A T cell repertoire timestamp is at the core of responsiveness to CTLA-4 blockade.

Biology of the response to anti-CTLA-4 involves the dynamics of specific T cell clones. Reasons for clinical success and failure of this treatment are still largely unknown. Here, we quantified the dynamics of the T cell receptor (TCR) repertoire, throughout 4 weeks involving treatment with anti-CTLA-4, in a syngeneic mouse model for colorectal cancer. These dynamics show an initial increase in clonality in tandem with a decrease in diversity, effects which gradually subside. Furthermore, response to treatment is tightly connected to the shared and public parts of the T cell repertoire. We were able to recognize time-dependent behaviors of specific TCR sequences and cell types and to show the response is dominated by specific motifs. We see that a single, specific time point might be useful to inform a physician of the true response to treatmentThe research further highlights the importance of temporal analyses of the immune response.

Introduction

The inhibition of negative immune regulation to treat cancer is now an essential tool in the oncologist's toolbox. The cytotoxic T lymphocyte-associated antigen 4 (CTLA-4) (Leach et al., 1996; Rowshanravan et al., 2018) and programmed cell death protein 1 (PD-1) (McDermott and Jimeno, 2015; Postow et al., 2015) were the first to be targeted, whereas other receptors for antibody therapeutics are constantly being sought (Sharma et al., 2017; Tang et al., 2018). However, response rates of patients treated with the currently available drugs are not optimal, and a treating physician is left with little knowledge as to the expected course of response (Pai et al., 2019; Pallocca et al., 2019).

Importance of immune checkpoint inhibition in CRC

Recently, colorectal cancers (CRCs) were shown to be promising candidates for immunotherapy (Ganesh et al., 2019). In those cancers, the use of anti-CTLA-4 has been shown to increase the diversity and the composition of the human T cell repertoire (Cha et al., 2014). Unlike other cancer types, immunotherapy treatment of CRC shows long-term improvement and remission. This has been suggested to be linked with the central role the immune system takes in this cancer type (Galon et al., 2006, 2007). As such, immune checkpoint inhibition has been the focus of other research, with other targets such as LAG3 (Long et al., 2018), TIM3 (He et al., 2018), and TIGIT (Harjunpaa and Guillerey, 2020), which are currently tested and are in various phases of clinical trials. These findings, although mostly done in humans, led us to design a study focused on the changes that the T cell repertoire goes through over the time course during and after treatment with anti-CTLA-4.

Repertoire Sequencing in CRC

To see changes in the immune response, and especially changes that are the result of immunotherapy, it is interesting to look at repertoire sequencing (Rep-seq, Benichou et al., 2012, or AIRR-seq, Rubelt et al., 2017). Such works often include two metrics used extensively by multiple groups to quantify the T cell repertoire: diversity and clonality. The origin of these measurements is metrics used to assess and quantify populations in various ecosystems, where the plurality of species is of great importance. Diversity is a measurement that tells us of the total number of species (in this work, total number of different T cell receptors [TCRs]). Diversity in essence does not take into account the number of items (number of T cells) that belong to each species (to each T cell clone). Clonality, on the other hand, does include information regarding the number of items in each species (the number of T cells in each clone). Clonality is higher when a specific species includes more items (when a specific T cell clone is more abundant).

Several works focused on diversity and clonality of the TCR repertoire following immunotherapy (Beer et al., 1988; Field et al., 2017; Hosoi et al., 2018; Robert et al., 2014; Valpione et al., 2020) and all found that the expected increase in clonality and decrease in diversity often precede a critical point in treatment. Other works, such as the computational analysis of CRC performed by Cha et al. (2017) detected 1,940 distinct antibody peptides and 163 pairs of antibody peptides with a single amino acid variant that showed correlation with survival times of sick individuals. Other works focused on tumor-infiltrating lymphocytes (Treiner et al., 2003) found both in tumors and in the surrounding tissue, to study their TCR repertoire (Sherwood et al., 2013), or on sequencing tumor-draining lymph nodes (Matsuda et al., 2019) to differentiate metastasis and non-malign tissues. Recent works (Zhang et al., 2019) used single-cell RNA sequencing (scRNAseq) to further explore the tumor-infiltrating lymphocytes of patients with CRC, whereas the TCR repertoire itself might be used as a liquid biopsy alternative (Snir and Efroni, 2020). These studies show that the research community identified the importance of quantifying the TCR repertoire of patients with CRC. These findings and their conclusions led to the work presented here.

Time-dependent analysis

Over the past few years, several works studied time-dependent changes to the TCR repertoire; Oh et al. identified stages in treatment where diversification of the TCR repertoire takes place (Oh et al., 2017), whereas other works focused on the importance of a diverse TCR repertoire to avoid adverse events following immunotherapy (Cha et al., 2014) and showed that such events often occur after a drastic change in the diversity and clonality of the TCR repertoire (Subudhi et al., 2016). Although these articles, and others like them, highlighted the importance of timing and of time-dependent analysis of the TCR repertoire, a straightforward study to follow the TCR repertoire has not yet been published, despite its translational importance.

Research motivation

Our research aims to improve the understanding of changes within the TCR repertoire in mice treated with anti-CTLA-4, with a special emphasis on the time-dependent behavior of T cells. We present evidence that highlights not only the importance of choosing the correct type of analysis performed but also the timing of sampling. We also note several cell types according to their TCR association with mucosal-associated invariant T cells (MAIT) and invariant natural killer T cells (iNKT) cells that have been previously shown to accompany the immune response in treated mice. We believe the research solidifies the importance of a time-dependent analysis of immunotherapy treatment.

Results

Experiment design

To set up a system that allows the monitoring of tumor progression and tumor regression following anti-CTLA-4 checkpoint immunotherapy, in tandem with measuring the repertoire, we used the MC38 syngeneic mouse model. In an anti-tumor efficacy experiment, we used 6-week-old female C57BL/6 mice (Strain C57Bl/6NAnNCrl, Charles River). Freshly cultured MC38 tumor cells (106 in 100μL PBS) were implanted subcutaneously into the left flank of all animals. Following implantation, primary tumor volume was determined daily by caliper measurement using the formula W2 × L/2 (L = length and W = the perpendicular width of the tumor, L > W). Eight days after tumor inoculation, mice were randomized into groups of 15 mice in the anti-CTLA-4 treatment group and 5 mice in control (untreated) group (day 0). Animals were treated intravenously on days 1, 3, and 6 starting on the day of randomization. Vehicle control animals were treated with 10 mL/kg body weight PBS. Blood samples were collected on days 0, 7, 14, and 21 (Figure 1). A total of 4 × 20 = 80 blood samples were taken throughout the experiment, and tumor (4 samples) and spleen (20 samples) were taken at the end of the experiment. We isolated peripheral blood mononuclear cells (PBMCs) by density gradient centrifugation. Library preparation and next-generation sequencing of the repertoire were performed as previously described (Rosati et al., 2017). Briefly, 100 μL blood was spun down and pellets were further processed and lysed with ACK buffer to remove erythrocytes. The resulting PBMCs, tumors, and spleens were processed for their RNA contents. 125 μg from each RNA was applied to SMARTer-Mouse-TCR-Profiling-Kit (Clontech) to obtain DNA libraries, which we then sequenced using Illumine MiSeq (v3 kit). We analyzed the obtained reads for their Rep-seq data using MiXCR (Bolotin et al., 2015). This workflow provided us with TCR repertoire data for all samples. As is known for the specific anti-CTLA-4 antibody clone we used (9H10), all mice responded to treatment (see also Simpson et al., 2013).

Figure 1

Experimental workflow

C57BL/6 mice were implanted with freshly cultured MC38 colon tumor cells. One group (red arrow, horizontal, 15 mice) was treated with the anti-CTLA-4 clone 9H10 on days 1, 3, and 6, whereas the control group (5 mice) remained untreated. We have drawn blood samples once a week, starting day 0, during a period of 3 weeks (purple vertical lines). From these blood samples, we obtained PBMCs and processed their RNA to produce TCR repertoire sequencing. In addition, tumor (when available) and spleen samples were collected with animal sacrifice at the termination of the experiment. Tumor volumes were measured on days 0, 2, 5, 7, 9, 12, 14, 16, 19, and 21 (black vertical lines in the figure).

Changes in clonality and diversity are detectable at weekly sampling resolution and subside over time

Human studies showed that treatment with anti-CTLA-4 caused expansion of specific clones (Friese et al., 2020). To examine this phenomenon, we measured changes in tumor size, T cell clonality, and diversity in each mouse, in each time point (Figures 2A–2C). As expected, only treated mice showed a decrease in tumor size as their tumors receded (not all control mice lived to day 21). To study diversity, a population size metric that is often informative in repertoire analyses (see Transparent Methods), we calculated the inverse Simpson index of the TCR repertoire in alpha chain in each animal in each time point. We found that in treated mice, beta-chain diversity decreased significantly in day 7 (D7), and then gradually increased in D14 and D21 (Figure 2B). Diversity is inversely correlated to clonality (Figure 2C). When we quantified clonality in each mouse, we noticed a clear difference between the response rate in mice with low clonality (clonality <0.034) and mice with high clonality (clonality >0.034); mice with higher levels of clonality reacted faster to the treatment, and with better final outcome (e.g., a smaller tumor size) (Figure 2D). The same phenomenon in the clonality was observed in alpha chain measurements (Figures S1A–S1C).

Figure 2

Changes in clonality and diversity subside over time and are distinctive to different response groups

(A–C) Changes over time in tumor size (See also Table S1) (A), diversity index for beta chain (B), and clonality index for beta chain (C) in the control group and in the treated group.

(D) Changes over time in the average tumor size. The treated mice in this panel are divided into two groups, according to their clonality index—low clonality (blue lines) and high clonality (orange lines).

(E) The abundance levels (copy number) of the 10 most expanded clones in the control and treated groups in an amino acid view, in beta chain.

(F and G) (F) Fold changes of the 10 most abundant clones in the different response groups (early, mid, late, see text) and (G) combined.

See also Figure S1. Data of all panels are represented as mean ± SEM.

The control group did not display significant variability in its response throughout the experiment, whereas treated mice showed a significant increase in the frequency of the 10 most abundant clones in beta chain (Figure 2E). These changes are most noticeable in D7 compared with other time points. The same observation was seen in alpha chain measurements (Figure S1D). Such changes were not found in the control group, suggesting that there are specific clones relevant to the disease in question that could be linked to the response to treatment.

These results suggest that treatment with anti-CTLA-4 leads to an expansion of some clones and to the contraction of others, leading to an increase in clonality and a decrease in diversity at the time point immediately following treatment (D7 here). As different sample sizes usually create an incomplete representation of the repertoires, we examined the 10 most expanded clones in each sample (the assumption is that even though the number of sampled CDR3 sequences is different between samples, the top clones remain at the top). To quantify these changes within the top 10 clones, we used D0 as a baseline. That is, we divided the clonal frequencies obtained in D7, 14, and 21 with that of D0. In Figure 2F we show these fold differences in the clones as they appear in four groups: the control group and three other groups that were obtained from the treated group. Full explanation of the three groups is in the Transparent Methods section. Briefly, the names “early,” “mid,” and “late” stand, respectively, for the animals that responded early to anti-CTLA-4 treatment; the animals that responded later to treatment, and the animals that responded much later to treatment. See quantification in Transparent Methods. As seen in the figure, the response groups differ among themselves in their clonal behavior but differ greatly from the control group. This is averaged in Figure 2G, which clearly demonstrates an expansion of the dominant clones at D7, post-treatment. This expansion is absent from the control group. We can see the same behavior we discussed earlier—the only time point that demonstrates a significant increase in the top 10 clones is D7. Additionally, we see that whereas the control group presents a mostly constant abundance of the top 10 clones, the treated group experiences significant expansion of clones in D7. The same observations were also found in the alpha chain analysis (Figures S2E and S2F).

Repertoire-based stratification of mice is associated with response to treatment

We looked at the response to checkpoint inhibition treatment and its association with the identity of the T cell clones specific to each animal pre- and post-treatment. Here, we see a correlation between the abundance of public clones (clones that are shared between mice, see Transparent Methods for definition) and response to treatment. We also see a correlation between the response and the level of convergent recombination (CR) (Smyth et al., 2000), the phenomena in which different nucleotide sequences code for the same amino acid sequence. Identical TCRα or TCRβ amino acid (Harjunpaa and Guillerey) sequences originate from different nucleotide (Pallocca et al., 2019) sequences (Quigley et al., 2010).

T cell repertoire work often discusses two types of clones: Private and Public (Venturi et al., 2008) (Li et al., 2012). Private clonotypes are tagged as such if their CDR3 (amino acid) sequences appear in only one animal in the study; Public clonotypes are tagged as such if they appear in more than one animal. We further divide Public clonotypes into two subgroups as done before (Covacu et al., 2016); we define clonotypes as Public Inclusive if they appear in more than one phenotype group of mice and Public Exclusive if they are shared only within their own phenotype group. We found that the frequency of Public Exclusive clones is fundamentally different in treated mice (all three groups, early, mid, late) from that of the control group, that is, the treated groups have a relatively higher portion of clones that are shared only within the same group. Starting at D7 (Figure 3A), these differences are most apparent from the contrast between the control group and the early-responders. The differences between the Control group and mid, early, and late can be further seen at D21 (t test: p < 2.1 × 10−5; p < 1.8 × 10−4; p < 8.2 × 10−4, respectively). Early-responders further show a significant increase in their frequency from D0 to D21 (t test: p < 0.0014). Early- and mid-responders show a significant increase between D14 and D21 (t test: p < 5.1 × 10−4; p < 0.0082, respectively). In addition, early-responders demonstrate the highest level of Public Exclusive sequences immediately after treatment, suggesting that this set of clones may be instrumental to the response.

Figure 3

Repertoire-dependent response to treatment

(A) Frequency of Public Exclusive clones in each of the response groups over time. Data are presented as means ± SEM (∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001; Student's t test). For the sake of legibility, the panel highlights a small set of some of the significant differences between the different levels. There are many additional significant contrasts.

(B) Overlaps between samples. We used the 100 most abundant beta chain clones in each sample for the calculation. The metric we used is Morisita's index (see text). The panel highlights inner similarities within samples from the same animal.

(C) The average overlap within and between groups. Each rectangle represents the average number of overlaps within all samples in a specific combination. To calculate an average overlap, we used the average over all sample combinations between two groups.

(D) Changes, over time, in the averaged CR levels in the control (Rubelt et al.) and treated (blue) groups. Data are presented as means ± SEM. The panel indicates the significant increase in CR in the treated group at D14, in contrast with the significant decrease in CR in the control group on the same day. Interestingly, the curves meet at D21.

See also figure S2.

To quantify the similarities between samples, we calculated the number of sequences shared between sample pairs. To do that, we used the 100 most abundant clones and “Morisita's index” (Rempala and Seweryn, 2013) to calculate the overlap. An analysis of alpha chains did not show higher levels of overlaps between samples within the same animal (Figure S2A). However, an analysis of the beta chain demonstrated a distinct higher overlap within samples from the same animal (Figure 3B). Quantification of the overlaps between and within the different response groups (Figure 3C) demonstrates that the control group shares a relatively lower overlap with all treated groups. For example, the overlap between control and early-responders is lower than the overlap between mid-responders and early-responders’ samples (0.1541 and 0.1572, respectively, p < 0.013). Also interesting is the fact that samples within response groups share similar repertoires.

CR has been shown to behave differently under different conditions (Venturi et al., 2008) (Quigley et al., 2010). We, and others, have shown that an increase in CR levels correlates with an increase in sharing levels (sharing level is simply measured as the number of animals sharing a specific sequence) (Figure S2B). In the temporal behavior we report here, early-responders show a post-treatment (D7) reduction in CR levels compared with all three other groups (control, mid-responders, and late-responders). In the other groups, D7 CR levels increased (Figures S2C and S2D). When we compared the CR levels of control and treated groups, we identified a significant difference in D14 (Figures 3D and S2E). As T cells do not go through in vivo mutations, the increase in CR could mean either a thymic-based activity that increases in response to a peptide presented in the thymus, or, alternatively, an increase in the peripheral abundance of multiple T cell clones directed at the same peptide. Such an increase would lead to higher detection within sampling of the observed repertoire.

Network analysis of clones

The network structure of the relationships between clones is useful in visualizing and learning about the T cell repertoire. Here, we created a network representation of the top 100 clones and their abundance. To do that, we looked at the unique set of the 100 most abundant clones in each sample (this set includes ~3,200 sequences). We used this set to create a network in which the clones themselves are nodes, whose sizes are determined by the abundance of each clone. The clone size is normalized within mice; that is, for each mouse, we calculated the relative abundance of a clone compared with its frequencies in all other time points of the same mouse. Edges between nodes appear when the editing distance between sequences, calculated as Levenshtein distance, is 2 or smaller. We cluster sequences into groups of sequences with high similarity. Figure 4 shows the networks of these TCR sequences, such that every row represents the repertoires of control mice, early-responders, and late-responders, over time (time progresses from left to right). The represented network is that of the complete set of 3,200 sequences. For each sample, only sequences that appear in the sample are highlighted, with their frequency represented by node sizes and with clusters associated through color. The figure shows that early-responders follow a different longitudinal behavior. Specifically, in D7, where the control and late-responders’ networks demonstrate a constant longitudinal behavior, with multiple high-frequency clones, the early-responders show several large clones. This D7 behavior suggests that these are the clone clusters that provide better protection, or that respond better to anti-CTLA-4, for this group of mice.

Figure 4

Network analysis of clones

Networks of a set of sequences assembled by collecting the 100 most abundant clones in each sample. Three different groups are visualized (mouse D = control; mouse L = early-responders; mouse H = late-responders). For each mouse, we see progress in time from left to right. The leftmost network is D0, whereas the rightmost network is D21. The size of the node represents the relative abundance of a clone within a mouse. We connected and clustered AA sequences according to their Levenshtein distance. An edge between two nodes appears when the distance is smaller or equal to 2.

Response and behavior are related to presence of specific TCRs

Finally, we identified specific TCRs, whose sequences are associated with iNKT and MAIT. These sequences demonstrated time-dependent behavior across all phenotypes (Figure 5). iNKT cells were previously reported in the human blood pool, comprising 0.01%–1% of PBMCs (Gumperz et al., 2002), and were categorized as important immunoregulatory cells (Kronenberg, 2014; Van Rhijn and Moody, 2015). iNKT cells are some of the first cells to respond during an infection. Pre-clinical mouse studies have clearly demonstrated that iNKT cells play a critical role in tumor immunosurveillance (Smyth et al., 2000). Figure 5A shows that iNKT cells behave in a time-dependent manner in the four response groups. Interestingly, the reduction in iNKT levels from D7 to D14 is in all the treated groups (this is different from MAIT cell behavior, see Figure 5B and further discussion below) and is less significant in the control group. It is not possible to see any relevant behavior if we are not looking at the abundance of clones, but instead look only at the set of unique clones (Figure S3A). When we looked at the abundance of iNKT cells (from spleens) we identified a significant difference between treated and control mice (Figure 5C).

Figure 5

Response and behavior are related to the presence of specific TCRs and cell types

(A) The frequency of iNKT cells in alpha chain for all four different groups.

(B) The frequency of MAIT cells in alpha chain.

(C) The frequency iNKT cells in spleen samples in control and treated mice.

(D) The frequency of CASSLELGGREQYF (the most shared clone observed, see Table S2) in the different groups. Data of all panels are represented as mean ± SEM.

See also Figure S3.

MAIT cells (Treiner et al., 2003), were previously reported as an abundant population in humans, where they represent up to 10% of blood T cells, whereas their role in immunity remains unclear (Godfrey et al., 2019). Despite their similarities, there are some striking differences between iNKT cells and MAIT cells; for example, most mouse MAIT cells produce the cytokine IL-17A, and only a small subset of these cells produces the cytokine IFN-γ, whereas the opposite is true for iNKT cells. (Furthermore, in mice, iNKT cells outnumber MAIT cells 10- to 100-fold in most tissues, whereas the opposite applies for humans (Godfrey et al., 2019)). Figure 5B demonstrates how MAIT clone expansion is associated with response to immunotherapy. Although the unique repertoire of MAIT sequences did not show any significant difference (Figure S3B), MAIT abundance was higher at D0 in the group we identified as early-responders; they were significantly low in the mid-responders as well as in the late-responders. MAIT frequency does eventually go up post-treatment in those two groups. It may be that this initial abundance of MAIT cells in these specific animals is the reason they shared early response.

To identify a specific dominant group of clones, we isolated the set of clones that are within the 100 most abundant clones in D7 (which we previously identified as critical in the response) and that are shared between at least two time points of the same animal. Table S2 gives the list of 18 clones that we identified in this way in the treated mice. The sequence in the first line of the table is shared between seven of the treated mice (Figure 5D).

T cell repertoire annotation and motif discovery

We applied the TCRnet analysis as previously used to discover motifs in TCR sequencing data; this method is used to discover groups of homologous TCR sequences that are unlikely to be observed by chance and therefore are a result of antigen-driven expansion (Pogorelyy and Shugay, 2019; Ritvo et al., 2018). Our results indicate the presence of several large groups of TCR clonotypes having similar TCR sequences for both alpha and beta TCR chains. By treating repertoire as a set of homologous TCR clusters, we were able to profile the behavior of TCR groups that are likely to target the same antigen. These groups display a characteristic behavior with T cell abundance peaking at D7, and we were able to identify conserved CDR3 sequence motifs characteristic for TCR groups displaying the most prominent change in abundance.

Discussion

The experiments undertaken in this article were designed to characterize the dynamics of the T cell repertoire in mice treated with anti-CTLA. The results led us to discover that the repertoire is translatable into important metrics and to discover the importance of the frequency of sampling during treatment.

Response to immunotherapy treatment is difficult to quantify, mostly because multiple factors control this response. In the experiments we described earlier, we have indeed seen the animals go through different disease courses. We have seen varying responsiveness to immunotherapy treatment, even though the mice are of identical genetic background. All treated mice eventually eliminated the tumors, but to do so, they went through major differences in their tumor volumes. What we show here, is that by measuring the T cells' repertoire clonality, diversity, and other metrics relevant to the immune response, we identified time point D7 as the one time point that informed of significant differences in future responses. Although an increase in clonality might be expected, we show that this clonality might be completely overlooked if the measurement is taken too soon, or too late.

In addition, the pathological display of changes in the disease, in this case the tumor size as it has been measured in animals, and the display of changes to the quantified behavior of the immune system, through changes in the repertoire, do not appear simultaneously, but at different times. This detachment between the two metrics adds to the importance of placing our attention over the correct temporal behavior.

We show, through network analysis of the dynamics of available TCRs and their sequence similarities, that connections between clones increase and decrease with time. These emergent networks show an association between clones and the general rate of response to treatment.

We then look at Public (shared between multiple animals) and Private (unique to one animal) clones, where we see that among the Public clones, a subset of clones that we termed “Exclusive,” as they are exclusive to a specific phenotype, are expanded within the group of early-responders, over the 4 weeks duration of the experiment; This expansion is in stark contrast to the behavior of clones that are exclusive to the control group. The Control-Exclusive clones diminish over time, perhaps as they are not boosted (unlike the treated groups) by immunotherapy (Figure 3A).

We further identified unique cell types and followed their dynamics throughout the experiment. Recent articles showed the importance of MAIT cells in response to immunotherapy treatment. Here, we see that the MAIT cells we identified out of the repertoire were tightly connected to drug response (Figure 5B). It could be that these cells are influenced directly by the drug itself. In a similar manner, and following the interest in iNKT cells and their involvement with cancer response (Patel et al., 2020; Yamashita et al., 2018), we see that TCR repertoire data identifies iNKT cells in a manner that shows expansion of iNKT clones in response to treatment. Specifically, this expansion is in D0 and D7, making iNKT cells first responders to the tumor. The expansion pattern changes in D14 through 21, but it does not change in the group of mice that responded late. Interestingly, the drop in iNKT clone sizes from D7 to D14 is unique to the treated mice groups, and is less significant in the control group. If we do not look at clone sizes, but instead only at clone lists (that is, without registering their abundance) the effect disappears (Figure 5A). Also, when we studied iNKT cells in spleens of treated and of untreated mice, we saw a major difference in their abundance (Figure 5C).

We used motif discovery to identify group of TCR sequences whose expansion is driven by antigens. As Figure 6 shows, both TCR alpha and TCR beta represent such clonotypes. We can see the temporal behavior of the clonotypes identified in Figure 6, and in Figure S4. As the figure indicates, the clonotypes demonstrate extensive expansion in the treated group at D7. That expansion is missing from the control group, which means it is driven by the treatment. Furthermore, this behavior is not homogeneous across all treated mice, but rather is more apparent in a subset of treated animals. Finally, motif analysis showed the presence of several large groups of TCR clonotypes, with similar TCR sequences for both alpha and beta TCR chains. These groups peak at D7 and include conserved CDR3 motifs most likely involved in the immune response. We intend to further explore this venue of results.

Figure 6

Cluster of clone abundance

Both TCR alpha (left) and TCR beta (right) show differences in clone sizes.

See also Figure S4.

Together, we present strong evidence that the timing and response to immunotherapy are connected to metrics available from the T cell repertoire. Within that general term, the T cell repertoire, we see specific behaviors in the days surrounding a specific time point, which in this experiment is D7, and which undoubtedly calls for specificity in different settings of different tumors and different treatments. We also see specific behavior of clonal networks, clonal sharing, and iNKT and MAIT cells. Further research is needed to identify more T cells and TCR sequences involved in the immune response to the types of cancer covered here, updating personalized cancer therapy with another layer of information to be used in decision-making. We believe that this research should lead to research in human samples and in clinical settings, to allow physicians access to the temporal behavior of the immune response. This knowledge might help physicians make well-informed decisions regarding treatment choices. The translational implications of these findings might be critical. It is technically straightforward to collect blood from human patients before, during, and after drug treatment. This is not routinely done at the moment. The ability to obtain T cell repertoire data from these samples is, at least technically, straightforward as well. If we are able to translate our findings here, of an indication available from these T cell repertoire data into an indication available from human data, that could provide the physician with an additional layer of reconnaissance, informing that in spite of tumor progression, T cell response is still at full strength. Decisions regarding treatment might need to be re-evaluated.

Limitations of the study

Although the study design included responders and non-responders, the final results showed that all mice presented complete response. Although our initial goal has been the stratification of responders and non-responders, this complete response led to two outcomes: (1) a limited number of tumors to work with (three out of the control group; two out of the treated group) and (2) the impossibility to contrast responders and non-responders. Additionally, the small number of available tumors prevented the follow-up of TIL repertoire with peripheral blood repertoire. Another limitation is inherent to the difference between a strain of mice and the heterogeneity of human samples and human phenotypes. The distances between repertoires of the same strain are not expected to represent the distances between individual patients.

Despite the above-mentioned limitations, we believe that we have demonstrated response-related repertoire metrics. This, together with the temporal nature of our data, is the result presented in the article.

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Sol Efroni (sol.efroni@biu.ac.il).

Materials availability

The syngeneic mice model used in this study was generated by the Charles River Discovery Research Services in Germany, Study Number: P559A1, Experiment Number: S006. We did not generate any unique reagents throughout the current study.

Data and code availability

The raw data reported and analyzed in this article have been deposited in the Sequence Read Archive in National Center for Biotechnology Information (NCBI) under the BioProject accession number “NCBI: PRJNA658348”. The code used in this study is available from the corresponding author, Sol Efroni (sol.efroni@biu.ac.il).

Methods

All methods can be found in the accompanying Transparent Methods supplemental file.