Validation of commercial ELISAs for quantifying anabolic growth factors and cytokines in canine ACD-A anticoagulated plasma

Abstract

Platelet-rich plasma has been studied extensively in dogs, but validation of enzyme-linked immunosorbent assays (ELISAs) for quantifying anabolic growth factors and inflammatory cytokines in canine plasma prepared with citrate-based anticoagulants is not available. We performed a validation of commercial ELISAs for transforming growth factor–beta 1 (TGF-β1), platelet-derived growth factor–BB (PDGF-BB), vascular endothelial growth factor (VEGF), tumor necrosis factor–alpha (TNF-α), and interleukin–1 beta (IL-1β) for use with canine plasma prepared with acid–citrate–dextrose, solution A (ACD-A). Platelet-poor plasma (PPP) anticoagulated with ACD-A as well as PPP anticoagulated with ACD-A and spiked with the relevant canine recombinant proteins were evaluated with each ELISA to calculate the efficiency of spike recovery. Replicates of the spiked PPP were also assessed in 2 additional assays to quantify intra-assay and interassay precision. The efficiency of spike recovery was within 75–125% of the expected concentration for the TGF-β1, PDGF-BB, and VEGF ELISAs. The intra- and interassay variability were <25% for the TGF-β1, PDGF-BB, VEGF, and TNF-α ELISAs. The TGF-β1, PDGF-BB, and VEGF ELISAs demonstrate acceptable efficiency of spike recovery and intra- and interassay variability, whereas the TNF-α and IL-1β ELISAs did not meet industry standards of performance with ACD-A anticoagulated canine plasma.

Introduction

Platelet-rich plasma (PRP) contains an increased concentration of platelets in comparison with the whole blood from which PRP is prepared.^3,5 PRP is commonly injected into sites of musculoskeletal injury in human and veterinary patients in an effort to facilitate tissue healing via platelet α-granules that produce and release anabolic growth factors such as transforming growth factor–betas (TGF-β), platelet-derived growth factors (PDGF), and vascular endothelial growth factor (VEGF).^2,3 Although these growth factors promote cell proliferation and maturation in vitro and may stimulate tissue healing in vivo, results with investigational studies of PRP use in human and veterinary subjects provide inconsistent results.^{1,4,8,10,12,15,16}

Variability in the different PRP preparations investigated is one potential explanation for inconsistent therapeutic results when evaluating the efficacy of PRP.^18,20 Studies characterizing multiple human, equine, and canine PRPs have all demonstrated substantial variability in the cellular composition of different PRPs from the same species, including in platelet, leukocyte, and erythrocyte concentrations.^7,9,13,23 In addition, studies of human and equine PRPs confirm notable variability in the growth factor composition of different PRPs.^19,20,23 The variability in growth factor delivery is in part attributable to the cellular composition of the PRP.^7,11,18 However, growth factor production is also likely influenced by intentional exogenous platelet activation using substances to cause their degranulation and release of growth factors.^19,20 Ultimately, further study is needed to better understand how cellular composition of the PRP, different methods of PRP preparation, and intentional PRP activation influence growth factor release.

The elution of TGF-β1, PDGF composed of 2 B chains (PDGF-BB), VEGF, and other anabolic growth factors has been quantified from 2 different canine PRP preparations using commercial enzyme-linked immunosorbent assays (ELISAs).^17,22 However, the ELISAs used in those studies were not validated for use with canine plasma samples containing citrate, an acknowledged limitation of one of those studies. ¹⁷ This is relevant because most PRPs, including the PRPs in the 2 previous studies, are prepared using citrate-based anticoagulants. Acid–citrate–dextrose, solution A (ACD-A) is the anticoagulant used most commonly in the preparation of canine PRP, and sodium citrate is the second most commonly used anticoagulant. ⁷ As a result, the growth factor concentrations of canine PRPs remain somewhat unclear.

Although greater characterization of growth factor and cytokine concentrations of different PRPs is needed, we are unaware of ELISAs that are validated for quantification of anabolic growth factors and inflammatory cytokines with citrated canine plasma samples. Accordingly, we performed a validation of TGF-β1, PDGF-BB, interleukin–1 beta (IL-1β), VEGF, and tumor necrosis factor–alpha (TNF-α) ELISAs for use with canine ACD-A anticoagulated plasma samples. Our specific goals were to assess the reproducibility of the parameters of the 4-parameter logistic (4PL) standard curves, quantify intra- and interassay precision, and quantify the efficiency of recovery for spiked samples. We hypothesized that there would not be significant differences among the standard 4PL curves generated from the different assays, that intra- and interassay coefficients of variation would be <20%, and that the efficiency of recovery would be within 20% of the ideal recovery values of 100%.

Materials and methods

Animal use was approved by the Institutional Animal Care and Use Committee of the University of Georgia. Whole blood was collected from 8 skeletally mature Beagles in order to prepare the PRP. Each Beagle was sedated with nalbuphine HCl ^a (0.5 mg/kg) and dexmedetomidine hydrochloride ^b (0.005 mg/kg) given intravenously. A 5-cm, 18-ga intravenous catheter was placed in the jugular vein. Syringes were preloaded with 7.15 mL of ACD-A and then filled with 55 mL of whole blood. The ACD-A used in our study contained (per 10 mL) 0.073 g of citric acid (anhydrous), 0.22 g of sodium citrate (dehydrate), and 0.245 g of dextrose (monohydrate). The anticoagulated whole blood was mixed on a blood tube rocker for 15 min. The whole blood was then spun for 10 min at 1,000 × g. ^c After this first spin, the plasma was transferred to a new tube and spun again with the same parameters to pellet any remaining cellular contents. The pellets were discarded. The remaining platelet-poor plasma (PPP), made from different dogs, was pooled, mixed for at least 5 min on a blood tube rocker, and platelets, red blood cells (RBCs), and leukocytes were counted ^d to confirm that negligible cellular components remained. One-milliliter aliquots of the PPP were frozen at −80°C so that there was 1 pool (in aliquots) of ACD-A anticoagulated PPP that had negligible cellular content. PPP was prepared and used for the validation because manufacturer instructions recommend additional centrifugation of all plasma samples to create PPP for use with ELISAs so as to mitigate the likelihood that cellular contents or debris interfere with assay performance. ²⁰

Five different commercial canine ELISAs were used to quantify growth factors: TGF-β1, ^e PDGF-BB, ^f IL-1β, ^g VEGF, ^h and TNF-α. ⁱ Briefly, 4 runs were performed with each ELISA. In run 1, the concentration of the relevant protein was quantified in duplicate in the PPP sample. This run was also used to quantify the concentration of such protein in PPP that was spiked with a canine recombinant protein in 30 replicates. From these data, the efficiency of spike recovery was calculated. The 30 replicates of spiked PPP from run 1 as well as 16 replicates of spiked PPP from runs 2 and 3 were used to assess intra- and interassay precision. In run 4, a standard curve was generated using the same canine recombinant protein used to generate a standard curve in each of the previous 3 runs. The standard curves generated with each of these 4 runs were compared to assess the repeatability of the standard curves.

With regard to specific details of ELISA performance, PPP assessed using the VEGF, IL-1β, and PDGF-BB ELISAs was diluted 2-fold, 5-fold, and 20-fold, respectively, prior to assessment. PPP assessed using the TNF-α ELISA was run undiluted based on manufacturer’s instructions. For the TGF-β1 ELISA, the PPP was acid-activated and then diluted prior to assessment according to manufacturer’s instructions. The PPP used with TGF-β1 ELISA was acid-activated by: 1) addition of 10 μL of 1N HCl to 40 μL of sample; 2) incubation for 10 min at room temperature; and 3) addition of 10 μL of 1.2N NaOH/0.5N HEPES (N-2-hydroxyethylpiperazone-N-2-ethanesulfonic acid). The activated PPP was then diluted 60-fold according to manufacturer’s instructions.

After dilution and activation (if relevant), aliquots of the PPP were subsequently spiked to a concentration of 250 pg/mL, 313 pg/mL, 1,022 pg/mL, 250 pg/mL, and 62.5 pg/mL with recombinant canine TGF-β1, VEGF, IL-1β, PDGF-BB, and TNF-α, respectively, for use in assay 1. PPP was spiked to the same concentration for use in runs 2 and 3 except for the TNF-α and VEGF ELISAs. For runs 2 and 3, the PPP used with the VEGF ELISA was spiked to 600 pg/mL, and, for TNF-α, the PPP was spiked to 125 pg/mL. The canine recombinant standard included with each ELISA was used for such purpose with the exception of the PDGF-BB ELISA. The PDGF-BB kit was a human ELISA kit with a human recombinant standard. A canine recombinant protein ^j standard was obtained separately for this ELISA and used to generate the standard curve as well as to spike the corresponding PPP samples. These same canine recombinant proteins were reconstituted, diluted, and serial dilutions run in duplicate to generate the standard curves with performance of each assay. For every assay, optical density was measured ^k at 450 nm and 570 nm for each sample and the value at 570 nm subtracted from that obtained at 450 nm to correct for any optical impurities in the plates. Runs 2 and 3 were run identically, side by side, on the same day, whereas runs 1 and 4 were performed on separate days from each other and from performance of run 2 and run 3. The same operator performed all of the ELISAs in our study.

Concentrations of all samples were calculated based on 4PL standard curves ^l generated using the standard run with that assay. Further, all protein concentrations were based on the mean optical density of the 2 replicates that were immediately adjacent to each other in each assay. As a result, 15 replicate measurements from run 1, and 8 replicate measurements from runs 2 and 3, were acquired for spiked plasma samples.

A nonlinear regression model ^m was used to quantify and compare the 4 parameters from the 4PL curves. Pairwise comparisons were performed of the 4 parameters (a–d) between the 4 different standard curves for each ELISA. The model generated estimates for the 4 parameters (a–d) and the differences between these parameters for the different standard curves generated from each of the 4 runs. If the 95% confidence interval for the estimated difference between the parameters (a–d) from 2 different standard curves did not include zero, the parameter was considered significantly different between the 2 standard curves. With 4 parameters compared between each pair of runs, this resulted in 24 pair-wise comparisons among the 4 runs for each ELISA. If the algorithm for this model would not converge, graphical exploration of the data and removal of outliers was performed and the model repeated until convergence was achieved.

The mean protein concentration, the standard deviation, and the coefficient of variation (CV) for the ACD-A anticoagulated spiked samples were calculated for each assay to characterize the intra-assay accuracy and precision. In addition, the protein concentrations for all spiked samples were pooled from all 3 runs, and the mean, standard deviation, and CV were calculated to characterize the interassay precision.

The difference between the protein concentration for the unspiked PPP and the corresponding spiked PPP was calculated for run 1. The ratio of this difference to the expected spiked concentration was calculated and expressed as a percentage to represent the efficiency of spike recovery. If the protein concentration in the unspiked sample was below the lower limit of quantification, then the concentration was considered 0 pg/mL.

Results

A complete blood count of the pooled PPP sample confirmed negligible cellular contents with cell counts of: platelets = 9 × 10⁹/L; leukocytes = 0.1 × 10⁹/L; and RBCs = 0 × 10⁹/L. The industry-accepted standard for the efficiency of spike recovery using immunoassays is 80–120% of the actual concentration or up to 75–125% of the actual concentration if measuring values close to the upper or lower limits of quantification (ULOQ, LLOQ).^6,14 The TGF-β1 and PDGF-BB as well as the VEGF assays met this latter standard of performance (Table 1). Conversely, this standard of performance was not met with the TNF-α and IL-1β ELISAs with spike recoveries of 65.3% and 144.4%, respectively (Table 1).

Table 1.

The difference between the measured protein concentration in spiked platelet-poor plasma (PPP) and unspiked PPP was calculated and expressed as a ratio to the expected spiked concentration to represent the efficiency of spike recovery for the 5 ELISAs.*

	ELISA
	PDGF-BB	VEGF	IL-1β	TNF-α	TGF-β1
% recovery	121.2	100.3	144.4	65.3	89.9

*

IL-1β = interleukin–1 beta; PDGF-BB = platelet-derived growth factor–BB; TGF-β1 = transforming growth factor–beta 1; TNF-α = tumor necrosis factor–alpha; VEGF = vascular endothelial growth factor.

The accepted standards for performance for intra- and interassay CV are <20% or <25% at the ULOQ or LLOQ.^6,14 The TGF-β1, PDGF-BB, VEGF, and TNF-α ELISAs met the more rigorous of these 2 standards of performance with intra- and interassay CVs <20% (Table 2). Although the IL-1β ELISA had an intra-assay CV of 22.2%, which meets the more lenient standard for acceptable performance, the interassay CV of 27.2% did not meet even the most lenient standard of performance (Table 2).

Table 2.

Coefficients of variation for measurements of replicate samples of acid–citrate–dextrose, solution A (ACD-A) anticoagulated plasma.*

Run no.	PDGF-BB	VEGF	IL-1β	TNF-α	TGF-β1
1	3.1	12.1	22.2	7.2	4.0
2	4.7	10.2	11.8	5.7	7.7
3	6.3	11.2	6.4	3.4	3.1
Pooled	5.7	15.9	27.2	5.8	10.0

*

Three runs with spiked platelet-poor plasma (PPP) samples were performed with each ELISA. The coefficients of variation (CVs) for measurements of the spiked PPP concentration were calculated for each run to represent the intra-assay CVs. Values from all 3 runs were pooled and the CV calculated to represent the interassay CV. Note that the interassay CV was calculated based on pooling measurements from runs 2 and 3 only for the vascular endothelial growth factor (VEGF) and tumor necrosis factor–alpha (TNF-α) assays because the PPP was spiked to a different concentration in runs 2 and 3 than in run 1. IL-1β = interleukin–1 beta; PDGF-BB = platelet-derived growth factor–BB; TGF-β1 = transforming growth factor–beta 1.

The algorithm used to calculate the 4PL curve equations converged for all assays except for assay 1 with the TNF-α ELISA. For this assay, the algorithm would not converge to a solution even when trying multiple methods of convergence and multiple methods for supplying starting values. In order to accommodate and allow the use of the 4PL model, one data point that was visually most different from the remainder was removed, thus enabling model convergence.

Using these models of the 24 estimates of differences in the 4PL parameters, 1 was significantly different among the 4 standard curves for TGF-β1. Four estimates of differences in 4PL parameters were significant for VEGF, 7 were significantly different for PDGF-BB, 8 were different for IL-1β, and 10 were different for TNF-α. Composite graphs of the standard curves from all 4 runs were generated for each ELISA. Visual assessments of standard curve similarities were not consistently congruous with statistical assessment of 4PL parameter comparisons. Visually, the 4 standard curves were adequately similar with all ELISAs except for IL-1β, and the 4 standard curves appeared most similar for TNF-α (Fig. 1). Conversely, the greatest number of significant differences in standard curve parameters based on statistical analysis was with the TNF-α ELISA.

Figure 1.

Composite graph demonstrating high repeatability of the 4 standard curves obtained from each of 4 runs with the tumor necrosis factor–alpha (TNF-α) enzyme-linked immunosorbent assay (ELISA). OD = optical density.

Discussion

Previous standards for acceptable performance of ELISAs have been suggested and include an efficiency of spike recovery that should be 80–120% of the expected value, or potentially as low as 75–125% of the expected value, particularly when the values are at the lower and upper limits of detection.^6,14,21 Likewise, it has been suggested that intra- and interassay variability should be <20–25%.^6,14,21 No standards have been established with regard to acceptable similarity of standard curves generated in different assays to our knowledge.

For the TNF-α ELISA, the efficiency of spike recovery of 65.3% did not meet the established standard of performance. However, the intra- and interassay variability met industry standards with CVs of 3.4–7.2%. Because the intra-assay precision was high, we conclude that the relative amounts of canine TNF-α can be compared among different samples evaluated in the same run, although the measured concentration of TNF-α is likely to be lower than the actual concentration. As for consistency of the standard curves, the TNF-α ELISA had the greatest variability among the standard curves generated from the 4 runs based on the statistical comparisons of the 4PL curve parameters. However, this finding is incongruent with subjective assessment of the TNF-α standard curves that appear to be highly similar (Fig. 1). Accordingly, the validity of the statistical methodology used in this study to compare the standard curves is unclear, and we conclude that assessment of spike recovery efficiency and intra- and interassay variation are more meaningful assessments of ELISA utility.

For the TGF-β1 ELISA, the efficiency of spike recovery of 89.9% easily met the standard of acceptable performance. The CVs for intra- and interassay precision were all ≤10%. The established industry standard for acceptable repeatability is a variation of up to 20%. The data suggest that there are small differences in the standard curves generated with each of the 4 runs that manifest as a statistically significant difference in one of the 4PL parameters between 2 of the different standard curves. Given that type I error is cumulative, it would be expected that at least 1 of 24 pairwise comparisons would be different by chance alone. Hence, we conclude that the standard curves were not substantially different from one another. Given the high efficiency of spike recovery, the low intra- and interassay variability, and the consistency of the standard curves, we conclude that the TGF-β1 ELISA used in this study performs adequately to quantify TGF-β1 in ACD-A anticoagulated canine plasma.

For the PDGF-BB ELISA, the efficiency of spike recovery was within the accepted 25% of the theoretical spiked concentration. The intra- and interassay CVs were well within the accepted standard of 20%. ⁶ In the assay-to-assay comparisons of the 4PL curve parameters, 7 of the 24 parameters were significantly different. However, all of the significant differences were for comparisons of assay 1 to the other assays, and subjective assessment of the standard curves did not reveal substantial deviations from curve to curve. Given the uncertain validity of our methodology for comparing the standard curves, the acceptable efficiency of spike recovery, and acceptable intra- and interassay variability, this ELISA provides acceptable performance with canine ACD-A anticoagulated plasma.

Results for the VEGF ELISA were similar to that for PDGF-BB. The efficiency of spike recovery was also within 20% of the expected value. The intra- and interassay precisions were within the acceptable limit of 20%. ⁶ Four of the 24 pairwise standard curve parameter comparisons were significantly different. However, visual inspection of these curves did not indicate major differences between the standards. We conclude that this assay provides adequate assessment of canine VEGF in canine plasma samples anticoagulated with ACD-A.

For the IL-1β assay, the efficiency of spike recovery of 144.4% did not meet industry standard. The intra-assay CV was 22.2% and the interassay CV was 27.2%, which is outside the most lenient acceptable standard of 25%. ⁶ Eight of the 24 parameter comparisons were statistically significantly different in plate-to-plate comparisons, and subjectively these standard curves demonstrated substantial variability. We thus conclude that this assay does not meet acceptable standards and requires further optimization for use with canine ACD-A anticoagulated plasma samples.