A223 tumor line was described previously (21). Tumor cells were injected into wild-type (WT) C57BL/6 (B6) (Stock no. 000664) (Jackson Laboratories). Both male and female mice (6-8 weeks) were used for the study. All mice were maintained under specific pathogen-free conditions in the vivarium facility of University of Colorado Anschutz Medical Campus (Aurora, CO) or in the UPMC Hillman Cancer Center Animal Facility (Pittsburgh, PA). Animal work was approved by the Institutional Animal Care and Use Committee of University of Colorado Anschutz Medical Campus (AMC) (Aurora, CO) and University of Pittsburgh (Pittsburgh, PA).

A223 cells were cultured and trypsinized as described earlier (19). A223 cells (1×10⁵) were resuspended to a final volume of 100μL in 50% Matrigel (Corning)/50% PBS and injected subcutaneously into one flank of each mouse. Tumor length and width were measured with calipers, and tumor volume was calculated as (length×width²)/2. When tumor size reached ~250-350mm³, tumor-bearing mice were treated with anti-PD-L1 (200µg/mouse/time diluted in PBS, clone 10F.9G2, BioXCell, Catalog# BE0101) via intraperitoneal (i.p.) injection three times (2-day interval) or PBS only as vehicle control (control group). Of note, as shown in our previous study (19), all the tumor-bearing mice in the control group had tumor growing out, and no substantial differences were observed in tumor growth among control group. To assess treatment effects, relative change in tumor volume (RCTV) was calculated as the change in tumor volume (TV) from the start of treatment (TV₀) to the TV at day n (the endpoint of control group) (TV_n) divided by TV₀ (RCTV=[TV_n−TV₀]/TV₀). For example, if tumor-bearing mice were treated on day 12 (the start of treatment), day 14, and day 16 with anti-PD-L1, and tumors were collected and analyzed on day 20, the RCTV would be calculated as follows: RCTV= (TV_day20−TV_day12)/TV_day12. Based on the RCTV, anti-PD-L1 treated recipients were divided into responders (RCTV<0), slow progressors (0<RCTV ≤ 1.5) and non-responders (RCTV>1.5).

Tumors were harvested from 3 R (1R-3R) and 4 NR (4NR-7NR) tumor-bearing mice and single-cell suspensions were prepared as described previously (19). Single-cell suspensions were subjected to EasySep™ Mouse CD8a Positive Selection Kit II (StemCell Technologies, Catalog#18953) according to manufacturer’s instructions to purify CD8 T cells followed by genomic DNA isolation. DNA concentration and purity were determined by NanoDrop™ One^C (Thermo Fisher Scientific). Genomic DNA samples were submitted to Adaptive Biotechnologies for amplification and sequencing of TCRβ chain CDR3 regions using the ImmunoSEQ platform (Adaptive Biotechnologies) at survey resolution (22). The same amount of genomic DNA was used for all 7 samples. The TCRβ chain CDR3 sequences and corresponding V, D, and J segments were delineated using the algorithm developed by ImMunoGeneTics (IMGT) collaboration (23). Data were analyzed using ImmunoSEQ Analyzer platform (http://www.adaptivebiotech.com/immunoseq/analyzer) and also extracted for further analysis using R (version 4.1.0) or using immunarch (version 0.6.7) package of R. Only in-frame TCR rearrangements were used for clonotype analysis. Details of the samples are included in Supplemental Table 1 .

Clonal expansion index was calculated by dividing the clonal frequency of each clonotype in top-20 expanded clonotypes by the sum (∑) of all singleton clonotype frequency, %   i n   e a c h   s a m p l e ∑ i = 1 n s i n g l e t o n %   i n   e a c h   s a m p l e   where n is the total number of singleton clonotypes. Singleton clonotype was defined as a given clonotype identified only once in a given sample. Simpson clonality is the square root of the sum over all observed rearrangements of the square fractional abundances of each rearrangement (Simpson clonality = ∑ i = 1 n P i 2 where n is the total number of rearrangements, ‘i’ is each rearrangement and P_i is productive frequency of rearrangement ‘i’). Simpson clonality is also the square root of Simpson’s D and is robust across differences in sampling depths. Simpson’s D (also known as Simpson’s dominance index) is the sum over all observed rearrangements of the square fractional abundances of each rearrangement (Simpson’s D = ∑ i = 1 n P i 2 ).

Normalized TCR Richness is calculated by dividing the number of unique productive TCR sequences with the number of total productive TCR sequences.

Data were presented as mean ± SEM. Statistical significance was calculated with unpaired t test or Fisher’s Exact test (24). Analysis was performed using GraphPad Prism version 9.3.1 for Windows (GraphPad Software).

One limitation of single-cell TCR sequencing is that only a few thousand cells were analyzed. To further test our findings of mutual exclusivity of TCR clonotypes in R vs. NR, we performed TCRβ CDR3 DNA sequencing using Adaptive Biotechnologies’ immunoSEQ platform, which allowed us to examine much more productive TCRβ CDR3 sequences. In total, we sequenced 3 responders (1R, 2R, 3R) and 4 non-responders (4NR, 5NR, 6NR, 7NR) by isolating CD8 TILs from 7 individual mice that were treated with anti-PD-L1 ( Figure 1A ). The total numbers of sequenced templates and productive rearrangements were shown for all 7 samples ( Supplemental Table 1 ).

Consistent with single-cell TCR sequencing data, we found that CD8 TIL samples underwent clonal expansion to a similar extent in R vs. NR groups ( Figure 1B ). To better delineate the relative abundance of all TCR clones in the entire repertoire, we also employed repClonality function of immunarch package to analyze our TCRβ sequencing data. We found that the vast majority of TCRβ clonotypes in CD8 TIL samples belonged to hyperexpanded or large clones ( Figure 1C ), demonstrating clonal expansion regardless of R or NR state. We showed the top 10 TCR clonotypes (including VDJ usage and CDR3 sequences of TCRβ) for all 7 samples and their abundance in each sample ( Supplemental Table 2 ).

To better delineate the clonal expansion status of R vs. NR CD8 TILs, we compared the average clonal frequency of the top 20 expanded clonotypes between R vs. NR samples (N=60 for R vs. N=80 for NR). Our data showed no differences in the average clonal frequency of top 20 clonotypes ( Figure 1D ). The number of productive TCR templates varied significantly in different samples due to sampling variation ( Supplemental Table 1 ). In order to minimize the effects of sampling variation, we calculated “clonal expansion index” using a normalization method (See details in Method). Again, we did not identify any significant differences in clonal expansion index between R vs. NR CD8 TILs ( Figure 1E ). Taken together, we conclude that both R and NR CD8 TILs underwent clonal expansion to a similar extent.

Diversity is an important global characteristic of TCR repertoires, which can be generally considered as the number of unique TCR (richness) and their relative abundances (evenness). Various metrics can be calculated to capture one or both properties of TCR diversity. These matrices can quantitatively characterize TCR repertories and the distribution patterns of TCR abundances in a T cell population, which include Simpson Clonality, Simpson’s D (also known as Simpson’s dominance index) and TCR richness. Simpson clonality is a method of quantifying the shape of a repertoire, ranging between 0 and 1, where values approaching 1 indicate a nearly monoclonal population. Simpson’s dominance index ranges from 0 to 1, where values approaching 0 correspond to a polyclonal, infinitely large, perfectly even repertoire and values approaching 1 correspond to a nearly monoclonal sample, where one clone dominates. TCR richness is defined as the number of unique productive TCR rearrangements in a given sample; however, this parameter can be influenced by sampling variation. Hence, we calculated the normalized TCR richness (See details in Methods). Overall, we found no significance differences in Simpson clonality ( Figure 1F ), Simpson’s dominance index ( Figure 1G ) or normalized TCR richness ( Figure 1H ) between R vs. NR CD8 TILs. We conclude that there is no significant difference in TCR diversity between R vs. NR CD8 TILs.

In line with our findings from single-cell TCR sequencing (19), we showed that the commonly expanded TCRβ (defined as >1% of the entire repertoire of a given sample) CDR3 a.a. sequences appeared to be almost mutually exclusive between R (n=30) vs. NR (n=32) samples ( Figure 2A ). Designation to R vs. NR group was determined by average abundance in R vs. average abundance in NR ( Table 1 ). To test if top TCRβ CDR3 a.a. sequences in R group are also observed in NR group and vice versa, we performed a Fisher’s Exact test (24) and found that top R TCRβ CDR3 sequences were much more frequently observed in R than in NR, and vice versa (P<0.0001). For instance, out of 30 R clonotypes, 26 of them were detected in R with a frequency of >1%, whereas only 4 of them were detected in NR with a frequency of >1% ( Figure 2B ; Table 1 ). When we expanded our analysis to clonotypes with a frequency of >0.5%, we still detected a significant difference between clonotypes in R vs. NR ( Figure 2C , top, P<0.0001), namely, the clonotypes in R group occurred much more frequently in R than NR, and vice versa. However, when we counted the clones with a frequency of >0.1% in the opposite outcome group or all clones regardless of their clonal frequency in either group, we did not identify a statistical significance for a given top CDR3 sequence to be observed in R vs. NR group ( Figure 2C , middle and bottom). Taken together, these data show that the top CDR3 sequences (>1%) in R group are also present in NR group, albeit at a much lower frequency ( Table 1 ).

Out of 6907 productive rearrangements identified in 7 samples, only 5 TCRβ CDR3 a.a. sequences were shared among the 7 samples ( Supplemental Table 3 ). The frequency of these 5 shared CDR3 sequences was very low except shared clonotype 1 ( Supplemental Table 3 ). We compared commonly expanded TCRβ CDR3 DNA sequences between R vs. NR and found that they also appeared to be almost mutually exclusive ( Supplemental Figure 1 ). Taken together, our data showed the mutual exclusivity of the top expanded TCRβ CDR3 sequences between R vs. NR, when high frequency clones were counted in the opposite outcome group. Of note, the top TCR clonotypes also differed in a recipient-specific manner within the same group (R or NR), suggesting a highly individualized anti-tumor immune response.

To obtain a broader overview of whether TCRβ features correlate with anti-PD-L1 response, we examined the usage of germline Vβ or Jβ gene segments in individual mice and between R vs. NR group. For each of the 7 samples sequenced for TCRβ, we showed the sum or average of productive rearrangement frequency for all TCR Vβ gene segments ( Supplemental Figures 2A, B ). To better compare the R vs. NR group, we separated the above two parameters into R (1R-3R) vs. NR (4NR-7NR) ( Supplemental Figures 2C, D ). By this comparison, we found a significant difference in the average of productive rearrangement frequency of TCRBV26-01 and TCRBV30-01 between R vs. NR ( Supplemental Figure 2E ).

Similar analysis was performed for TCR Jβ gene segments for all 7 samples ( Supplemental Figures 2F, G ) or R vs. NR samples ( Supplemental Figures 2H, I ). When we compared the average of productive rearrangement frequency for all TCR Jβ gene segments, TCRBJ01-04 and TCRBJ01-05 appeared to be more frequently used by R group; however, this difference was caused by drastic differences in a single R mouse (2R for TCRBJ01-04 and 3R for TCRBJ01-05) ( Supplemental Figure 2I ), consistent with a highly individualized anti-tumor immune response in different mice. Overall, we did not identify any preferential usage of certain Jβ gene segments that were significant between R vs. NR TILs.