Arstila *et al*. (1) estimated an average diversity of 9 × 10^{5} different β chains and 4.5 × 10^{5} different α chains in the human naı̈ve T cell repertoire. To calculate the total T cell repertoire diversity, the β-chain diversity was estimated within a certain variable (V) gene family, V_{α}12^{+}, comprising 2.5% of the total α-chain repertoire. Finding in this particular family an estimated total of 6 × 10^{5} different β chains (i.e., two-thirds of the total β-chain repertoire), Arstila *et al*. suggested that the total T cell receptor (TCR) diversity comprises at least (6 × 10^{5}) × 40 = 2.4 × 10^{7} different αβ combinations (1). The authors acknowledge that this is only a lower bound, because the calculation assumes that the β chains that do bind at least one V_{α}12 chain bind only one of the 4.5 × 10^{5} different α chains in the V_{α}12^{+} family. If each β chain found within the V_{α}12^{+} family were to bind an average of *n* different V_{α}12 chains instead, the total estimated TCR diversity would be *n*-fold higher than this lower bound.

Arstila *et al*. estimated an upper bound of 10^{8}different αβ combinations (1). Pre–T cells having rearranged a β chain expand 1000-fold before the α chain is rearranged, and only 10% of these cells leave the thymus to enter the mature repertoire. Thus, it was argued that each β chain can maximally pair with any of about 100 different α chains.

This is indeed correct for all descendants of any particular pre–T cell having rearranged a particular β chain—but another pre–T cell rearranging the same β chain may bind to 100 different α chains. Thus, to calculate the upper bound on TCR diversity, one has to consider the frequency with which identical β-chain rearrangements are expected. This frequency can be estimated from the turnover rate of the naı̈ve T cell repertoire. In human adults, the total body production of naı̈ve T cells has been estimated at about 10^{8} per day (2), a figure obtained from recovery rates following T cell depletion (2) and from an estimated 0.1% turnover (3) in a pool of 10^{11}naı̈ve T lymphocytes. Assuming that most of this production is of thymic origin (4) and that more than 90% of the cells die before leaving the thymus (1), this implies a daily production of at least 10^{9} pre–T cells. The 1000-fold expansion of the pre–T cells (1) before α-chain rearrangement implies that approximately 10^{6} β chains should be made every day. Because this is close to the Arstila *et al*. estimate of total β-chain diversity, every β chain should be rearranged about every day.

Over the 1000-day expected life-span (2, 3) of the progeny of a pre–T cell expressing a single β chain, therefore, 1000 recurrences of the same β-chain rearrangement might be expected. Hence, the upper bound for the total TCR diversity could easily be 1000-fold larger than calculated by Arstila *et al*. Such an upper bound, at 10^{11}, would allow almost every T cell in the naı̈ve repertoire to have a unique TCR. The true TCR diversity may be several fold lower, however, owing to factors such as proliferation after the α-chain rearrangement and possible restrictions in αβ-chain pairing.

## REFERENCES

# Diversity of Human αβ T Cell Receptors

*Response*: Keşmir *et al*. argue that although any developing TCR β chain will be paired at most with 100 different α chains, the same β chain may appear repeatedly and garner other sets of 100 α chains, increasing the total αβ TCR diversity from the 10^{8} we estimated (1). We studied the diversity of the human αβ TCR in the blood of healthy adult donors at a given moment, not over time. Also, we did not measure the upper limit of α-to-β pairing; our estimate was based on what is known of αβ T cell development and TCR rearrangement. Thus, the comment of Keşmir *et al*. actually goes beyond our data.

Because any expansion after α-chain rearrangement will increase only clone size, not diversity, the argument of Keşmir *et al*. hinges on the assumption that the estimated total turnover of naı̈ve T cells equals thymic production of pre–T cells. That assumption is incorrect, however, and ignores the well-documented role of post-developmental division in the maintenance of the naı̈ve T cell population, especially in adults. Murine T cells may go through up to six cell cycles after α-chain rearrangement even before emigrating from the thymus (2). Haynes *et al*., cited by Keşmir *et al*., specifically argued for “minimal contributions of the thymus to maintenance or reconstitution of the peripheral pool of T cells . . .” in humans [(3), p. 457], and showed that the presence or absence of thymic function and even the surgical removal of the thymus had no impact on the reconstitution of the T cell compartment, including the naı̈ve CD4^{+} cells, in treated HIV-infected individuals. Naı̈ve T cells, long after having completed TCR rearrangement, clearly have a considerable capacity for self-renewal.

The suggestion of Keşmir *et al*. can also be viewed as a question of clone size. If the size of the repertoire is 10^{8} different TCRs, as we suggest, the average clone among 10^{11} naı̈ve T cells would consist of 1000 cells, the progeny of a single intrathymic α-chain rearrangement after 10 cell cycles. These cycles should therefore be detectable in the naı̈ve T cell population, and indeed this appears to be the case. Studying the disappearance of cells damaged by therapeutic irradiation, McLean and Michie (4) concluded that, on average, naı̈ve T cells divide once every 3.5 years and die after 20 years, which suggests six post-thymic cell cycles in the life-span of an average naı̈ve T cell. Other experimental approaches have suggested higher division rates. From age 25 to 70 years the mean telomere length in the naı̈ve T cell population decreases from 9.5 kb to 8.0 kb, so an estimated loss of 50 to 100 base pairs (bp) per cell cycle translates to 7 to 13 divisions during the 20-year life-span of naı̈ve cells (5). De Boer and Noest have argued that this estimate of telomere loss is too high; their estimate, 35 to 70 bp per cycle (6), would mean 10 to 19 cycles. At any given time the fraction of naı̈ve T cells in cell cycle is 0.8% (7), which suggests a rate as high as one division per 125 days, or 60 cycles per life-span. The available data thus can easily accommodate 10 divisions producing the average naı̈ve clone.

Studies on the frequency of antigen-specific T cell precursors provide an independent line of evidence that points to a diversity close to what we proposed. A conservative estimate of the frequency of such precursors in the naı̈ve repertoire would be one per million; some studies have reported significantly higher frequencies (8, 9). Thus, a total repertoire of 10^{8} TCRs would predict an epitope-specific response to consist of 100 clones, while Keşmir *et al*.'s repertoire of 10^{11} TCRs predicts a composition of 100,000 responding clones. The existing literature is more compatible with our prediction (10–14). Thus, we submit that the phenomenon that Keşmir *et al*. postulate, although in principle possible, has little impact on the total diversity.