Normal Distribution of CD8+ T-Cell-Derived ELISPOT Counts within Replicates Justifies the Reliance on Parametric Statistics for Identifying Positive Responses

Abstract

Accurate assessment of positive ELISPOT responses for low frequencies of antigen-specific T-cells is controversial. In particular, it is still unknown whether ELISPOT counts within replicate wells follow a theoretical distribution function, and thus whether high power parametric statistics can be used to discriminate between positive and negative wells. We studied experimental distributions of spot counts for up to 120 replicate wells of IFN-γ production by CD8+ T-cell responding to EBV LMP2A (426 – 434) peptide in human PBMC. The cells were tested in serial dilutions covering a wide range of average spot counts per condition, from just a few to hundreds of spots per well. Statistical analysis of the data using diagnostic Q-Q plots and the Shapiro-Wilk normality test showed that in the entire dynamic range of ELISPOT spot counts within replicate wells followed a normal distribution. This result implies that the Student t-Test and ANOVA are suited to identify positive responses. We also show experimentally that borderline responses can be reliably detected by involving more replicate wells, plating higher numbers of PBMC, addition of IL-7, or a combination of these. Furthermore, we have experimentally verified that the number of replicates needed for detection of weak responses can be calculated using parametric statistics.

ELISPOT Assays in 384-Well Format: Up to 30 Data Points with One Million Cells

“…Q-Q plots ( Figure 4 ) for experimental spot counts in replicate wells were linear confirming the results of Shapiro Wilk test. This proves that the spot counts in replicate wells of 384-well plates, as previously reported for 96-well plates [ 14 ], follow Normal distribution function. Therefore, in 384-well plate as well as 96-well plate assays, parametric statistics (including the Students’t-test or ANOVA) can be used to compare test results.…”

“…, one-third of the absolute numbers per well. Because of the inherent variability between replicate wells, which for 96-well plates was shown to follow a Normal distribution [ 14 ], a direct well-to-well comparison between 96-well and 384-well format is best done by establishing whether variability among replicates follows the same rule in 384-well format. Furthermore, the Normal distribution of spot counts in replicate wells allows one to determine whether parametric statistics can be used to compare antigen-induced spot counts with spot counts in the medium control wells, identifying cut-offs for positive responses.…”

High Reproducibility of ELISPOT Counts from Nine Different Laboratories

“…The Normal Distribution of ELISPOT counts among replicate wells on the RP confirms what we previously reported using IFN-γ transfected CHO cells [ 24 ], and was also observed involving higher numbers of replicate wells, testing primary PBMC for peptide-induced recall responses of CD8 cells [ 25 ]. Therefore, parametric statistical analysis, including the use of the Students’ t-test is suitable for comparing different experimental conditions of ELISPOT tests, such as identifying positive antigen-induced spot counts over the medium control.…”

Exposure to wild-type AAV drives distinct capsid immunity profiles in humans

“…Normal distribution of the samples was tested using the D'Agostino-Pearson normality test. All data shown were analyzed by nonparametric tests, with the exception of the FluoroSpot data in Figure 4D (80) and the data in Figure 6E, which were analyzed with a Student's t test. A Mann-Whitney U test was used for the comparison of data sets composed of 2 groups, except in Figure 3B, in which a Wilcoxon signed-rank test was used because of the absence of variance in the control group.…”