Neoadjuvant Immunotherapy for Hepatocellular Carcinoma

Introduction

Immunotherapy has revolutionized the treatment of many cancers including hepatocellular carcinoma (HCC). HCC is unique among cancers due to its resistance to chemotherapy; instead, tyrosine kinase inhibitors (TKIs) have been the mainstay of treatment for HCC for the past two decades.¹ However, immunotherapy agents targeting programmed death receptor (PD-1) and its ligand (PD-L1) have shown significant clinical activity as a second-line therapy in patients whose disease had progressed following treatment with TKIs, and a potentially higher survival when combined with CTLA-4 blocking antibodies.^2–4 More recently combination approaches with PD-(L)1 blocking antibodies given with either the VEGF antibody bevacizumab or lenvatinib, a multi-targeted TKI, have showed significant activity in the first-line setting far superior to the historic TKI comparator, sorafenib.⁵ While these agents have changed the treatment paradigm, the majority of patients with HCC still fail to respond to therapy, and we lack a firm understanding of the heterogeneity of clinical response as well as validated biomarkers to help define who will benefit from these therapies.

PD-L1 immunohistochemistry (IHC) has been used in other malignancies as a biomarker of response, though in most malignancies it has relatively poor predictive power, and in HCC it remains controversial whether it is useful in guiding therapy.⁶ Additionally, given that HCC can be diagnosed radiographically, the use of a tissue-based biomarker will require a paradigm shift in the clinical approach to diagnosing and treating HCC. Pre-treatment and on-treatment biopsies from patients receiving therapy in the context of a clinical trial is the ideal way to identify predictive biomarkers, though given the heterogeneous nature of tumors—particularly in tumors where there has been a response to therapy and significant intratumoral necrosis—the utility of small-bore needle biopsies in defining pre-treatment predictors of response, as well as dynamic changes induced by therapy, is questionable. Even in clinical trials with mandated biopsies, the tissue quality is often suboptimal, and typically high-quality pre- and on-treatment biopsies are successfully collected from only a minority of patients.

Neoadjuvant clinical trials offer a unique setting not only to improve outcomes for patients with localized HCC, but to better understand how novel therapies work in patients with HCC, which can inform biomarkers and identify optimal combinatorial therapeutic approaches. The standard approach for patients with localized disease and intact liver function is surgical resection; however, even in patients with tumors smaller than 2 cm the risk of recurrence nears 70%,⁷ and no perioperative therapy has demonstrated significant survival benefit to date.⁸ Given the success of immunotherapy in the advanced HCC setting,^4,5 and the clinical efficacy of perioperative immunotherapy in other malignancies that has led to FDA approvals in the past year,^9–13 we would hypothesize that perioperative immunotherapy may significantly benefit patients with resectable HCC, and recent trials support this hypothesis.^14–16 Here we review the current understanding—and scientific and clinical promise—of perioperative immunotherapy in HCC.

The Clinical and Scientific Potential of Perioperative Therapy in HCC

Neoadjuvant trials offer a unique clinical setting to accurately define the mechanism of action of novel immunotherapies and combination approaches.¹⁷ In general, neoadjuvant therapy is used with two goals in mind: (1) reducing the chance of recurrence through elimination of micrometastatic disease not visualized on imaging, and (2) cytoreduction of the primary tumor to facilitate subsequent definitive therapy. Additionally, response to therapy may be used as a predictive and prognostic marker and to guide clinical decisions around adjuvant therapy. However, brief pre-operative therapeutic intervention trials, often termed “window-of-opportunity” trials, may be designed not necessarily with the intention of downstaging the tumor, but rather to study the effect of therapy. This said, these trials—and this therapeutic approach—do not represent an exploitation of humans for the purpose of research; they in fact may have the potential to improve recurrence-free survival while at the same time not significantly delaying surgery.^14,18–22 Tumor shrinkage and/or frank necrosis prior to surgery is sometimes observed in this setting, but the primary target of therapy is micrometastatic disease that can seed eventual recurrence since the clinically evident tumor will be surgically removed.

Additionally, response rates to immunotherapy in the neoadjuvant setting across cancer types appear equivalent to, if not higher than, those observed in the unresectable and/or metastatic setting.^14,18–23 Neoadjuvant immunotherapy alone or in combination with chemotherapy is now standard therapy for triple negative breast cancer, and promising phase 2 and phase 3 clinical trial data in multiple other tumor types will likely lead to broadening of this approach in the coming years.^9–13 There are many reasons why operative candidates may be more responsive to immunotherapy than patients with more advanced disease. Operative candidates are typically treatment naïve and have more robust immune systems; not only does more advanced cancer potentially induce systemic immunosuppression through a variety of mechanisms, but immunotherapy is often given following many other lines of therapy that may have further taxed the immune system. Additionally, smaller resectable tumors typically demonstrate less intratumoral heterogeneity (including mutational/neoantigen heterogeneity) and less loss of heterozygosity that may enable immune evasion.^24,25 Indeed, there are numerous clinical studies in many cancer types demonstrating that tumor burden inversely correlates with response to immunotherapy.^26–28

With improving rates of screening of high-risk individuals, more HCC patients are being identified with early-stage HCC amenable to surgical resection, however, the vast majority of these patients’ tumors recur, and to date perioperative therapy has demonstrated enough of a significant survival advantage to warrant widespread adoption.^8,29,30 A number of studies are underway assessing the efficacy of adjuvant immunotherapy following curative-intent treatment of HCC, including the registrational CheckMate 9DX study looking at the use of nivolumab after resection or ablation; clinical efficacy of adjuvant immunotherapy has already led to its approval by the FDA in multiple other cancers including non-small cell lung cancer and melanoma.^31,32 However, neoadjuvant—which we can more accurately call “perioperative” immunotherapy since anti-PD-(L)1 antibodies have long half-lives and will remain in circulation post-operatively—is posited to be superior to adjuvant immunotherapy alone. Use of immunotherapy preoperatively augments anti-tumor immunity, likely due to priming of new T cell responses against tumor neoantigens while the tumor still remains in vivo, while in the adjuvant setting the only neoantigens that remain are from micrometastatic disease, which may translate to less immune priming and activation.^11,33 Pre-clinical modeling lends credence to this argument, with improved survival and enhanced tumor-specific CD8⁺ T cell activation in mice administered PD-1 blockade pre-operatively compared with those that received only post-operative therapy.³⁴ Limited clinical evidence in melanoma and lung cancer also supports the use of neoadjuvant over adjuvant immunotherapy and, in an albeit small cohort, there are data showing augmented T cell clonal expansion when immunotherapy is administered pre-operatively.^11,18

Additionally, neoadjuvant treatment therapy may be superior to treatment in the adjuvant setting, as surgery itself can be an insult to systemic immunity. Post-operative immunosuppression, associated with surgical stress, is mediated by elevated alarmins such as damage-associated molecular patterns (DAMPs) and peripheral cytokines including IL-8 and IL-6, leading to local and systemic inflammation.^35,36 This suppresses cellular immunity, hindering both de novo T cell priming and activation and potentially further limiting efficacy of adjuvant immunotherapy with checkpoint blockade.^37–39 Resection and disruption of the tumor microenvironment and associated draining lymph nodes may also impact the efficacy of immunotherapy, as PD-1 targeting agents may in part act to aid in T cell expansion at the site of tumor-draining lymph nodes, and surgical disruption of normal lymphatics may decrease the efficacy of adjuvant immunotherapy compared with neoadjuvant.

In addition to providing potential clinical benefit, neoadjuvant trials provide a unique window to characterize the dynamic effect of immunotherapy in vivo. As opposed to the scant biospecimens available from clinical trials in the metastatic setting, neoadjuvant trials provide blood and tissue specimens pre-treatment and post-treatment (time of resection), greatly facilitating characterization of the effect of an intervention.⁴⁰ Using emerging immune monitoring capabilities including platforms such as multiplex IHC and spatial transcriptomics, we can characterize the changes induced in the three-dimensional interface between tumor, stroma and immune infiltrate at baseline and post-treatment, so as to understand predictors of response and spatial determinants of resistance that may be targeted by more optimal combination approaches.^41–47 Additionally, single-cell technologies enable us to characterize at the transcriptomic level the complex heterogeneity within the intratumoral lymphoid and myeloid cell lineages, enabling us to resolve very small subpopulations, such as the stem-like T cell populations that we now believe are responsible for the efficacy of PD-1 blockade.^48–51 Additionally, through neoantigen calling and incorporation of T cell receptor (TCR) sequencing, we can resolve the tumor-specific T cells from the conglomerate of tumor-infiltrating T cells, the majority of which are bystanders and not reactive to tumor neoantigen.^18,52 Resolving and phenotyping the pre-existing T cells, as well as those de novo clones induced by therapy, will enable us to more precisely target our immune modulation to activate this all-important subset, with the aim to increase response to therapy while decreasing toxicity associated with pan-activation of the immune system.

There is increasing understanding of the key role myeloid cells have in promoting a permissive immune milieu to grow, and determining response to immunotherapy. It has long been known that myeloid cells—specifically macrophages and granulocytes—comprise a large portion of the cellular composition of tumors of multiple tumor histologies.^48,53,54 Recent deep profiling of the immune compartment within HCC lesions has demonstrated that Kupffer cells, the tissue resident macrophage within the liver, are lacking from the tumor microenvironment, and rather macrophages derived from inflammatory monocytes recruited to the tumor make up the majority of the intratumoral macrophage population.⁵³ These macrophages have a resolutely immunosuppressive phenotype and likely aid in promoting tumor growth, and potentially resistance to therapy. There are numerous agents in clinical development targeting intratumoral macrophages, but many of these agents (e.g. antibodies targeting the CSF1-R receptor present on all macrophages) deplete the entire lineage rather than the specific subset of immunosuppressive monocyte-derived macrophages. Depletion of macrophages capable of priming anti-cancer immunity along with the suppressive subsets is likely one reason for the net neutral benefit that these therapies have demonstrated in clinical trials to date.⁵⁵

Neoadjuvant Therapies Demonstrate Early Signs of Efficacy

Neoadjuvant as well as adjuvant TKI administration has failed to show benefit in the randomized phase 3 setting.^8,56 There is, however, a strong pre-clinical rationale for the combination of TKI with immunotherapy, and this has shown efficacy in the advanced setting.⁵⁷ One of the first published reports of the use of neoadjuvant immunotherapy in HCC was from a cohort of patients with locally advanced disease, where nivolumab was given in conjunction with the TKI cabozantinib to assess the potential to downstage tumors in order to enable curative-intent resection.¹⁶ This trial enrolled 15 patients with HCC for whom upfront surgical resection was not recommended due to high-risk features such as multifocal disease, portal vein invasion, or tumor diameter >10 cm. Patients received a 2-week lead-in with the TKI cabozantinib, followed by 4 treatments with nivolumab added to the TKI, and initial restaging was performed at 10 weeks. Twelve patients successfully underwent resection; in five a major pathologic response (defined as ≥90% necrosis) was noted, with one achieving a pathologic complete response. Patients in this study underwent biopsies at the beginning and end of the 2-week cabozantinib run-in period, enabling assessment of the effect of cabozantinib monotherapy prior to adding PD-1 blockade. After 2 weeks on TKI there was an expansion of peripheral blood effector and memory T cells, and biopsies demonstrated transcriptomic evidence of a more activated immune milieu induced by cabozantinib. After the addition of nivolumab and subsequent resection, the tumors in responders showed a robust immune infiltrate and a preponderance of tertiary lymphoid structures, B cells, and plasma cells when compared with the tumors in patients who failed to achieve a pathologic response.

While the trial with cabozantinib was for patients who were deemed poor resection candidates at presentation, two window-of-opportunity trials for patients deemed resection candidates up-front were recently reported that showed promising pathologic and immunological response to more brief interventions.^14,15 One trial treated patients with two doses of the anti-PD-1 antibody cemiplimab, administered 3 weeks apart, followed by surgery as soon as 22 days following the initiation of immunotherapy; all patients underwent pre-treatment biopsies and regular blood collection, and following surgery patients received an additional 8 cycles of adjuvant therapy.¹⁴ Notably this trial mandated that patients be up-front surgical candidates; though these were not small isolated lesions, 62% were BCLC-A while the remainder had multifocal disease or other high-risk features. The primary endpoint of this trial was significant tumor necrosis (STN), defined as necrosis of 70% or more of the tumor at time of resection. As with the prior trial, therapy was well tolerated, and all patients went to the operating room; only one experienced a 2-week delay due to the emergence of immunotherapy-related pneumonitis prior to initial planned day of resection. The trial enrolled 21 patients, and 20 tumors were successfully resected; 4 (20%) had achieved STN, with 3 of those 4 achieving a pathologic complete response (pCR). Three additional patients had between 50–70% necrosis of the resected tumor, while the remaining 13 patients had 30% or less necrosis, so an exploratory analysis was pursued looking at patients who had more or less than 50% necrosis. Mandatory core-needle biopsies on this trial enabled direct comparison between pre- and post-treatment tissue. Similar to the data from the trial of nivolumab and cabozantinib, a significant immune infiltrate was seen in responders compared with non-responders, and this increased immune infiltrate was also noted at baseline among responders, suggesting that a pre-existing immune infiltrate—so-called “hot” tumors—may be predictive of rapid response to immunotherapy. Also similar to the previous study, an enrichment in B cells and plasma cells was seen among responders, recapitulating findings from many studies in the advanced setting which have seen a correlation between the presence of B cells and response to checkpoint blockade.¹⁵

In this trial the primary endpoint was safety and tolerability; all 27 patients enrolled received 3 pre-operative infusions of nivolumab, administered every 2 weeks, and patients randomized to the combination arm received a single infusion of pre-operative ipilimumab. Following surgery nivolumab was administered every 2 weeks, with those in the combination arm continuing to receive ipilimumab every 6 weeks. Therapy was overall well tolerated, though, as has been seen in the advanced setting, patients receiving the ipilimumab combination had nearly double the rate of serious immunotherapy-related adverse events (irAEs). In this study a major pathologic response (MPR) was also defined as 70% or more necrosis, which was seen in 6 of the 20 patients who underwent resection; 5 of these showed a pCR. Of the 7 patients who did not undergo surgery, the majority were due to disease progression, and none of these cancellations was due to irAEs. This trial had the longest follow-up of the three trials discussed here, and notably, after a median follow-up of 2 years there had been no recurrences seen in patients with MPR, while half of the remaining patients had developed recurrence. This trial similarly found a higher immune infiltrate and increased B cells in MPR patients compared with those without significant pathologic response. Interestingly, similar to the trial with cemiplimab, preexisting immune infiltrate appeared to predict response to nivolumab monotherapy without a significant increase in the number of T cells following therapy, whereas in the nivolumab/ipilimumab combination cohort there was a notable increase in the T and NK cell infiltrate following therapy in the MPR patients that was not seen in non-responders. While this is a small cohort, these data suggest that PD-1 monotherapy may benefit only patients who have pre-existing immune infiltrate, while patients with “cold” tumors may stand to benefit from the more toxic combination with ipilimumab.¹⁵ This is notable given the data supporting the potential of CTLA-4 blockade to induce de novo T cell clones.¹⁹ Beyond PD-1 and CTLA-4 blockade, there are numerous trials looking at other combinations, including combination therapies often used in the metastatic setting, incorporating antibodies or TKIs targeting VEGF, and novel combinations of PD-1 and targets such as LAG-3 which have shown synergistic efficacy in melanoma (Table 1). These trials and others will enable us to better understand the efficacy of these combinations in the advanced setting, where there is a significant need for biomarkers and more optimized combinations given that no combined therapy to date shows response rates over 50%.

Table 1 Neoadjuvant Immunotherapy Trials Underway in HCC

Challenges in the Neoadjuvant Space

These three trials highlight both the potential of immunotherapy to prolong recurrence-free survival, and the limitations in this approach. One pressing issue that must be resolved to make way for regulatory approval of neoadjuvant agents for HCC is the lack of a validated endpoint. Pathologic endpoints have been validated in tumors such as breast and lung cancer where neoadjuvant chemotherapy is standard, but extrapolating response to cytotoxic therapy to immunotherapy is a stretch. Validating pathologic endpoints such as pCR and MPR anew in the age of immunotherapy will be required. In HCC there is no validated endpoint from previous TKI or chemotherapy trials, so it is not surprising that early trials have disparate endpoints. The cemiplimab trial and the nivolumab±ipilimumab trials chose 70% necrosis as cutoffs, extrapolating from a case series in with patients underwent resection following locoregional therapy with chemoembolization.⁵⁸ The cabozantinib trial, conversely, borrowed the commonly used MPR definition of 90% or more necrosis seen in NSCLC trials.⁵⁹ For exploratory purposes the cemiplimab trial noted that 7 patients had 50% or more necrosis while the remaining patients had 30% or less necrosis in the post-treatment tumors, and explored that as a potential cutoff; this 50% cutoff has been used as the definition of MPR in other studies reported in abstract form.⁶⁰ Traditional radiographic responses are sometimes used in the clinical setting to identify when patients receiving neoadjuvant therapy should proceed to the operating room, but when we are considering very brief interventions, where pseudoprogression is a real possibility, the two neoadjuvant papers’ response data highlight how this type of radiographic assessment must be refined.^14,15 The cemiplimab trial had multiple patients who had pathologic responses and evidence on contrast-enhanced MRI of necrosis without significant response as per standard RECIST criteria which rely on tumor size alone. The nivolumab±ipilimumab trial similarly had one patient with frank progression by RECIST on imaging who was found on pathologic assessment to have a pCR with growth of the tumor bed associated with immune infiltrate. Larger trials are warranted to define and validate clinically significant pathologic and/or radiographic responses.

The optimal duration of therapy is also a key question. In the three reports we highlight here, using the respective pathologic endpoints identified, there was an inverse correlation between the time that people received therapy prior to surgery and the rate of pathologic response. Since the aim of neoadjuvant therapy is not so much to kill the tumor as to prime an immune response against micrometastatic disease, a briefer intervention may not only be equally beneficial but also decrease the risk of a pre-operative irAE derailing the scheduled surgery. With PD-1 blockade alone toxicities typically show up after a month or more of therapy, so a brief intervention may mean that should a patient develop a toxicity it would happen after surgery and not risk delaying surgery, while ipilimumab-containing regimens often result in toxicity developing earlier on during treatment. As was the case with one patient in the cemiplimab trial, an early case of pneumonitis immediately prior to surgery delayed resection significantly, and such delays in certain circumstances may lead to patients no longer being surgical candidates. Conversely, in the nivolumab±ipilimumab trial 4 of the 7 patients who did not have surgery were found to have progressive disease on imaging prior to surgery, suggesting that a longer treatment provides a “biological test of time” enabling identification of patients who will rapidly recur, and should therefore be spared the morbidity of a hepatectomy. Some have expressed concern about the use of immunotherapy in HCC patients who may subsequently undergo transplant, as may occur in a patient who is found to have additional small lesions upon reimaging following induction immunotherapy. However, while there is valid concern for complications, the highly immunosuppressive lymphodepleting regimens used in the post-transplant setting will likely abrogate any concern for increased toxicity, and small case series to date have supported the use of immunotherapy in HCC patients who may subsequently undergo transplantation.⁶¹

The use of immunotherapy based on biomarkers and/or underlying etiology of HCC must also be investigated to see if these factors should guide decisions about whether to use immunotherapy in the pre-operative setting, and if so which therapy is best. In other malignancies tumor size, histologic classification, and/or lymph node metastasis guides stage and treatment decision in the neoadjuvant space, but in HCC we know that the vast majority of patients with even the smallest tumors (less than 2 cm) with no radiographic evidence will recur postoperatively, and may therefore benefit from perioperative therapy that may help eradicate micrometastatic disease. While a recent report suggested that in patients with advanced non-alcoholic steatohepatitis (NASH)-related HCC, immunotherapy is significantly less efficacious than in patients with other etiologies, this finding has not been fully supported in these albeit small neoadjuvant trials.^14,62 In the cemiplimab trial there was significant necrosis seen in tumors resected from both viral (HBV and/or HCV-related) and non-viral (alcohol and/or NASH-related) tumors, suggesting that etiology alone does not define the immune response to cancer, though again, larger trials must be performed to support either conclusion.

While standard biomarkers, such as size or underlying etiology may not prove helpful, in-depth analysis of tissue from these trials will aid in identifying biomarkers, such as baseline T cell infiltration, described in the setting of the ipilimumab trial. These findings support the increased use of diagnostic biopsies not only to confirm the radiographic finding of HCC but also to inform development and personalization of therapeutic approach. There are numerous clinical trials now looking at combining these standard immunotherapies with radiation (e.g. NCT03916627), myeloid modifying agents (e.g. NCT04123379) and dozens of other approaches, and given the vast number of agents in development in the immunotherapy space, in-depth understanding of how these agents work and for whom will be key to tailoring therapy to individual patients.

The Unique Potential to Learn in the Neoadjuvant Setting

While the studies discussed here support the idea that, for at least a subset of patients, preoperative therapy likely portends significant clinical benefit, these studies in which pre- and post-treatment biospecimens are available offer a unique setting to investigate the dynamic effects of these therapies. Groups leading similar trials in other fields have already spearheaded revolutionary characterization of tumor-specific immune cells that will inform our development of immunotherapies that target only the minute subset of lymphocytes capable of tumor-recognition while avoiding the toxicity associated with pan-immune activation.^18,52 Through use of spatial transcriptomics and multiplex IHC we will garner further understanding of the immune-tumor-stroma interface which is the target of numerous novel cancer therapeutics. The insights from these trials will benefit development of therapies not only for the minority of patients who are surgical candidates, but also the majority of patients for whom curative intent surgery is not an option.

Conclusion

In conclusion, neoadjuvant immunotherapy is a promising clinical approach for the treatment of resectable HCC, and also offers the opportunity to characterize how novel therapies work in vivo in humans. Trials of adjuvant immunotherapy will read out in the coming year, and this will further help define the optimal perioperative therapy, given that many neoadjuvant trials opt to continue immunotherapy following resection. These perioperative trials will aid us in defining the optimal duration of therapy and biomarkers to guide therapeutic decisions to improve both progression-free and overall survival in HCC patients across stages.