In fabricating protein microarrays, rapid evaporation can occur when printing nanoliter quantities of protein solutions. Rapid evaporation, in turn, can produce abrupt changes in the solution conditions, exposing the protein solute to entirely different environments and compromising the functional integrity of the microarray [xv
]. An example is given in , which shows the effect of adding small amounts of dimethyl sulfoxide (DMSO) to the printing solution on the pHLA A2-Ig complexes printed on the microarray. DMSO is typically used to enhance the solubility of hydrophobic peptides in preparing peptide-loaded HLA complexes [xxvii
]. DMSO is also known to induce protein unfolding in aqueous solution at high concentrations on the order of 40% (v/v) or more [xxix
]. We found that adding only 0.33% (v/v) DMSO to a printing solution containing the pHLA A2-Ig complex () reduces PE-labeled W6/32 antibody bound to the printed complex by more than 90%, independent of peptide loading with a specific peptide of interest – in this case, CMVpp65 – or with endogenous peptides resident in the “unloaded dimer” [vi
]. A similar effect is observed for FITC-labeled anti-Igλ antibody binding to the CMVpp65-loaded dimer (). The mean spot fluorescence intensity in this case is reduced by nearly half when the dimer complex is printed from solutions containing 1.4% (v/v) DMSO compared to solutions without DMSO.
Figure 2 Effect of dimethyl sulfoxide (DMSO) on the binding of anti-HLA (clone W6/32) and anti-Ig antibodies binding to microarray spots printed with pHLA A2-Ig dimers. From left to right: PE-labeled W6/32 (20 μg/ml) contacted with spots printed with 0.1mg/ml (more ...)
Since monoclonal W6/32 antibody recognizes the folded HLA domain of the complex [xxx
], while the anti-Igλ antibody recognizes the folded Ig domain of this complex, these results suggest that the presence of DMSO at dilute concentrations in the printing solution leads to the destabilization of the complex as a whole, rather than acting locally, for example, to render the antigen-recognition interface inaccessible. This deleterious effect is not observed when printing much larger (microliter) volumes of solution containing DMSO at dilute concentrations; thus implicating the printing/rapid evaporation of nanoliter solution volumes that leads to the loss of functionality.
We also considered the loss of functionality of the pHLA A2-Ig complex due to reaction with the amine-reactive NHS groups on the substrate by first pre-coating the entire slide with anti-Igλ antibody, then blocking with excess ethanolamine to quench unreacted NHS groups before printing the pHLA A2-Ig complex. We still observed very low levels of W6/32 antibody binding to the printed complex, and thus concluded that this reaction does not contribute significantly to the lability of the pHLA A2-Ig complex.
The strategy we adopt here circumvents printing the labile HLA molecule altogether, but takes advantage of the unique structure of this complex by printing instead the more robust and stable anti-Igλ antibody, which binds to the Ig domain of the complex, and enables printing a single solution on each spot across the entire microarray. Moreover, the TCR-pHLA molecular interactions that define T-cell antigen specificities take place in solution, rather than at the cell-microarray substrate interface, allowing for further control and optimization of the assay. Both factors contribute to the reliable quantitative determination of antigen-specific frequencies.
Quantifying antigen-specific cell capture on the pHLA A2-Ig microarray
We first examined the ability of this microarray assay to capture T cells specific for three peptide antigens: CMVpp65 from cytomegalovirus, and M1.58 and PA.46 from influenza A virus [xxxi
]. depicts 5×5 subarrays of 25 anti-Igλ antibody spots with cells captured from these enriched populations when the cell suspension in each case was incubated with the cognate peptide-loaded HLA A2-Ig complex. Also shown is a 5×5 subarray of 25 anti-Igλ antibody spots contacted with the CMVpp65-enriched cell population incubated with the unloaded dimer (negative control). Essentially no cell binding to the anti-Igλ antibody spots is observed in this case. Similar results were obtained for the M1.58-enriched and PA.46-enriched cell populations when incubated with the unloaded dimer (images not shown).
Figure 3 Comparison of pHLA A2-Ig dimer microarray and FACS assays of the frequency of antigen-specific CTLs in enriched T-cell populations. (a) Representative images of cell capture on 5 × 5 subarrays of 25 anti-Igλ spots and anti-CD3 spots. From (more ...)
A 5×5 subarray of anti-CD3 antibody spots used for an internal calibration of the assay is also shown in . Normalizing the mean intensity of the anti-Igλ antibody spots against the mean intensity of anti-CD3 antibody spots takes into account experimental variations in cell labeling across individual cell populations or from one experiment to another, as well as differences in capture efficiencies from slide to slide. Importantly, this normalization allows a quantitative assessment of the frequency of antigen-specific CTL in the total T-cell population. As illustrated in , the frequencies derived from the microarray assays using the enriched CTL populations are essentially equivalent to those obtained by flow-activated cell sorting (FACS; see Figure S2
for plots). The enriched CMVpp65-specific CTL frequency measured by FACS is 71.5% compared to 67.9±4.2% measured in the microarray assay. Similar agreement is obtained for the enriched PA.46-specific CTL frequency: 69.5% by FACS compared to 67.0±4.0% from the microarray assay; and for the enriched M1.58-specific CTL frequency: 13.2% by FACS compared to 14.6±1.3% from the microarray assay, which shows the quantitative accuracy of the microarray assay for these enriched cell populations.
Quantitative detection of low-frequency CTL
The ability to detect and quantify antigen-specific CTL at low frequencies on the order of 0.01% or less is critical to characterizing broad-based immune responses [xxxii
]. To evaluate the sensitivity of the microarray assay to low-frequency CTL, we measured CMVpp65-specific T-cell capture from suspensions of CMVpp65-specific CTL in CD8 T cell-depleted autologous PBMCs. Assays were carried out for five different concentrations of CMVpp65-specific T cells: 0%, 0.01%, 0.1%, 1% and 10% of the total cell number. The cell suspension containing only CD8 T cell-depleted autologous PBMCs quantifies the background fluorescence intensity from endogenous fluorophores in the anti-Igλ antibodies and BSA printed on the microarray. Parallel assays in which each cell suspension is incubated with the unloaded dimer measure non-specific cell capture.
As shown in , the mean spot intensities for cell suspensions incubated with the unloaded dimer are not significantly different from the background fluorescence intensity for the lower CMVpp65-specific CTL frequencies. In contrast, some non-specific cell capture is evident at CTL frequencies of 1% and 10% where the mean spot intensities are significantly greater than the background intensity. The quantitative impact of non-specific cell capture on the antigen-specific CTL frequency is, however, negligible. For example, the extent to which the mean spot intensity exceeds the background intensity for cells captured from the 1% CMVpp65-specific cell suspension incubated with the unloaded dimer is comparable to that for cells captured from the 0.01% cell suspension incubated with the CMVpp65-loaded dimer, and likewise for the 10% cell suspension incubated with the unloaded dimer relative to the 0.1% cell suspension incubated with the CMVpp65-loaded. Thus, the correction for non-specific cell capture is approximately two orders of magnitude less than the actual value of the antigen-specific CTL frequency at these higher frequencies.
Figure 4 Quantified mean spot fluorescence intensities of cells captured from suspensions containing different concentrations of CMVpp65-specific CTLs incubated with either the CMVpp65-loaded dimer or the unloaded dimer. A CMVpp65-enriched cell population containing (more ...)
The mean spot intensities in corresponding to average numbers of CMVpp65-specific T cells captured on each spot decrease monotonically as the frequency of those cells decreases from 10% to 0.01%, with the mean spot intensity for the lowest frequency still greater than the background fluorescence intensity. Nonetheless, the mean spot intensities for CMVpp65-specific T-cell capture from the 0.1% and 0.01% cell suspension are not statistically different than those corresponding to non-specific cell capture from these suspensions (p-values are given in caption), indicating a lower detection limit that is greater than 0.1%, or 100 CMVpp65-specific CTL in a total population of 100,000 CD8 T cells, the cell seeding condition used in these assays. Moreover, at the lowest antigen-specific CTL frequency of 0.01%, most of the microarray spots will not display captured antigen-specific T cells. This situation is seen in where the lower limit of the error bar for the mean spot intensity corresponding to CMVpp65-specific cells captured from the 0.01% cell suspension extends below the background fluorescence intensity. Clearly assays for low-frequency CTL can produce sparsely populated microarrays, especially when the total cell population is small, and this must be taken into account in order to obtain an accurate quantification of antigen-specific CTL frequencies.
To this end, we account for the uneven distribution of cells captured across the spot replicates by considering the variation in spot fluorescence intensities in the microarray assay. shows that variability for M1.58-specific T cells captured from suspensions of CFSE-labeled, M1.58-specific CTL in CD8 T cell-depleted autologous PBMCs. The absence of cells captured on a spot is defined to be the highest spot intensity observed over the entire distribution of spot intensities at 0% M1.58-specific CTL in CD8 T cell-depleted autologous PBMCs incubated with the M1.58-loaded dimer. This cutoff intensity is nearly five standard deviations above the mean spot intensity, and establishes a stringent criterion for cell capture at the higher (non-zero) frequencies. In addition, this cutoff intensity is more than two standard deviations above the mean spot intensity for non-specific cell capture from the undiluted, M1.58-enriched cell suspension incubated with the unloaded dimer – the negative control for M1.58-specific T cell capture in – and thus, also indirectly accounts for non-specific cell capture from this cell suspension at the low CTL frequencies considered. It is important to note that non-specific CD8+ T cells are present at a frequency nine times higher than the frequency of M1.58-specific CD8+ cells since the M1.58-enriched cell suspension contains only 10% M1.58-specific T cells.
Figure 5 Individual spot intensities corresponding to CFSE-labeled M1.58-specific T-cell capture on a 10 × 10 subarray of 100 anti-Igλ antibody spots as a function of M1.58-specific CTL frequency. (a) Spot intensities corresponding to cells captured (more ...)
At 10% M1.58-specific CTL, we find that all 100 spot replicates have fluorescence intensities above this cutoff intensity indicating that captured cells are detected on each spot, although the variability in spot intensities is substantial. Considering a total population of 100,000 cells in the suspension, as many as 10,000 M1.58-specific CTL are captured, or 100 cells captured per spot on average. As the frequency of M1.58-specific CTL in the suspension decreases, captured cells are detected on fewer spots. At the lowest frequency of 0.01% M1.58-specific CTL, only two spots have fluorescence intensities higher than the cutoff, indicating that at least two cells have been captured out of a total 10 M1.58-specific CTL in the suspension. This lower bound sets the minimum efficiency for antigen-specific cell capture at 20%, which is much greater than the efficiency based on the fractional area of all the spots of 3.8%. The result implies that the cells sample the microarray surface more efficiently than expected from the quiescent settling of cells from the suspension. If the 10 M1.58-specific cells were randomly distributed over the 100 spots, the probability of capturing all of them on just two spots is approximately (1/50)8 or ~ 10−14. The random probability of capturing three M1.58-specific cells on just two of the 100 spots is about 2%, and it is ~ 10−4 for four cells on two of 100 spots. We conclude, therefore, that only two or three M1.58-specific cells were likely captured in this assay; thus establishing that the sensitivity of 0.01% antigen-specific CTL frequency corresponds to detecting the capture of two or three antigen-specific T cells.
An important implication of the large variability in spot intensities in is that mean spot intensities will not provide an accurate quantitative measure of low-frequency antigen-specific CTL, even after taking into account the unoccupied spots on sparsely populated microarrays. An alternative measure is the sum of spot intensities for only those spots displaying captured cells. Using this measure, a linear dependence over four orders of magnitude (R2=0.998) is obtained by plotting the logarithm of this sum as a function of the logarithm of the frequency of M1.58-specific CTL (). Our finding that the linear dependence holds even when the target cells are sparse in number compared to the number of microarray spots is especially significant for clinical and research applications where small volumes and consequently small numbers of target cells are dictated by limitations on sample size.
Profiling influenza-specific CTL in peripheral blood
We also evaluated the microarray assay for simultaneously detecting and quantifying antigen-specific CD8+ CTL frequencies in more complex cell populations from peripheral blood by comparing the assay with flow cytometry for three influenza A epitopes: the immunodominant M1.58 peptide, and the subdominant NA.75 and PA.46 peptides [xxxiii
]. For the microarray assays, the spots displaying captured cells were identified by accounting for the variability in spot fluorescence intensities, as described above, and normalizing by the sum of spot intensities for all anti-CD3 antibody spots. Non-specific cell capture on the microarray was taken into account as described above and found to be less than the number of cells detected by FACS that bind the unloaded dimer.
The results in show that multiple antigen-specific CTL can be detected simultaneously and reliably for a variety of antigen-specific responses, including responses to subdominant as well as immunodominant epitopes. The M1.58-specific and PA.46-specific CTL frequencies were found to be 3.1% and 2.1%, respectively, in microarray assay compared to 1.5% and 0.67%, respectively, determined by FACS dimer staining. A significantly lower frequency of NA.75-specific CTL was obtained in both assays: 0.19% in the microarray assay compared to 0.07% determined by FACS dimer staining. The consistently higher frequencies derived from the microarray assays presumably reflect the high efficiency of cell capture on the microarray coupled with a higher level of non-specific binding of the unloaded dimer in the FACS analysis, which together predict comparatively higher antigen-specific CTL frequencies in the cell suspensions in contact with the microarray.
Figure 6 Detection of M1.58-, NA.75-, and PA.46-specific CTL in CD8+ CTL isolated from peripheral blood of A2-positive human donor A8 by pHLA A2-Ig cellular microarrays and FACS. For each assay, the level of non-specific binding to the unloaded dimer was subtracted (more ...)