Anti-angiogenesis and Immunotherapy: Novel Paradigms to Envision Tailored Approaches in Renal Cell-Carcinoma

Although decision making strategy based on clinico-histopathological criteria is well established, renal cell carcinoma (RCC) represents a spectrum of biological ecosystems characterized by distinct genetic and molecular alterations, diverse clinical courses and potential specific therapeutic vulnerabilities. Given the plethora of drugs available, the subtype-tailored treatment to RCC subtype holds the potential to improve patient outcome, shrinking treatment-related morbidity and cost. The emerging knowledge of the molecular taxonomy of RCC is evolving, whilst the antiangiogenic and immunotherapy landscape maintains and reinforces their potential. Although several prognostic factors of survival in patients with RCC have been described, no reliable predictive biomarkers of treatment individual sensitivity or resistance have been identified. In this review, we summarize the available evidence able to prompt more precise and individualized patient selection in well-designed clinical trials, covering the unmet need of medical choices in the era of next-generation anti-angiogenesis and immunotherapy.


Introduction
Angiogenesis inhibition remains one of the most active approaches in the treatment of advanced kidney tumors. Although tumor heterogeneity can be a therapeutic obstacle [1] angiogenesis-related mechanisms represent a truncal event in renal cell carcinoma (RCC) biology, particularly in clear-cell histotype. Indeed, the alterations of the HIF/VEGF axis are deemed to be the fundamental target [2], even aiming at overcoming drug resistance [3]. This evidence explains the clinical success of sequential  [13,[40][41][42]. The color code represents the presence of a given prognostic factors for each model: PS, Hb, LDH, corrected Ca, time from diagnosis to treatment (red) for MSKCC model; PS, Hb, corrected Ca, time from diagnosis to treatment, NE, PLT (green) for IMDC model. Subsequently, Heng et al., in a series of 645 patients with advanced renal cell carcinoma, identified six prognostic factors significantly related to a worse prognosis (IMDC, or Heng model). This system derives from a retrospective analysis conducted on patients with metastatic renal cell carcinoma treated with sunitinib, sorafenib or bevacizumab + interferon alfa-2a. Patients who had received a first line of treatment with cytokines and VEGF/VEGFR inhibitors as second-line treatment were also included in the analysis. Six prognostic factors have been identified: Karnofsky PS, low hemoglobin level, high corrected serum calcium, period from diagnosis to treatment < 1-year, high absolute neutrophil count, and high platelet count. Subjects were divided into different subgroups according to clinical risk: favorable (n = 157), for whom the median overall survival (OS) was 43.2 months and the 2-year OS was 75%; intermediate (n = 440), characterized by a median OS was 22.5 months and the 2-year OS was 53%; poor risk (n = 252) in which the median OS was 7.8 months and the 2-year OS was 7% [13,42] ( Figure 1).

From the Cytogenetics to the Mutational Landscape of RCC
Despite the considerable efforts made to stratify patients from a prognostic standpoint by using clinical criteria, efficient prognosticators for characterization represent an unmet medical need, especially when considering the plethora of new immunomodulatory and anti-angiogenic drugs available to date. Cytogenetics pioneered the molecular investigation of patient stratification, based on Xp11.2 translocation and deletion or chromosomal aberration on 3p and 14 in RCC-impacted clinical outcomes [43][44][45]; the incidence of Xp11.2 translocation is very low, but should be searched for systematically in young patients [46]. Chromosome 3 harbors several putative oncogenes and oncosuppressors, the biological relevance of which is highlighted by von Hippel-Lindau(VHL)/HIF-1α axis, PBRM1, BAP1, SETD2 prognostic role [16,45,[47][48][49]. Furthermore, numerous chromosome alterations in terms of chromosome gain or loss (i.e., gain of 7q, loss of 9p, 9q and 14q) have been highlighted and associated with worse survival (p < 0.001), with a prognostic but not predictive role [50].
Next, several novel biomarkers are currently being evaluated to assess the prognostic and predictive value for different response of renal malignancies treated with antiangiogenic-TKI and immunotherapy. Molecular markers can be classified according to their physiological location into tissue and soluble factors [51]. Among the above-mentioned traditional histological features, carbonic anhydrase IX (CaIX) [52], CXCR4 [53,54], HIF-1α and HIF-2α [55] have been reported to predict response to sorafenib or sunitinib as well as improved progression-free survival (PFS), despite no consistent impact on OS being reported. Specifically, tumor shrinkage gained by sorafenib treatment significantly differed between CaIX high vs. CaIX low cases (−13% vs. +9%) [52]. Moreover, D'alterio et al. and Guo et al. independently revealed CXCR4 expression level to be significantly correlated to sunitinib response and improved PFS in patients treated with sorafenib, respectively (median PFS 20 vs. 6 months, in CXCR4 low/high , p = 0.038) [53,54]. Furthermore, patients' stratification-according to HIF-1α expression level-was also able to predict improved PFS in the HIF-1α high over the HIF-1α low sunitinib-treated-subgroup (42.0 weeks vs. 30.4, respectively, p = 0.034) [56].
The assessment of the soluble factors evaluation has also been extensively investigated in the prognostic stratification attempts, uncovering VEGF/VEGFR, LDH, IL-6, IL-8, osteopontin (OPN), HGF and TIMP1 to be significant drivers of a patient's prognosis and response to therapy [62][63][64][65]. High serum VEGF levels reflected an aggressive tumor biology and kept an independent prognostic value in a multivariate analyses including MSKCC score and ECOG PS, while being able to predict a better clinical outcome over the unstratified population (p = 0.015) [66]. Low baseline levels of sVEGFR3 and VEGF-C were also predictive of improved PFS upon sunitinib treatment. (median PSF 36.7 weeks and 19.4 weeks in sVEGFR 3 low/high , respectively; moreover, the median PSF was 46.1 weeks and 21.9 weeks in VEGF-C low/high patients, respectively [62]. Next, IL-6, osteopontin, and TIMP-1 were integrated in a prognostic model including selected clinical variables and showing higher prognostic accuracy than IMDC model (concordance-index 0.75 vs. 0.67, respectively) [65]. Ancillary, emerging evidences uncovered nucleotide polymorphisms (SNPs) of IL-8, HIF-1α and VEGF axes to significantly impact the therapeutic outcome in RCC [67,68] as in several TKI sensitive tumors [69][70][71]; however, no validation has been achieved in statistically powered clinical studies [55,72].
A recent report highlighted the emerging role played by non-coding RNA, such as miRNA in RCC; in the frame of this thinking, clinically and prognostically relevant RCC subgroups were reflected by distinctive miR expression levels [73][74][75][76]. For example, Heinzelmann and colleagues identified a signature, including miR-451, miR-221 and miR-26a, which separated between metastatic and non-metastatic clear cell RCC [77]. Functionally, miRs orchestrate crucial steps in immunosurveillance and modulate cancer immune checkpoints by influencing cells of the immune system and tumor cells [78,79]. In RCC, miRs were shown to regulate HLA-G [80] and PD-L1 expression [81]. Additionally, there is a growing body of literature highlighting the prominent role of miRs in angiogenesis-related signaling [76,[82][83][84]. For instance, miR-195 and miR-221 regulate the expression of VEGFR2 in various tissues, including RCC [85][86][87][88][89]. Accordingly, miR-221 expression was part of signatures predicting the response of RCC patients towards TKI/anti-angiogenic therapy in two independent studies [87,90].
Evidence from tissue and circulating pro-angiogenic factors matches with familial VHL syndrome disease-phenotype: hypoxia-inducible factors overactivation constitutes a fundamental proof of principle in hereditary clear-cell RCC (ccRCC), but also elicited comprehensive genomic characterization of sporadic tumors, by focusing on pro-angiogenic mechanisms. In ccRCC, decreased VHL activity correlated with enhanced HIF-1α expression, as well as with the consequent hyperactivation of VEGF, PDGF, TGF-α, thus leading to increased PI3-K/PKB/mTOR signaling, and tumor progression [91][92][93]. Undoubtably, the biological knowledge related to VHL pathway-driven investigation inspired novel therapeutic windows [94,95]. However, several data and meta-analyses revealed that VHL gene alteration holds neither prognostic, nor predictive value in subjects suffering from ccRCC [91,96].
The dismal impact on clinical outcome exerted by VHL per se can likely also be explained by the complex genomic architecture driving the malignant phenotype of RCC. Indeed, several additional genetic alterations were also frequent in ccRCC, such as somatic mutation of chromatin remodeling genes including PBRM1, SETD2 and BAP1 (38%, 13% and 11% of cases, respectively), mutation of PI3K-AKT-mTOR pathway genes (occurring in 16% of patients) comprising PTEN, MTOR and PIK3CA, loss of CDKN2A, and mutation of TP53 (16.2% and 2.6%, of subjects, respectively) [16]. CDKN2A loss, BAP-1 and TP53 mutation are associated with poorer survival in ccRCC. The poor prognostic role of CDKN2A loss has also been confirmed in papillary and chromophobe RCC histological subtypes [16]. Conversely, PBRM1 loss-of-function mutations correlated with less aggressive behavior and with better PFS and OS in advanced patients [97][98][99]. In a retrospective study, Kapur et al. revealed PBRM1 to be significantly predictive for improved median OS (10.6 vs. 4.6 years) when compared to BAP1 mutational status. Consistently, data from TCGA confirmed the UTSW cohort by showing median OS of 5.4 and 1.9 years for PBRM1 vs. BAP1 mutated cases, respectively [98]. Next, genomic annotation-model based uncovered the independent prognostic value harbored by any TP53, BAP1 and PBRM1 mutation to be relevant in improving the MSKCC model in patients treated with first-line TKI [100]. Likewise, the IMmotion150 trial, which compared, in a three-arm fashion, sunitinib over atezolizumab monotherapy and atezolizumab plus bevacizumab in treatment-naive RCC, revealed PBRM1 mutations to be correlated with improved survival in the sunitinib arm. Additionally, the ICI response prediction to anti-PD1 identified by PBRM1 mutational status apparently parallels the behavior reported in TKI-treated patients [97], warranting further statistically powered trials aimed to clarify the predictive value of PBRM1 [25].

Molecular Classification
Gene expression profile parallels genetic and genomic alterations and impacts the clinical outcome. The mRNA expression patterns differ among major histological subtypes as well as among each RCC subtype. Proteomics-based subtyping of ccRCC, either according to Brannon [102], or KIRC analysis (four clusters m1-m4) [103] consistently deconvolute the biologic taxonomy of disease phenotype. Moreover, the combination of the singular subtypes can dissect three different clinical behaviors: (1) good prognosis group (cluster ccA, CC-e.2, and m1), involved chromatin modifier genes mutations, such as PBRM1; (2) poor prognosis group (cluster ccB, CC-e.3, m3), associated with higher expression of CDKN2A and hypoxia-related genes, chromatin remodeling genes mutation including SETD2 or BAP1, PI3K/AKT/mTOR pathway genes mutations, epithelial-mesenchymal transition, hypermethylation, and a metabolic shift with higher glutathione and dipeptide levels; (3) intermediate prognosis group (cluster 3, CC-e.1, m2, and m4) associated with BAP1 mutations and base-excision repair [55]. Additionally, data obtained from 942 surgical series pinpoint a molecular signature consisting of 16 genes that could predict post-surgery relapse and could be translated into clinical trials [104]. Unsupervised hierarchical cluster analysis identified different biological pathways, including vascular, cell growth or division, immune response, and inflammation phenotypes. In line with previous data, vascular and immune response phenotypes were associated with a better outcome, whereas higher expression of proliferation and differentiation genes and markers associated with inflammatory responses were associated with a worse outcome [104]. Overall, it is worth highlighting that all the above-mentioned data were generated by analyzing prognostic implications obtained from non-metastatic settings. Conversely, Beuselinck et al. performed a multi-omics analysis and identified four molecular tumor subtypes able to predict clinical outcome and response to sunitinib in metastatic ccRCC: ccrcc1 ("c-myc-up") and ccrcc4 ("c-myc-up and immune-up") characterized by the upregulation of MYC targets and shorter PFS, OS and poorer response to sunitinib; ccrcc2 ("classical") and ccrcc3 ("normal-like") with a higher expression of the pro-angiogenic HIF-VEGF-VEGFR-pathway, longer OS and better TKI response. Characteristically, the ccrcc4 subtype had a strong inflammation, BAP1 mutation, sarcomatoid dedifferentiation and decreased angiogenesis dependency, and significantly poor survival and response to sunitinib and pazopanib [105,106]. The four molecular subtypes could explain the different outcome in the IMDC risk group. The IMDC good risk group was enriched for the ccrcc2 subtype and higher angioscore; conversely, the IMDC poor risk group was enriched for the ccrcc4 subtype and lower angioscore. Nevertheless, no correlation was found in the immune score across IMDC risk groups [107]. Given that the existence of an angiogenesis-addicted, and immune-inflamed phenotypes seems to correlate with the presence of peculiar genomic signatures [108], it is tempting to speculate an ancillary role played by specific mutated genes. Remarkably, PBRM1 mutational status and boosted angiogenesis in ccrcc2-3 seem to have more interactions among themselves than would be expected for a random set of molecular interactions [26,108]. The different clinical outcome obtained in sunitinib-treated patients compared to avelumab alone or in combination with bevacizumab remains to be fully elucidated, and might be explained by an underlying angiogenesis-driven mechanism in this subgroup over PBRM1 wild type phenotype [25].
In a comprehensive interrogation of available datasets carried out by Hakimi et al., four clusters were also identified, shedding more light on the peculiar features of the tumor microenvironment (TME) and substantially extending the insights regarding the role of angiogenesis signatures in predicting TKI response. Specifically, this analysis highlighted the role of macrophages fingerprint within the TME and uncovered a putative angiogenesis high macrophages low signature to be one fundamental determinant predicting prognosis and, likely, impacting response to TKI [26]. This piece of evidence might support clinical decision while selecting approaches based on mono-vs. combination-therapy and anti-angiogenesis vs. ICI-inhibitors based approaches, also pinpointing the unexplored efficacy of CSFR1-targeting [26]. These data need to be confirmed in appropriately designed studies to be translated into clinical practice.

Therapeutic Window Driven by Angiogenesis and the Immune System Targeting Current Challenges
The treatment scenario of mRCC has largely evolved in recent years, translating into an outcome improvement achieved by targeting VEGF/VEGFR pathways (bevacizumab, sorafenib, sunitinib, pazopanib, axitinib, cabozantinib and lenvatinib) [126][127][128][129][130][131][132], mTOR signaling (everolimus and temsirolimus) [133,134] and immunocheckpoint inhibitors comprising anti PD1/PD-L1 (nivolumab, pembrolizumab, avelumab, atezolizumab) [11,[59][60][61] and anti-CTLA4 (ipilimumab) [135] alone or in combination therapies ( Figure 2). Due to the dynamic plethora of therapeutic options available to date, it is critical to identify criteria driving personalized approaches. Indeed, real-life clinical practice faces the significant challenge of patient selection by tailoring a TKI-vs. ICI-based and mono-vs. combination-therapeutic strategy [55,136]. Currently, besides the obvious impact of clinical  Due to the dynamic plethora of therapeutic options available to date, it is critical to identify criteria driving personalized approaches. Indeed, real-life clinical practice faces the significant challenge of patient selection by tailoring a TKI-vs. ICI-based and mono-vs. combination-therapeutic strategy [55,135]. Currently, besides the obvious impact of clinical individual risk profiling and comorbidities potentially influencing the safety issues, the single decision-making tool is represented by the risk stratification considered by the regulatory agencies.
Intermediate and high-risk might warrant a cabozantinib-containing regimen according to the CABOSUN study [4] that evaluated only this setting of patients. Cabozantinib, as a small molecule halting several tyrosine kinase receptors such as VEGFR-2, MET and AXL, as well as other potentially relevant kinases including RET, KIT, and FLT3, has been evaluated in the CABOSUN phase II multicenter study. In total, 157 subjects with intermediate/high risk stratified by Heng profiling were randomized to receive cabozantinib or sunitinib [4,136,137]. The CABOSUN trial met the primary end-point, showing improved PFS in the experimental arm (median PFS 8.2 months vs. 5.6 months with cabozantinib over sunitinib, respectively, supporting cabozantinib as a potential first-line treatment option for patients with advanced ccRCC of intermediate or poor risk [138]. Cabozantinib has been uncovered to also be effective in metastatic non-clear RCC in a retrospective cohort study investigating naïve and refractory cancers and all IMDC model risk groups. The median PFS was 7.0 months, and median OS was 12.0 months [139]. The combination study of nivolumab with ipilimumab (CheckMate-214 study) including all-comers showed an ICI benefit in the intermediate/high-risk population only, apparently with a detrimental effect in low-risk patients, where sunitinib conferred an improved clinical outcome [9,134]. The phase 3 trial included 847 patients with untreated advanced RCC who were randomly assigned to receive either nivolumab in combination with ipilimumab, or sunitinib. In the latest update, presented at 2020 Genitourinary Cancers Symposium at median 42 months of follow-up, the combination immunotherapy continued to be associated with improved OS and PFS compared to sunitinib arm (median 47.0 vs. 26.6, and 12 vs. 8.3 months, respectively, and 42-month rates of 52% versus 39%, and 35% versus 19%, respectively). PFS curves plateaued after 30 months at around 35% with nivolumab plus ipilimumab. An exploratory efficacy analysis restricted to the 249 favorable-risk participants established sunitinib to be more active when compared to ICI in this patient subset, gaining a median PFS of 27.7 vs. 17.8 months and ORRs of 54% vs. 29% [140].
However, the clinical and pathological features not entirely mirroring the complex biology of the tumor should be adapted to the novel agent's era. Specifically, Heng criteria [13] and prognostic factors were developed to inform patients about their prognosis and in order to compare the results of different trials [141]. Conversely, such stratification tools are not expected to perform efficiently in therapeutic strategy selection. The Checkmate 214 study represented a paradigm shift, with the potential to picture and weigh the single prognostic factors quantity over the global additive effect on the clinical outcome [9,134]. Moreover, the platelet count and the calcium levels had a more significant impact than was usually observed [142].
The recently published data regarding the combination of anti-angiogenic and anti-PD1 treatment (axitinib in combination with pembrolizumab [59] or avelumab [60]) compared to sunitinib demonstrated a benefit from the combination across the population, regardless of risk class and PD-L1 expression.
In the phase III KEYNOTE-426 study, the majority of patients displayed intermediate or poor risk disease as assessed by IMDC criteria and sarcomatoid features in 18% of the patients. PFS was 15.1 and 11.1 months in the pembrolizumab/axitinib and in the sunitinib group, respectively. Pembrolizumab plus axitinib demonstrated effectiveness and good safety for patients with clear cell mRCC, with an impressive 59% objective response rate.
The Javelin renal 101 study dichotomized patients into two classes-PD-L1 positive/negative -choosing immunohistochemistry expression as class boundary and by declaring as co-primary outcome OS and PFS assessment in PD-L1 positive . Avelumab/axitinib performed better than sunitinib in terms of both PFS and ORR, regardless of PD-L1 expression [60]. Conversely, a trend of enhanced efficacy within the PD-L1 positive subgroup was observed in the atezolizumab plus bevacizumab arm compared to sunitinib in the IMmotion151 trial (median PFS 11.2 vs. 7.7 months, respectively; p = 0.0217) [61]. Preliminary results of a new combination of TKI (cabozantinib) plus anti-PD1 (nivolumab) promise a clinically meaningful benefit (NCT03141177) and warrants further investigation regarding the chance of anti-angiogenic strategies combined to ICI. Additional information aiming to clarify whether this approach might benefit as pure synergic strategy or by intercepting a broader disease spectrum irrespective of patients' selection remains as-of-yet unknown.
Collectively, evidence generated by the above-mentioned trials uncovered neither the risk class nor the PD-L1 expression as being efficient in predicting the response to therapy. Thus, several omics attempts retrospectively analyzed the available data. Nonetheless, as a first in class prospective study, the IMmotion150 phase II emphasized the translational role of TME deconvolution at the transcriptomic level, suggesting that the outcome prediction with anti-angiogenic drugs and ICI is applicable upfront in mRCC [25]. In detail, a gene expression profiling fingerprint has been proposed according to different phenotypes, clustered using expression ranks boundaries of pro-angiogenic, pre-existing immuneand myeloid tolerogenic-associated molecular subgroups [25]. Consequently, as expected, angiogenic blocking by sunitinib was highly active in Angiogenesis High patients, whilst atezolizumab alone seems to halt tumor activity in immunogenic cancers and dismal myeloid inflammation (Teff High Myeloid Low ). Regarding the combination of ICI plus sunitinib, although the authors comprehensively demonstrated a direct impact of immune-and inflamed-infiltration (Teff High Myeloid High ) [25], it is still debated whether combining anti-angiogenic and immunological checkpoint inhibitors without proper selection, more than what would be necessary, constitutes a synergistic strategy [143]. Nonetheless, robust and compelling preclinical [28,143] and clinical [26] evidence supports the biological ecosystem dissection as the future driver of patient selection for choosing candidates among ICI/anti-angiogenic strategies: different biological RCC behaviors pinpoints the tight correlation existing by intermediate/high risk profile, tumor angiogenesis and indirect immune-tolerogenic milieu. The roadblocks in standardizing biomarkers in clinics are due to the lack of data able to deconvolute RCC biological characteristics derived from prospective studies. Moreover, additional caveats restraining the real-life translation of the biological RCC taxonomy are constituted by the patient population heterogeneity and by the absence of a clinical stratification model accounting for next-generation immune-targeted therapy. Statistically powered clinical studies are expected to be carried out, aimed at further validating the promising pioneering results [144]. State-of-the art molecular dissection of RCC subtypes should guide clinical trials' designs, in order to efficiently tailor the best therapeutic option upfront. An Achilles' heel of the modern approach proposed might be the applicability of changing clinical tools; however, the rational and efficient use of the novel agents available would also prevent the inevitable financial toxicity of the integrated stepwise RCC management. These data will be more than a determinant in a dynamic evolving sequential treatment strategy, thereby deeply impacting further therapy. Details from the most recent clinical trials are summarized in Table 1.

Conclusions
We are currently entering the third era of mRCC therapy with the challenging aim of combining immune-immune and immune-VEGFR-TKI, which is a direct portrait of the peculiar underlying pathophysiology of disease, being dependent on angiogenesis and the close connection between cancer cells and the immune system. The lack of direct comparisons, as well as different study designs and patient stratification, considered as major limits could also represent a caveat in order to better tailor clinical decisions. Nonetheless, though prognostication is mandatory, biological correlates are highly needed. Using immunotherapy, it is mandatory to design clinical trials with a robust immunological background.