain CSF259-93 gene sequence RFPS00910 (GenBank accession no. KC192661). Comparison with homologous gene sequences from F. psychrophilum strain JIP02/86 (YP_001296792.1) and Flavobacterium branchiophilum strain FL-15 (YP_004843271.1). Predicted amino acid sequence is above the nucleotide sequence. Nucleotide differences within the F. branchiophilum sequence are underlined, and nonsynonymous amino acid changes are listed above the nucleotide sequence.

Analytical specificity

Flavobacterium psychrophilum isolate CSF 259-93, which was originally collected from a BCWD field case and frequently used in infectivity studies (Hadidi et al., 2008), was primarily used to test analytical specificity and sensitivity. An additional 209 F. psychrophilum isolates cultured from various field cases and identified by F. psychrophilum16S rRNA gene-specific PCR (Wiklund et al., 2000) were also examined for qPCR assay specificity. Twenty-three ecologically and taxonomically related bacterial isolates, obtained from farmed rainbow trout and identified by 16S sequencing, were used as nontarget bacterial isolates (Table 1). Bacteria had been maintained at −80 °C in TYES media supplemented with 10% (v/v) glycerol. All bacterial cultures described were performed on TYES media for 5 days at 15 °C. Cultured bacteria genomic DNA was extracted using a QIAprep Spin Miniprep kit (Qiagen) and eluted in 50 μΜ Tris/EDTA. Isolated DNA was quantified using a Nanodrop ND-1000 spectrophotometer (Thermoscientific).

Table 1. Nontarget bacterial species examined by qPCR to test assay specificity
Isolate IDYearTissueLocationGenBank accession no.ID%, Top-BLAST hit (accession no.)qPCR detection
  1. Bacterial isolates were previously collected from rainbow trout tissue and identified by 16S sequencing. GenBank accession number of each isolate is listed as well as top-BLAST hit.

  2. a

    Yellow-pigmented bacteria.

ARS-137-11a2011GillWV, USA KC192662 99% Flavobacterium branchiophilum FL-15 (FQ859183.1)
ARS-124-11a2011KidneyID, USA JX827612 99% Flavobacterium psychrolimnae strain NW109 (JF915319.1)
ARS-150-11a2011GillUT, USA JX827619 99% Flavobacterium aquatile strain NBRC 15052 (AB517711.1)
CSF-031-10a2010KidneyID, USA JX827625 99% Flavobacterium columnare strain XF (GU296112.1)
CSF-298-10a2010KidneyID, USA JX827623 99% Flavobacterium columnare strain XF (GU296112.1)
CSF-321-10a2010KidneyID, USA JX827624 98% Flavobacterium cheniae strain NJ-26 (NR_044198.1)
ARS-190-11a2011SkinID, USA JX827630 98% Flavobacterium sp. NBRC 101304 (AB681441.1)
ARS-151-11a2011GillUT, USA JX827620 99% Flavobacterium sp. HME6123 (HQ891304.1)
ARS-125-11a2011SpleenID, USA JX827613 99% Flavobacterium sp. HME6012 (HM149211.1)
ARS-135-11a2011SpleenID, USA JX827614 98% Flavobacterium sp. HME6012 (HM149211.1)
ARS-136-11a2011GillWV, USA JX827615 99% Flavobacterium sp. 1084B-08 (HE612093.1)
ARS-168-11a2011KidneyNC, USA JX827621 99% Sphingomonas sp. TSBY-38 (DQ151830.1)
ARS-171-11a2011SpleenNC, USA JX827622 99% Sphingomonas aerolata strain R-36940 (FR691420.1)
ARS-148-11a2011SpleenUT, USA JX827618 99% Massilia sp. Hp37 (JN637332.1)
ARS-188-11a2011GillID, USA JX827629 100% Chryseobacterium indoltheticum strain LMG 4025 (NR_042926)
ARS-145-11a2011SpleenUT, USA JX827616 99% Chryseobacterium jeonii, strain R-36526 (FR682715.1)
ARS-123a-11a2011KidneyID, USA JX827610 99% Exiguobacterium undae (AB334767.1)
ARS-123b-112011KidneyID, USA JX827611 99% Epilithonimonas tenax strain EP105 (NR_041912)
ARS-183-112011SpleenWV, USA JX827626 99% Aeromonas sobria, strain JCM 2139 (AB472942.1)
ARS-185-112011SpleenWV, USA JX827627 99% Comamonas sp. 12022 (AF078773.1)
ARS-187-112011GillID, USA JX827628 99% Brachybacterium tyrofermentans strain, CNRZ 926 (NR_026272.1)
ARS-146-112011SpleenUT, USA JX827617 99% Pseudomonas trivialis strain, KOPRI 25674 (HQ824913.1)
CSF007-821982UnknownID, USA- Yersinia ruckeri biotype 1 (Evenhuis et al., 2009)

Polymerase chain reaction was performed using hydrolysis probes (TaqMan®; Applied Biosystems) in 15 μL reaction mixtures consisting of 7.5 μL TaqMan Universal Master Mix II + UNG (Life Technologies), 0.67 μM forward and reverse primers, 0.17 μM probe, 1.5–4.5 μL sample, and nuclease-free water added to volume. Between 15 and 30 ng of bacterial DNA template and an average of 375 ng of input DNA from trout tissue samples were tested per reaction. The input DNA varied between trout DNA samples due to variation in DNA extraction and tissue availability; thus, results were normalized to 100 ng of sample DNA. Samples were run in duplicate with water used as a no-template control in 384-well full-skirt PCR plates (Phenix Research) covered with TempPlate RT® adhesive film (USA Scientific). An ABI 7900HT real-time thermal cycler (Applied Biosystems) was used to amplify the target sequence using absolute quantification with the machine's default program of initial denaturation for 15 min at 95 °C, 40 cycles for 15 s at 95 °C and 1 min at 60 °C.

Sensitivity, linearity and repeatability

Analytical specificity, linearity and repeatability of the assay were tested using serial, 10-fold dilutions of CSF 259-93 between 3.1 × 107 genome equivalents (GE) and 0.3 GE per reaction. GE were calculated based on the F. psychrophilum genome of 2 900 735 nucleotides, with a 32.5% cytosine-plus-guanine content and molecular weight of 895 884 304 Da (Wiens et al., unpublished data). The dilution series within the linear range was repeated with each sample set as a positive control to confirm assay reproducibility. Seven independent standard curve assays were run on at least three separate days. Standards were assayed in triplicate for several experiments; however, reproducibility was judged to be sufficient using duplicate wells. Assay amplification efficiency was determined from the slope of the standard curve, as previously described (Bustin et al., 2009). For assays that quantified F. psychrophilum DNA in fish tissue samples, a standard curve of purified CSF 259-93 bacterial genomic DNA was used to convert cycle quantification (Cq) into log10 GE 100 ng−1 of input sample DNA.

Quantification limit and test for inhibition by factors in spleen tissue

Assay performance was evaluated in the presence of splenic DNA and spleen homogenate to determine the quantification limit and to examine whether host DNA or tissue inhibits low-level quantification of bacteria. qPCR was performed directly on purified CSF 259-93 DNA serially diluted tenfold from 3 × 106 ng to 30 ng μL−1 in 40 ng μL−1 spleen DNA or dH20 to serve as controls. To examine the inhibitory effect of host tissue on DNA extraction and qPCR, cultured CSF 259-93 bacteria at 0.5 optical density were diluted 100 to 10−6 in PBS and added in a 1 : 1 volume ratio with spleen homogenate prepared from 10 mg of tissue diluted 1 : 10 (w/v) in filter-sterilized PBS using a mini-bead beater (Biospec Products). Dilutions were performed in quadruplicate and separately extracted. Each dilution sample was run in duplicate. Bacterial concentrations were confirmed by plating 100 μL of the homogenate on TYES media in triplicate and counting the 10−5 dilution. The quantification limit for the procedure was examined by plotting Cq against bacterial concentration (log10[colony-forming units (CFU) rxn−1]). The CFU value of the last linear point on the standard curve was determined to be the assay limit. Reaction efficiency was calculated from the slope of the line, as described previously.

Diagnostic specificity and sensitivity

qPCR was performed on tissues collected from naturally occurring BCWD outbreaks and from two separate BCWD infectivity trials at the NCCCWA, to examine diagnostic specificity and sensitivity of the assay.

Splenic and renal tissue were collected for DNA extraction from 216 rainbow trout mortalities from three trout-rearing aquaculture facilities believed to be affected by BCWD. Fifty-two of these fish were sampled for bacterial culture based on clinical appearance of BCWD. Additionally, 137 rainbow trout that were associated with BCWD outbreaks but were apparently healthy, were collected approximately 3 weeks postinitial outbreak and euthanized with tricaine methane sulfonate MS-222 (Sigma) for bacterial culture and qPCR. DNA was extracted using the DNeasy Blood and Tissue Kit (Qiagen) from 10 mg frozen (−80 °C) spleen or kidney or 100 μL tissue homogenate (combination of spleen and kidney) and eluted in 200 μL Tris/EDTA. Tissue homogenates, spectophotometry and qPCR were performed as described previously. Cultured bacteria that were preliminarily identified as F. psychrophilum based on characteristic culture morphology and appearance on TYES media were restreaked and confirmed by F. psychrophilum-specific PCR (Wiklund et al., 2000).

Artificially infected rainbow trout were examined as part of an ongoing selective breeding program to develop a genetic-line with increased BCWD resistance. The first experimental trial consisted of five 400-g rainbow trout from each ARS-Fp-R and ARS-Fp-S line. Fish were challenged by intraperitoneal injection with 9.6 × 106 CFU per fish of F. psychrophilum CSF 259-93, as determined by plate count, or PBS, in a volume of 500 μL. Fish were euthanized day 5 postinfection with 250 mg L−1 MS-222. Harvested spleens were divided and splenic bacterial loads determined by qPCR or plate count. Correlation between qPCR and plate counts was examined, and PCR amplification efficiency was calculated, as described previously.

A second set of experiments examined mortality differences in selectively bred rainbow trout belonging to four genetic lines, ARS-Fp-R (pooled from n = 43 families), ARS-Fp-S (pooled from n = 11 families), ARS-Fp-C (pooled from n = 9 families), and an ARS-Fp-C × R cross (pooled from n = 10 families). Two hundred, approximately 3.4 g, fish from each line were challenged with 5.8 × 106 CFU per fish F. psychrophilum strain CSF 259-93 in chilled PBS or PBS, in a volume of 25 μL. Simultaneously, 30 families of the ARS-Fp-R line fish (n = 2234 fish) were challenged with the same dose as the pooled families to validate family phenotype. A total of 40 fish per tank were supplied with 2.4 L of 13 °C flow-through water. A combined total of 40 mortalities from the challenge experiments were collected on day 5 (n = 10 R line, n = 4 C × R line, n = 4 C line, and n = 2 S line) and day 9 (n = 7 R-line, n = 4 C × R line, n = 5 C line, and n = 4 S line) postchallenge to assess application of the assay in examining disease pathophysiology in fish selectively bred for BCWD resistance. One ARS-Fp-R line mortality (day 5) was excluded from analyses due to advanced postmortem autolysis. DNA was extracted from harvested spleens, and qPCR was performed as described previously.

Statistical analyses

Statistical analysis and graphing were performed using graphpad prism version 5.0. All Cqs were normalized to DNA concentration and CFU and GE data were log10-transformed before statistical analysis. The relationship between GE and CFU for experiment two was analyzed through Pearson correlation. A one-way anova was used to compare bacterial loads in differentially susceptible family lines in the second experimental trial. All statistical analyses were run with a significance level of < 0.05.


Analytical specificity

All 210 F. psychrophilum isolates examined were positive by qPCR, with a mean Cq of 17.12 and standard deviation (SD) of 1.5. Results were consistent over multiple reactions and different days. There was no measurable amplification from the 23 non-F. psychrophilum bacterial isolates tested (Table 1).

Sensitivity, linearity and repeatability

The sensitivity and linear range of the assay were evaluated by plotting GE vs. Cq of serially diluted CSF 259-93 DNA. A positive linear relationship was demonstrated between 3.1 GE and 3.1 × 106 GE (Fig. 2), indicating a linear dynamic range for the assay covering seven orders of magnitude with a mean amplification efficiency of 101.6 ± 1.6% SD (n = 7 experiments). The assay was repeatable both within and among runs as the coefficient of variation was less than 5% within the linear range of the standard curve (Table 2).

Table 2. Repeatability of qPCR assay
GEInterrun CVIntrarun CV
No. exp.Mean CqSDCV (%)No. samplesMean CqSDCV (%)
  1. qPCR was performed on duplicate standards run on multiple days to measure interrun variation with known dilutions of Flavobacterium psychrophilum CSF 259-93. Intrarun variation was measured in triplicate in a single experiment. Mean cycle threshold (Cq) and SD were determined at each dilution, and results were compared using the coefficient of variation (CV).

  2. a

    Dilutions were determined to be beyond assay linearity.

Figure 2.

Standard curve showing linearity of the qPCR assay for known concentrations of Flavobacterium psychrophilum CSF 259-93. Data points represent the average cycle threshold (Cq) at each dilution for F. psychrophilum serially diluted tenfold between 3.1 × 106 GE and 3.1 GE, based on seven independent experiments. Error bars denote standard deviation. Mean efficiency of the reaction was 101.6 ± 1.6% SD (R2 = 0.999).

Quantification limit and test for inhibition by factors in spleen tissue

The quantification limit of bacteria added to spleen homogenate, measured using 1.5 μL of a 200-μL DNA elution, was 486 ± 146 SD CFU (n = 3). Mean amplification efficiency of purified F. psychrophilum CSF 259-93 DNA serially diluted in spleen DNA was 89.4 ± 2.0% SD. Amplification efficiency from CSF 259-93 bacteria diluted in spleen homogenate was 95.3 ± 2.9% SD. All efficiencies were within the acceptable range for qPCR standards (Purcell et al., 2011).

Diagnostic specificity and sensitivity

All mortalities examined from putative BCWD farm outbreaks that were culture positive (28/52, 54%) were also qPCR positive (mean GE 130 528 ± 106 583 SD). Flavobacterium psychrophilum was detected by qPCR in 24 dead fish that were culture negative (mean GE 106 492 ± 49 595 SD) and in 132/144 (92%) mortalities from which only qPCR was performed (mean GE 111 242 ± 95 804 SD). Bacterial DNA was detected in an additional seven samples, but Cq values were beyond the linear range of the assay (Cq > 34.6).

Five of the 137 subclinical rainbow trout harvested live from the end of BCWD outbreaks were culture positive. Of these, two were qPCR positive (mean GE 8786 ± 12 335 SD), one sample had detectable bacterial DNA beyond the linear range of the assay, and two were qPCR negative. Purified bacterial DNA isolated from cultures of the latter was tested by qPCR and yielded a typical signal of Cq 15.8 SD 0.11, indicating that failure to detect was not due to lack of assay specificity. Flavobacterium psychrophilum was detected in an additional 13 fish by qPCR that were negative by culture (mean GE 58 ± 65 SD) and an additional five samples measured beyond the linear range of the assay.

In the first laboratory challenge experiment, F. psychrophilum was detected in the spleens of 6/10 inoculated fish by qPCR and culture. All PBS-injected fish were negative by qPCR and culture. There was significant correlation between GE and CFU (Fig. 3), indicating agreement between the two methods for bacterial quantification (Pearson correlation, r = 0.83, < 0.05). Amplification efficiency was 94.9%, which is similar to the 95.3% efficiency observed when the assay performance was tested in vitro for inhibition by spleen homogenate.

Figure 3.

Comparison of qPCR and culture methods to quantify splenic bacterial loads in six rainbow trout infected intraperitoneally with Flavobacterium psychrophilum CSF 259-93. Positive correlation between GE (GE 100 ng DNA−1) and CFUs (CFU mg spleen−1) indicate agreement between the two methods for bacterial quantification at day 5 postinfection (Pearson's correlation, r = 0.83, < 0.05).

To determine whether bacterial loads in four selectively bred lines of rainbow trout were different at time of death, qPCR data were collected from 39 mortalities in experiment two. No differences in bacterial load were found between mortalities collected on days 5 and 9 postchallenge and thus data were pooled (students t-test, = 0.19). No significant differences in log10 GE were found among the four differentially susceptible genetic lines (1-way anova,= 0.90) (Fig. 4). The average GE for all mortalities was 127 800 ± 140 900 SD (n = 39). Nineteen of 20 PBS-injected fish were negative by qPCR, and the one remaining fish had Cq of 38.3 that was beyond the linear range of the assay.

Figure 4.

Comparison of Flavobacterium psychrophilum (Fp) splenic loads as measured by qPCR, in 39 intraperitoneally challenged fish mortalities. Rainbow trout were bred for different susceptibilities to BCWD. The lines are resistant (ARS-Fp-R), resistant × control (ARS-Fp-R × C), control (ARS-Fp-C), and susceptible (ARS-Fp-S). Horizontal black line is the mean.


The signature sequence for F. psychrophilum was selected among multiple suggested targets, as it represented a single-copy gene and was present in the genomes of both F. psychrophilum strains examined. Consistent analytical specificity across 210 F. psychrophilum isolates and lack of detection in 23 non-F. psychrophilum bacterial isolates obtained from rainbow trout farms support its use as a diagnostic target.

Assay inhibition has been described in a number of qPCR assays performed from tissue, including for F. psychrophilum (del Cerro et al., 2002; Opel et al., 2009). Results from this study indicate that neither bacterial DNA extraction nor amplification was preferentially inhibited by splenic host compounds. Splenic tissue was used in these experiments based on prior experimental and diagnostic assays that identify the spleen as a major organ affected by BCWD (Nematollahi et al., 2005). We expect similar amplification efficiency to be present in other organs affected as well, but this remains to be validated.

Diagnosis of BCWD can be complicated as F. psychrophilum is difficult to culture from infected fish (Michel et al., 1999) and can be inhibited by fungi and less fastidious bacteria when culturing from dead fish. Cultures taken from frozen samples can produce false-negative results due to decreased bacterial viability (Ewald & Eie, 1992). In this study, diagnostic sensitivity of qPCR is suggested to be greater than that of culture methods in fresh mortalities and frozen-stored samples, as bacterial antigen was detected in all culture-positive and culture-negative fish tested with clinical signs of BCWD. Additionally, 132/144 (92%) of fish tested only by qPCR were positive, demonstrating the assay's efficiency in detecting F. psychrophilum in samples from BCWD-related mortalities.

Bacterial loads quantified by culture and qPCR from splenic tissue of subclinically infected experimental fish were positively correlated, and qPCR detected F. psychrophilum in 9/12 (75%) culture-positive samples from euthanized experimental and farm fish. Positive qPCR from subsequently cultured samples indicates that assay failure was not related to inability to identify these isolates specifically, but rather unequal bacterial distribution in tissue or sensitivity of the assay, as only few colonies were observed at the lowest plated dilution on TYES agar.

Bacterial loads were not significantly different in experimental challenge of rainbow trout selectively bred for divergent BCWD susceptibilities. Results suggest that for acute mortalities, bacterial burden may not directly influence survival differences between genetic lines. However, considerable variation was observed within lines, as was observed with naturally infected fish, and it will be of interest to more precisely determine the kinetics of organ-specific bacterial loads after laboratory challenge. The development of this qPCR assay will additionally be important for evaluating farm trials of ARS-Fp-R line fish and will aid diagnostic investigation of BCWD.


We acknowledge technical contributions from Joel Caren and Jessica Thompson and the receipt of strains from Drs. Scott LaPatra, Doug Call and Tim Welch. The authors would like to thank Dr. Esteban Soto and anonymous reviewers for their critical review of the manuscript. This work was supported by Agricultural Research Service CRIS Project 1930-32000-005 ‘Host-Pathogen and Environmental Interactions in Cool and Cold Water Aquaculture’ and Agriculture and Food Research Initiative competitive grant no. 2012-67015-30217 from the USDA National Institute of Food and Agriculture. The authors have no conflict of interest to declare. Mention of trade names or commercial products in this publication is solely for the purpose of providing specific information and does not imply recommendation or endorsement by the U.S. Department of Agriculture. USDA is an equal opportunity employer.