We thank Drs. Ortiz and Kang for performing this principal component analysis (PCA). We are pleased that the data of our 12 well-defined CTL clones of identical specificity enabled this analysis to be made. To date it has been difficult to make decisions about optimal screening parameters for CTL clones because previous studies included few clones and compared few parameters.

It is positive that K-means clustering led to assignments of individual CTL to the same clusters that we defined. Ortiz and Kang included seven of the variables we analyzed in functional studies, but they did not consider our analysis of structural characteristics, including determination of mean fluorescence intensity as well as on- and off-rates of MHC–multimer binding. We agree with this decision because we found no correlation between structural and functional parameters and multimer binding did not distinguish between low and high avidity CTL in our analyses. This is particularly relevant because functional characteristics were retained when TCR of representative clones were expressed as transgenic receptors in recipient lymphocytes.

Ortiz and Kang identified avidity (positive association) and IL-13 (negative association) as the variables best predictive of CTL cytotoxic activity against tumor cells. We agree with this conclusion, but still recommend a different strategy for early selection of clones based on considerations that are not apparent in PCA. Two reasons underpin our recommendation that Th1-polycytokine secretion patterns should be assessed in early screening of CD8+ clones.

First, when many T cell clones must be analyzed it is important to select the good and eliminate the bad clones as rapidly as possible. Discrimination must be made early on to avoid extensive CTL culture, saving time and money. While high avidity (i.e., high peptide sensitivity) is an excellent parameter to predict good tumor recognition, its measurement requires almost 2 × 105 CTL, which must be expanded for 6–7 wk from a single cell. Alternatively, Th1/Th2 cytokines can be measured with 1 × 104 CTL obtained in 3–4 wk.

Second, as we discussed, several studies of virus-specific CTL showed that Th1-polycytokine secretion of IFN-γ, IL-2, and TNF-α was positively associated with improved virus control and correlated with greater peptide sensitivity. Likewise, our antitumor clones producing this cytokine triplet show higher avidity. Good tumor killing can be confirmed in the selected CTL. We also suggested that selection of higher avidity CTL might be improved by measuring Th2 cytokines. The analysis of Ortiz and Kang supports this contention and confirms that production of IL-5 and IL-13 is negatively associated with good cytotoxicity, whereby IL-13 provides better discrimination. Use of Th1/Th2 cytokine patterns for early clone selection is attractive because many factors can be assessed with multiplex arrays without need for more cells. The limiting factor is cost.

Based on these arguments, it would be of interest to us to see an additional PCA in which cytotoxicity is replaced with avidity to determine the best parameters to rapidly identify high avidity CTL.