For decades, adoptive cell transfer (ACT) of tumor-infiltrating lymphocytes (TILs) has achieved durable tumor regressions, although not consistently. However, the ACT infusion product, which is isolated from tumors and expanded ex vivo, comprises a heterogeneous population of T cells. Recently, the observation that specific T cell states can mediate response to immune checkpoint blockade (ICB) has highlighted the importance of cellular phenotypes in T cell therapies. Now published in Science, Krishna and Lowery et al. investigated the contribution of particular T cell states in ACT efficacy.

To pinpoint which T cell phenotypes correspond with ACT response, Krishna and Lowery et al. assessed infusion products given to patients with late-stage metastatic melanoma, who later either had a complete response (responder) or experienced disease progression (non-responder). First, the researchers analyzed the infusion products from the TILs with CyTOF, uncovering several T cell clusters. One, which was enriched in responder compared to non-responder infusion products, displayed a stem-like phenotype marked by low CD39, CD69, and TIM3 expression, and high CD44, CD27, and CD28 expression. Notably, CD39 and CD69 expression alone could stratify patients by therapeutic outcome. Although infusion products among all patients were predominantly CD39⁺CD69⁺ (double positive, DP), CD39^-CD69^- (double negative, DN) T cells were significantly more abundant in responding patient products. Higher infusion DN cell numbers, as well as the DN/DP cell ratio, correlated with progression-free survival in this patient cohort, implicating a role of the underlying TIL phenotype in clinical outcomes.

Further characterizing these cellular states, the researchers conducted transcriptomic, proteomic, and epigenetic analysis on DN and DP subsets. DN cells expressed markers characteristic of stem-like T cells: TCF7, IL7Ra, and KLF2 along with CD62L, CD27, and SLAMF6. Conversely, DP cells expressed MKI67, TNF, IFNG, and increased levels of TOX, PD-1 and TIM3, suggesting an activated, differentiated phenotype. TCF7, a stemness marker associated with ICB response, was expressed five-fold higher in DN than DP cells. Epigenetic differences were also observed between DN and DP cells using ATAC-seq, with open chromatin regions in DN cells corresponding with TCF7 and KLF4 loci. Supplementing these findings, single-cell RNA sequencing of ACT products uncovered two super-clusters: the first was enriched in ACT responders and displayed a CD39^-CD69^- signature, while the second was enriched in non-responders and featured a CD39⁺CD69⁺ signature. Taken together, these findings indicate that the DN cells possess a stem-like phenotype, which may improve ACT response.

Next, the authors characterized functional differences between isolated DN, single positive (SP; CD39^-CD69⁺ or CD39⁺CD69^-), and DP TILs. Following in vitro stimulation, DN cells expanded into a diverse population of DN, SP, and DP cells; DP cells remained primarily DP. CD69, consistent with its role as an activation marker, was upregulated in all subsets following stimulation, but soon declined in the DN group, while CD39 more specifically appeared to indicate a terminally differentiated state. Over several rounds of proliferation, DN cells expanded 1000x more robustly than DP cells and could continually recognize tumor antigen from an autologous tumor cell line. Furthermore, when adoptively transferred into mice bearing B16F10 tumors, CD39^-CD69^- sorted Pmel T cells better maintained tumor control and survival than CD39⁺CD69⁺ cells, even at lower cell numbers. Thus, DN TILs appeared functionally, as well as phenotypically poised to mediate ACT response.

To generalize these findings, the authors assessed prior datasets from patients treated with ICB. In this setting, higher pretreatment TIL CD39^-CD69^- signatures also correlated with improved response. Interestingly, the CD39^-CD69^- TILs identified within ACT products matched the phenotypes of memory and progenitor-exhausted T cells previously associated with ICB responsiveness. Correspondingly, the DP ACT TILs phenotypically linked to terminally exhausted TILs from ICB, which have been linked to worse outcomes. Building from these results, the authors established a “fitness score” for T cell phenotypes (calculated as the difference between the DN and DP signatures) associated with both ICB and ACT response.

Next, the researchers analyzed the infusion product phenotypes in the context of tumor antigen specificity, critical for T cell-based therapies. Previously, CD39 was reported as a potential proxy for neoantigen recognition, and as expected, neoantigen-specific T cells within ACT products were primarily CD39⁺CD69⁺; however, CD39-negative cells were also observed. Importantly, the proportion of DN cells within neoantigen-specific infusion products was substantially greater in responder compared to non-responder infusion products; responding patients received 23x higher numbers of neoantigen-specific DN cells than did non-responders. To further probe the intersection between T cell phenotype and neoantigen specificity, the team performed paired scRNA/scTCR sequencing on ACT products. In a responding patient, neoantigen-specific TCRs were distributed between DN and DP clusters. This patient’s neoantigen-specific T cell clonotypes generally had high fitness scores and were detected long term in the bloodstream, while one clonotype with a low fitness score was eliminated fairly quickly. In the non-responding patient, neoantigen-specific TCRs mostly clustered within the DP cluster, shared poor fitness scores, and declined rapidly following infusion.

To rule out the impact of TCR avidity in transferred cell persistence, the authors assessed the phenotypes of a NY-ESO-1 TCR-transgenic ACT infusion product in a patient with synovial cell sarcoma who showed a complete response. Of the top 20 clonotypes detected within the infusion product, 70% were enriched in the DN population and 30% were in the DP population. Notably, the DN-enriched clones were detected out to five years after infusion within patients, decreasing slowly over time, while DP-enriched clones declined more abruptly. Thus, similar to TIL ACT infusion products, DN or DP phenotypes were observed despite arising from a single TCR. Furthermore, along with the TIL infusion product results, these findings suggested that infusion product phenotype and ACT efficacy may be linked through persistence of the transferred cells in the body.

In summary, Krishna and Lowery et al. uncovered that TIL phenotypes – specifically CD39^-CD69^- cells – may hold the key to ACT efficacy through their enhanced persistence after infusion. This same cell population correlated more broadly with ICB response, indicating that the functional importance of T cell state may be relevant to multiple therapies featuring T cell antitumor function. As such, ascertaining ideal T cell populations will likely inform engineering strategies for next-generation immunotherapies.

Write-up by Alex Najibi, image by Lauren Hitchings.

Meet the researcher

This week, first co-authors Sri Krishna and Frank Lowery answered our questions.

What prompted you to tackle this research question?
Sri Krishna and Frank Lowery: Under Dr. Rosenberg’s leadership, the Surgery Branch uses cellular immunotherapies to treat cancer. One of Surgery Branch’s most successful adoptive cell therapy (ACT) clinical trials was using tumor-infiltrating T lymphocytes (TILs) to treat patients with metastatic melanoma – a trial in which 56% of the patients responded to ACT, with a quarter of them having complete long-term tumor regression. We have an obligation to learn from this experience to extend ACT’s benefit to treat other difficult-to-treat cancers – especially in patients with metastatic epithelial cancers that are the leading causes of cancer mortality. So, a straightforward question for us was, what did we get right in those clinical trials? What were the essential immunologic components of these patients’ treatment products that made the difference? While this question is continuously being taken on by researchers, typically in the setting of immune checkpoint blockade, we had a unique opportunity in that we could specifically look at the T cells that led to clinical responses. Also, recent advances in single-cell analytical tools meant we could look for relatively rarer phenotypes that may have been masked by traditional bulk analyses. So, we were able to study the T cell markers and gene expression profiles of “infusion products” of this melanoma cohort between complete responders and non-responders through a new prism. The most rewarding moment in the study was when a patient whose metastatic melanoma completely regressed following TIL therapy in 2012 reached out to us to let us know he was pleased that we had done this work – his message made us incredibly emotional and was exceptionally gratifying.

What was the most surprising finding of this study for you?
Sri Krishna and Frank Lowery: A major focus recently in immunotherapies is in identifying tumor mutations and the T cells that recognize them (neoantigen-specific T cells). Our own prior studies, as well as those from multiple groups, have nicely shown that neoantigen-specific, tumor-reactive T cells often express the markers CD39 or PD-1, and that T cells exhibiting these markers are enriched for tumor-reactive T cells. On the other hand, there have also been studies mostly in context of immune checkpoint immunotherapies suggesting that more memory-like T cells are responsible for patient responses, which often lack these “exhaustion” markers. Because our patients received only T cells as a treatment – a heterogenous, yet finite, isolatable, analyzable product – we specifically asked the question, which of these subsets are associated with immunotherapy response? We found that T cell subsets that were enriched for tumor mutation-reactivity were no different between responders and non-responders to ACT.

Scientifically, the biggest surprise was that this smaller subset of memory-like T cells that were CD39^- was higher in responders. This is a bit counterintuitive; we often want to give patients treatment products containing T cells that are the most enriched with reactivity, but it seems like, at least in this cohort, those T cells were not fully responsible, and instead a smaller pool of “stem-like” T cells were. It also presents a problem for the field; finding this small population of less differentiated T cells to give to patients might be challenging. We are working on solving these questions.

What was the coolest thing you’ve learned (about) recently outside of work? 
Frank Lowery: During quarantine I was fortunate to get to spend a lot of time at home with my 2-year-old son. It was great to be with him for such an extended period, to really bond with him and see how his personality and interests changed by the day – vehicles, dinosaurs, space, swimming, watching baseball and movies about vehicles (Cars 3 is our favorite in the trilogy, but Michael Caine’s turn as Finn McMissile in Cars 2 is not to be missed). It’s incredible to see how sometimes he’s a mini-me and other times he’s this total original. While the pandemic has led to so much global suffering, I was incredibly privileged to get this chance to watch him grow up close. Hopefully we make enough progress with cancer immunotherapy that when it comes time for my kids to pick their vocations, they can decide to pursue anything they want and the need for breakthroughs in cancer research is no longer so pressing.

Sri Krishna: The coolest thing I learned about outside of work are the clinical trial results from the Moderna and Pfizer/BioNTech’s COVID19 vaccines. Those two figures from the phase 2/3 trials showing remarkable immunity and >90% protection against the disease should be framed in some museum so one day we can marvel. It really is a testament to the scientific community in the US and worldwide. The way we have learned so much about the epidemiology of the virus and developed treatments and vaccines against the disease in such a short time is truly spectacular. My gratitude to all the researchers, frontline healthcare, and essential workers who are helping us in these difficult times.