Supplementary material for Attuil et al. (2000) Proc. Natl. Acad. Sci. USA 97 (15), 8473-8478.

Before PCR amplification, proteinase K (Appligene, Strasbourg, France; final concentration 250 mg/ml) was added to each 20-ml sample containing a sorted cell, and the tubes were incubated for 1 hr at 50°C and then for 5 min at 95°C. For the first round of PCR, a 30-ml mixture prepared in the manufacturer’s 1´ PCR buffer was added to the same tubes to obtain the following final concentrations: 100 nM each for oligonucleotides VB10a and JB1.5, 200 mM of each dNTP (Promega), 2.85 mM MgCl₂, and 0.9 unit of Taq polymerase (Boehringer Mannheim or Perkin-Elmer). The reaction begins with an initial cycle of 30 sec at 95°C, 4 min at 59°C, and 2 min at 72°C, followed by 44 cycles of 10 sec at 95°C, 45 sec at 59°C, and 1 min 30 sec at 72°C, and ends with 5 min at 72°C.

For the second round, 0.5 ml of the first PCR product was added to a 50-ml mixture prepared in the manufacturer’s 1´ PCR buffer containing as final concentrations: 100 nM each of oligonucleotides VB10b, JB1.2, and JB1.4, 200 mM of each dNTP, 1.75 mM MgCl₂, and 0.5 unit Taq polymerase. The reaction begins with 2 min (or 5 min) at 95°C and 5 sec at 72°C, then 35 cycles of 10 sec at 95°C, 1 min at 61°C, 30 sec at 72°C, and finally 5 min at 72°C.

PCR products were purified by using PCR purification columns (Boehringer Mannheim) according to the manufacturer’s instructions. Sequencing reactions were performed on the purified PCR products by using an Applied Biosystems Prism Dye Terminator Cycle Sequencing Ready Reaction Kit (Perkin-Elmer) and an internal primer specific for VB10 (5¢-AGGCGCTTCTCACCTCAGTCTTCA-3¢ from ref. 1). Samples were run on an Applied Biosystems 373 DNA sequencer (Perkin-Elmer) and analyzed with Applied Biosystems PRISM Ver. 3.0 software.

For AT experiments in which the CW3-selected TCR repertoires originating from single or multiple donors were to be compared, two categories of irradiated recipient mice were set up. In the first, each mouse within a pair was injected with spleen cells prepared from a single donor, such that each received the number of cells equivalent to a half spleen. In the second, each of three mice in a group was injected with cells pooled from the spleens of six donor mice, for the cell equivalent of two spleens. One day after the transfer, all recipient mice as well as normal DBA/2 control mice were immunized by injection of 500 P815-CW3 cells. A first flow cytometric analysis of PBL was performed on day 19 to monitor the CW3 response, and mice were killed on days 20, 21, or 26 to obtain lymphocytes for cell sorting. On day 19, there was already a substantial increase in the percent of VB10⁺ cells among CD62L- CD8⁺ cells for most (19/20) of the single-donor recipients (mean of 46.6%), for all of the multiple-donor recipients (58.1%), and for the four control immunized normal DBA/2 mice (59.3%) compared to the background levels (6.2%) in nonimmunized mice (Table 5). At this time, the proportion of CD8⁺ cells that expressed the CD62L- phenotype was somewhat lower in the AT groups (13.9% for single-donor and 13.1% for multiple-donor mice) than in the control immunized mice (42.5%). By the day of sorting, however, further expansion had occurred, and the proportion of cells that expressed the VB10⁺CD62L- phenotype characteristic of CW3-specific cells ranged from 22.9% to 60% of CD8 cells for all of the immunized mice selected for single-cell repertoire analysis (Table 6). Cells gated for the VB10⁺CD62L-CD8⁺ phenotype characteristic of CW3-specific T cells were sorted as single cells directly into PCR tubes for TCR repertoire analysis.

In initial experiments, the rearranged VB10 TCR junctional region DNA was amplified directly from individual cells by a previously described (1) seminested PCR protocol that uses primers specific for the VB10 and JB1.2 gene elements. This method amplifies VB10 TCRs that have rearranged to either JB1.1 or JB1.2. In the first pair of single-donor recipient mice analyzed, three and four TCRs with the canonical SxGxxx CDR3 motif were identified for mice 4A and 4B, respectively (Table 7). One VB10-JB1.2 TCR nucleotide sequence (1.2-42c) was shared between them. Because relatively few different TCRs were identified from this first pair of AT mice, we altered our original PCR protocol to amplify a broader VB10 TCR repertoire. In our modified nested PCR protocol (described in detail in Materials and Methods), the JB1.2 primer was replaced by one specific for JB1.5 in the first PCR and by a mixture of JB1.2 and JB1.4 primers in the second PCR. In this way, we were able to amplify efficiently VB10 TCRs rearranged to the JB 1.1, 1.2, 1.3, or 1.4 segments. The latter accounted for most (77%) of the TCRs expressed on a series of CW3-specific CTL (3).

(i) Generation of the Ag-specific preimmune repertoires of each donor mouse.

(ii) Shuffling, partitioning, and transfer of the combined donor repertoire to the recipient mice.

(iii) Generation of the immune-stimulated recipient repertoire by selection and expansion of Ag-stimulated clones.

(iv) Random sampling of a small recipient-specific number of cells from the Ag-stimulated repertoires.

A donor repertoire typically consists of a relatively small number (<1,000) of cells belonging to different clones. The process of generating such a repertoire is controlled by four parameters: the number of donor mice, the number of Ag-reactive cells per mouse, the repertoire type, and a fourth parameter whose function depends on the repertoire type. The two available repertoire types are called "equal clone size" and "unequal clone size." In the case of an equal clone size repertoire, the fourth parameter specifies the number of clones. An unequal clone size repertoire consists of a series of clones of exponentially decreasing size. In this case, the fourth parameter specifies a decay rate. For both repertoire types, the actual number of cells per clone corresponds to the nearest integer approximation of the exact (real number) clone size distribution leaving the total number of Ag-reactive cells per mouse unchanged. For an unequal size repertoire, the number of clones thus depends on the total number of cells, as for larger repertoires more clones will reach a size roundable to a nonzero integer. Note further that in the computer simulation, different clones always contain different TCR types, and that clones from different donors never contain the same TCR types. The terms clones and TCR types are thus synonymous in this context.

The combined donor repertoire of several mice is randomly split into subsets of equal size. Each subset corresponds to the input received by a recipient mouse. As in the previous step, the numerical procedure applied keeps the total number of cells unchanged, but as a consequence may create recipient subsets that differ in size by one. The immune-stimulated repertoire is assumed to consist of a sufficiently large number of cells such that size-dependent sampling effects will not occur. The relative abundances of different TCRs are therefore expressed as fractions. The clonal selection and expansion process can be modeled according to two different modes, called uniform and nonuniform. Both processes start with the random selection of a user-defined number of cells from the input repertoire. Note that it is possible to select multiple cells from the same clone at this stage. The number of different TCR types in the expanded repertoire may thus be smaller than the specified number of clonal selection events. In the uniform expansion mode, the Ag-stimulated cells are expanded to cell populations of equal size. In the nonuniform mode, an individual Ag-stimulated cell is expanded to a burst size of Ne^rln(B), where N is the minimal burst size, B the maximal burst size ratio, and r a random number between zero and one. Note that the minimal burst size is not considered a parameter of the model because it does not affect the relative sizes of the expanded cell colonies (which are represented as a set of fractions summing to one). In the final step, a random cell sample is extracted from the immune-stimulated Ag repertoire of each recipient mouse and characterized by TCR type distribution. Note that the number of extracted cells is a user-defined parameter, which may be different for each recipient mouse.

The same type of selection experiment can be simulated many times at once, in which case summary statistics of the results may be included in the output. These summary statistics consist of the means and the standard deviations of three observables: (i) the number of different TCR types in one recipient mouse, (ii) the standard deviation of the relative frequencies of the different TCR types in one recipient mouse, and (iii) the number of TCR types shared by at least two recipient mice having received the same donor repertoire.

The simplest models in which equal-size clones in preimmune repertoire are selected in a uniform manner were eliminated primarily because they predict a much more limited variation in the sizes of the selected clones than we found experimentally. When more variation is introduced at the level of the preimmune repertoire (unequal sized clones) but the selection is kept uniform, another problem emerges, namely the prediction of too many shared clones.

Our results clearly support models in which variability occurs at the level of TCR selection and clonal expansion during the CW3 response. In our model, the parameter termed "maximal burst size ratio" operationally defines the degree of nonuniform expansion and is intended to account for differences in cell numbers attained after individual stimulation events (e.g., a value of 1,000 means that burst sizes may vary by up to a factor of 1,000). No assumption is made about the underlying reason(s) for burst size differences, but these may include variable rates of clonal expansion (e.g., because of TCR affinity) and differences in the timing of the stimulation of individual cells. Models of preimmune repertoires composed of equal-size clones that are selected in a nonuniform manner fit the experimental data for some AT groups but not for all of them. For example, an 800-cell preimmune repertoire of 20 or 30 clones fits the data for pair 11A/B with 6 or 8 selection events over a range of 32-256 or 256-1,024 burst size ratios, respectively. Moreover, the clones selected under these simulation conditions varied widely in size, similar to the AT experiments. As an upper limit, increasing the number of selection events to 10 under maximal burst size ratio conditions that would give 4-5 detectable clones reduced the probability of finding shared clones. Similar conditions appeared to fit the AT data for pair 14A/B. However, similar simulation conditions tested over a wide range of selection events (two to eight events) and burst size ratios failed for pairs 12A/B and 13A/B, in that too many clones would have been detected under conditions that predicted shared clones.

Models that best fit the data are those in which the preimmune repertoire would be composed of unequal-size clones, and the selection would be operationally nonuniform. For example, a preimmune repertoire of 800 cells with a size distribution based on an exponential decay of 0.8 would consist of 28 different clones, and a nonuniform mode of 4-5 selection events with burst size ratios of 32-4,096 would allow a high probability (>90%) for the selection of shared clones corresponding to experimental data of the AT pairs 11A/B and 14A/B. Certain combinations were eliminated in that they result in repertoire sizes of less than 20 clones (e.g., exponential decay of 0.7 and 200 cells gives only 15 clones in the preimmune repertoire). As an upper limit, increasing the decay rate to 0.9 was found to be less favorable for selection of shared clones, because the preimmune repertoire would be composed of a larger number of clones with more similar clone sizes. The reverse trend for the exponential decay rate would optimize models for the multiple-donor AT group because repertoires with the highest decay rates would have reduced probabilities of selecting shared clones. Under the above conditions optimized for pairs 11A/B and 14A/B, the probability of selecting shared clones would be somewhat lower (63-81%) for pairs 12A/B and 13A/B, for which we detected only 1-2 distinct TCRs. For the multiple-donor group, the probability of selecting shared clones is even lower (62-73%), as would be required to fit the data. Although these conditions do not predict our AT data exactly, they seem reasonably close compared to other conditions, in that the probability is maximized for the AT pairs for which we found two shared clones and is minimized for the others.

1. Maryanski, J. L., Jongeneel, C. V., Bucher, P., Casanova, J. L. & Walker, P. R. (1996) Immunity 4, 47-55.

2. Cochet, M., Pannetier, C., Regnault, A., Darche, S., Leclerc, C. & Kourilsky, P. (1992) Eur. J. Immunol. 22, 2639-2647.

3. Casanova, J. L., Martinon, F., Gournier, H., Barra, C., Pannetier, C., Regnault, A., Kourilsky, P., Cerottini, J. C. & Maryanski, J. L. (1993) J. Exp. Med. 177, 811-820.

4. Casanova, J. L., Cerottini, J. C., Matthes, M., Necker, A., Gournier, H., Barra, C., Widmann, C., MacDonald, H. R., Lemonnier, F., Malissen, B., et al. (1992) J. Exp. Med. 176, 439-447.

5. Chothia, C., Boswell, D. R. & Lesk, A. M. (1988) EMBO J. 7, 3745-3755.