Synergism between anti-angiogenic and immune checkpoint inhibitor drugs: A hypothesis

Abstract

Hepatocellular cancer (HCC) and renal cell cancer (RCC) are singularly resistant to conventional chemotherapy drugs but therapies targeting the supporting stroma have significantly altered their management. Two recent trials combining anti-angiogenic (AA) agents with immune checkpoint inhibitors (ICIs)- the IMbrave150 and IMmotion151 – have reported impressive progress over targeted agents. It has been suggested that bevacizumab, by improving tissue perfusion, changes the immune suppressive tumour microenvironment to an immune stimulatory one where the ICIs can be more effective. This hypothesis proposes an alternative explanation: That bevacizumab, by increasing tissue hypoxia, amplifies the mutational burden of the tumour by stress-induced mutagenesis, creating a hypermutator profile, which is more vulnerable to the ICI drug, atezolizumab. Additionally, ICIs are known to cause hyperprogression in some tumours, and bevacizumab could provide further benefit by starving these rapidly proliferative tumours of blood supply and nutrients.

Introduction

The combination of bevacizumab (an anti-angiogenic drug) with atezolizumab (an immune checkpoint inhibitor) scored major success in two trials, IMbrave150 and IMmotion151, against hepatocellular (HCC) [1] and renal cell (RCC) cancers [2] respectively. These successes are particularly noteworthy as liver and kidney cancers are particularly resistant to conventional chemotherapy. Classical cytotoxic drugs work by interfering with the machinery in the nucleus that controls cell division. As organs specialized for waste disposal, the liver and kidney are superbly evolved to deal with toxins, including chemotherapy drugs. Cancer cells derived from them inherit similar properties. These drugs are thus eliminated efficiently before they can cause major damage. Oncologists have acknowledged the impenetrable nature of HCC and RCC to conventional chemotherapy after multiple failed trials. Their focus then shifted to targeting the cells in the surrounding stroma (such as vascular and immune cells) – and have achieved limited success. Anti-angiogenic (AA) drugs that cut off blood supply improve survival by a few months. These are either monoclonal antibodies such as bevacizumab, aflibercept and ramucirumab, or a multitude of small molecule tyrosine kinase inhibitors (TKIs) such as sorafenib, sunitinib, lenvatinib, axitinib, pazopanib and cabozantinib. (It should be noted that most TKIs are not pure AA drugs but are multikinases, targeting other receptors as well, such as RAF and PDGFR). Further success followed by use of immunotherapy drugs that stimulate host immunity against cancer cells, such as immune checkpoint inhibitors (ICIs) like pembrolizumab, nivolumab and atezolizumab. These work indirectly by activating the host cytotoxic T lymphocytes against the cancer cells and are thus pro-drugs. Recent trials suggest that the combination of an AA drug (bevacizumab) with an ICI (atezolizumab) is much more than the sum of its parts (Table 1). Interestingly, the combination also appears to work well in another hard-to-treat cancer – liver metastases (such as from lung cancer [15]).

There are AA TKI plus ICI combinations approved for treatment in RCC [16], but as mentioned earlier, TKIs have broad-spectrum activities and I shall mainly concentrate on bevacizumab as the prototype AA drug.

The rationale behind the combination, as pointed out in the trial publications [1], [2], is that the bevacizumab reverses VEGF-mediated immunosuppression, and by normalizing tumour vasculature and improving perfusion, can convert the immunosuppressive tumour micro-environment (TME) to an immune stimulatory one. Three key mechanisms related to VEGF–mediated immunosuppression are inhibition of dendritic cell maturation, reduction of T-cell tumor infiltration, and promotion of inhibitory cells in the TME [17]. Bevacizumab, by virtue of being an anti-VEGF antibody, is believed to reverse these mechanisms.

The conventional explanation of the mechanism of action of bevacizumab is that it shuts off blood supply to the tumours and causes cell death by hypoxia. However, alternative theories have been proposed. Jain [18] hypothesised that under specific circumstances, bevacizumab can do the opposite, that is, “normalise” tumour vasculature, improve perfusion and reduce interstitial pressure, resulting in better drug delivery to the tumour. Normalizing circulation converts the immune suppressive TME into an immune stimulatory one (see below).

Incidentally, it is known that bevacizumab has limited single agent activity and its efficacy improves when given in combination with chemotherapy drugs. Blagosklonny [19] has suggested that the chemotherapy drugs have anti-angiogenic properties, that resistance to these drugs occurs by elevation of VEGF, and bevacizumab works by blocking this increase. Bocci [20], however, accepts the conventional view point but suggests that accompanying chemotherapy drugs may actually contribute to anti-angiogenesis resistance.

Jain [18] acknowledges that normalizing vasculature is both dose and schedule dependant, and in other situations, bevacizumab leads to pruning of vasculature and tumour hypoxia, causing cell death. When the tumour vasculature is excessively pruned, the resultant hypoxia can aggravate an already immunosuppressive TME. This is characterized by increase in immunosuppressive regulatory T cells (Tregs), M2 tumour associated macrophages (TAM), and myeloid-derived suppressor cells (MDSCs), and reduced anti-tumour cytotoxic T lymphocytes (CTLs), M1 TAMs and mature dendritic cells (DCs). Hypoxia can also upregulate multiple immune-suppressive growth factors and cytokines (e.g., VEGF and TGF-β). Intratumoral hypoxia also compromises antigen presenting cell (APC) functionality and the stimulation of T cell responses. (Summarized in [21]). Despite the attractiveness of Jain’s hypothesis, augmented tumour hypoxia and immune suppression appears to be the more likely outcome of bevacizumab activity.

Hypothesis:

It is hypothesized that bevacizumab increases the efficacy of ICIs by

1. Increasing immunogenicity of the tumours by aggravating hypoxia; this causes stress-induced mutagenesis and increased mutation load.

2. Starving rapidly proliferative cancer cells during the hyperprogression precipitated by ICIs.

Generation of hypermutator profile by stress-induced mutagenesis (SIM) caused by hypoxia:

ICIs are more effective against “hot” immunogenic tumours, ie, tumours displaying an excess of “neoantigens”. What defines an immunogenic tumour is still a matter of investigation, but these are generally identified by predictive biomarkers such as percentage of cells positive for PD-L1 (or its variants like tumour proportion score (TPS) or combined positive score (CPS)), by microsatellite instability (MSI), or by the Tumour Mutational Burden (TMB). TMB [22] (measured as mutations per megabase (mutations/Mb) by next generation sequencing (NGS) or by actual number of mutations by whole exome sequencing (WES)) is a biomarker that indicates the number of tumor-specific somatic mutations in the tumor’s exome. High TMB tumors are “hot” tumours with more surface neoantigens, producing an enhanced immune response [23]. The KEYNOTE-158 trial showed that pembrolizumab, an ICI, is effective against tumours with high TMB irrespective of tissue of origin [24], and has been FDA-approved in this indication.

It is possible that bevacizumab can convert “cold” cancers into an immunogenic (“hot”) cancer. ICIs work by blocking the PD-1/PD-L1 interaction, and increasing PD-L1 level is one possibility. This has been shown in ovarian cancer, where bevacizumab-responsive cisplatin resistant ovarian cancer had higher levels of PD-L1 than the non-responsive tumour (58.7% vs. 39.3%) [25]. Similarly, sorafenib (an AA TKI) induced hypoxia with activation of the SDF1-α/CXCR4 pathway, and increased expression of PD-L1 in the cancer cells [26]. In addition, hypoxia has been shown to directly up-regulate – via HIF1α activation – the expression of PD-L1 by MDSCs, DCs and cancer cells to aid immune-suppression and evasion [27].

However, PDL1 is not used as a predictor for immunotherapy responses in HCC and RCC. The other predictor, TMB, is relatively low in RCC and does not seem to be predictive of survival, as shown in the MSKCC IMPACT cohort [28]. Data of HCC was not reported in this cohort, but in another study of 755 patients with advanced HCC [29], median TMB was low at 4 mutations/Mb. Only 6 tumors (0.8%) were TMB-high (conventionally accepted as >10 mutations/Mb).

Both HCC and RCC are highly vascular tumours. HCC’s vascularity relates to upregulation of HIF1a and multiple angiogenic factors such as VEGF, BMP4, PAI-1 and SCF [30]. The predominant clear cell subtype of RCC (ccRCC) arises from the inactivation of vHL gene with upregulation of HIF-1a; the resultant “pseudo-hypoxic” signal causes increased vascularity [31].

Paradoxically, despite their vascularity, HCC and RCC are highly hypoxic. Most mammalian tissues are maintained at 2–9% oxygen (O2) (on average 5.5%, 40 mm Hg). Tissue hypoxia is defined as reduced oxygen partial pressure ≤ 2% (≤15 mm Hg) [32]. The critical pO2 in tumours is proposed to be 8–10 mmHg [33]. Below this level, detrimental changes associated with reduced oxygen consumption have been observed (ATP depletion, intracellular acidosis, glycolysis and apoptosis). HCC is one of the most hypoxic tumors with median oxygen level as low as 0.8% [34]. Again, although ccRCC is a vascular tumour, mean pO2 value was 9.6 mmHg [35].

Despite Jain’s intriguing concept [18], there are several lines of evidence that AA drugs do, as predicted, inhibit angiogenesis and induce hypoxia. For instance, sustained sorafenib treatment has been shown to reduce microvessel density and promote intratumoral hypoxia; the resultant HIF-mediated cellular responses favor the selection of resistant cells adapted to the hypoxic microenvironment and leads to sorafenib resistance [36].

Similarly, multiple studies have documented that bevacizumab primarily induce hypoxic stress. Treatment of mice bearing colon cancer xenografts with bevacizumab resulted in depletion of tumor microvasculature, upregulation of HIF-1α, and increased pimonidazole staining, consistent with an anti-angiogenic effect and induction of hypoxia [37]. Another study with colon cancer xenografts showed that intratumoral hypoxia induced by bevacizumab treatment resulted in activation of HIF-1α protein and upregulated stanniocalcin 2 (STC2) [38]. AA treatment (sunitinib and bevacizumab) of breast cancer xenografts induce hypoxic conditions and increase in cancer stem cells, primarily mediated by HIF-1α (EMT induction) [39]. Other studies have consistently demonstrated reductions in microvessel density in HCC [40] and RCC [41].

Aggravation of hypoxia alters expression of multiple genes, both random and specific (the latter for the upregulation of adaptive resistance pathways) and also results in increased tumor cell invasiveness and selection of more invasive metastatic clones [42], [43], [44]. The increased aggressiveness post AA therapy is a possible explanation for the lack overall survival (OS) benefit despite the increased progression free survival (PFS) seen with AA agents [43]. For instance, two randomized clinical trials with bevacizumab and taxanes for the treatment of metastatic breast cancer (the E2100 study [45] and the AVADO [46] study) demonstrated small benefits in PFS but not an OS benefit. Similar induction of aggressive behavior has been noted in glioblastoma [47] and colorectal cancer [48].

Uedo et al. [49] demonstrated that in metastatic breast cancer treated with a combination of paclitaxel and bevacizumab, non-responders exhibited more severe hypoxia than responders (despite higher angiogenesis). The authors confirm that addition of bevacizumab results in acute hypoxia, and increased cytokine secretion associated with cancer progression, but the exact genetic mechanism was not studied.

The paradoxical effects of enhanced tumour aggressiveness and induction of an invasive phenotype with the use of bevacizumab have been reviewed [50]; the basis appears to be upregulation of resistance mechanisms by additional mutations; specific causes cited include increased epithelial-to-mesenchymal transition, and activation of MET receptor. The significance of these studies is that cancer cells do not “roll over and die” when stressed with hypoxia, but survive by acquiring additional mutations. The additional mutations contribute to increased aggressiveness.

As Uedo et al. [49] notes, the behavior of bevacizumab is a double-edged sword; it induces reoxygenation if there is proper vascular remodeling, otherwise it can lead to severe hypoxia because of destruction of the tumor vasculature.

In summary, it would appear that use of AA drugs leads to increased hypoxia, and this hypoxic challenge can result in additional tumour cell mutations and evolution of an aggressive phenotype.

Not all mutations are random. When bacteria and yeast are exposed to stress (such as by hypoxia or nutrient deprivation), they respond by actively inducing survival mutations. This is by a process called adaptive or stress-induced mutagenesis (SIM). It is now known that cancer cells when stressed, like bacteria, can also undergo SIM. Hypoxic stress activates adaptive mutagenesis, partly by downregulation of mismatch repair proteins; it induces genomic instability and a hypermutator profile (reviewed by Fitzgerald et al. [51]); optimum conditions for checkpoint inhibitors to be effective.

It is hypothesized that hypoxic stress caused by AA therapy leads to SIM and creates a hypermutator profile, and the induction of neoantigens and a higher mutation burden is a major factor in the improved efficacy of ICIs in combination with AA drugs.

There is limited data on mutational changes after exposure to AA therapy. Metastatic colorectal cancer monitored by liquid biopsy while on treatment with bevacizumab (and chemotherapy), developed mutations of CREBBP and FBXW7; these genes are associated with tolerance to hypoxia [52]. A report from the ICGC/TCGA Pan-Cancer Analysis of Whole Genomes (PCAWG) Consortium quantified hypoxia in 1188 tumours spanning 27 cancer types. It showed that elevated hypoxia is associated with increased mutational load across cancer types, irrespective of underlying mutational class [53].

A recent study [54] on cell lines showed that hypoxic stress resulted in transcriptional downregulation of DNA repair genes. Hypoxic conditions also increased the mutational burden which was characterized by an increase in frameshift insertions and deletions, and increase in the formation of potential neoantigens.

The downregulation of DNA repair genes on exposure to hypoxia have been reported earlier [55], [56]; this helps in the propagation of the new mutations. Hypoxic cells can thus acquire a mutator phenotype that consists of decreased DNA repair, increased mutation rate, and increased chromosomal instability [57].

Conforming to the same principle, gliomas treated with temozolomide (an alkylating agent) develop hypermutator profile [58]; similar changes induced in a case of neuroendocrine carcinoma was exploited to treat the patient successfully with an ICI, pembrolizumab [59].

In summary, it is hypothesized that use of bevacizumab increases hypoxic stress in tumours, ushers in stress-induced mutagenesis, which increases the mutational burden of the tumour, and converts a “cold” non-immunogenic tumour into a “hot” tumour, thus increasing the efficacy of ICIs.