Predictors of Response to Intravesical Therapy

Despite the 40-year reign of bacillus Calmette-Guérin (BCG) as the most effective immunotherapy in urologic cancers, a lack of clinical tools to predict treatment response has hampered progress in the field. Acting as an immunostimulatory agent against a multitude of phenotypically diverse non–muscle-invasive bladder cancers, response to BCG likely depends on both tumor characteristics as well as host factors. With a deeper understanding of the tumor biology as well as the mechanism of action underpinning immunotherapy, newer and more effective clinical tools are being constructed to improve patient selection.

Key points

•

Tumor characteristics such as grade, stage, concomitant carcinoma in situ, and other molecular markers are likely prognostic of oncologic outcomes, rather than predictive of response to intravesical bacillus Calmette-Guérin (BCG).
•

Host factors such as age, gender, genetic polymorphisms, and host microbiome may play crucial a role in the response to intravesical BCG.
•

Additional predictors of response can be deduced from further elucidation of the mechanism of immunostimulation by intravesical BCG.

Introduction

It has been more than 4 decades since Morales and colleagues first introduced the use of bacillus Calmette-Guérin (BCG) for non–muscle-invasive bladder cancer (NMIBC). Despite its reign as the most effective immunotherapy to date in urologic cancers, recurrence rates range between 32.6% and 42.1% and progression rates between 9.5% and 13.4%. ^, Progression clearly leads to compromised survival ; recurrent, nonprogressive disease is also problematic because of repeated procedures and intensive surveillance protocol. The lack of clinical tools to accurately predict response to therapy has hampered progress. Recent advances in the understanding of bladder cancer biology have provided a pathway to identifying early markers of response and of failure.

Although many investigators have focused on using intrinsic tumor characteristics to predict outcomes, because BCG renders its antineoplastic effects indirectly via stimulation of the immune system, host factors may also play a key role in treatment efficacy. In addition, accurate predictors of responsiveness ultimately rely on the understanding of BCG’s mechanism of action, which has been stubbornly difficult to elucidate. Nevertheless, several aspects of the BCG-elicited immunogenic reaction have been identified, measured, and linked to response. This article summarizes the published evidence on markers of response to intravesical immunotherapy with BCG.

Tumor factors

Two large studies by the Club Urologico Espano de Tratamiento Oncologico (CUETO) and European Organisation for Research and Treatment of Cancer (EORTC) groups attempted to link baseline clinicopathologic features to BCG response. In a cohort of 1062 patients, the CUETO group identified female gender, recurrent tumors, multiplicity, and the presence of carcinoma in situ (CIS) as predictors of recurrence, whereas recurrent tumors, high-grade tumors, T1 tumors, and recurrence on 3-month endoscopic examination were found to predict progression to muscle invasive bladder cancer (MIBC). From these findings, a scoring system was constructed to categorize the patients into 4 risk groups for recurrence (C index, 0.64) and progression (C index, 0.69–0.70).

In a similar study consisting of 1812 patients, the EORTC found that prior recurrence rate greater than 1 per year, tumor multiplicity (≥4 tumors), and tumor grade predicted early recurrence after BCG induction (C index, 0.65–0.67). Similarly, late recurrence was also predicted by prior recurrence rate greater than 1 per year and tumor multiplicity (C index, 0.56–0.59). In contrast, tumor grade and T stage were significant predictors of disease progression (C index, 0.64–0.72). One notable difference between the 2 cohorts studied was the difference in the maintenance BCG schedule. Although the CUETO group used a nonstandard maintenance protocol of 6 fortnightly treatments after induction, the EORTC study was conducted in patients who received 1 to 3 years of maintenance therapy in accordance with the Southwest Oncology Group (SWOG) (6 + 3) protocol. Nevertheless, it is noteworthy that these 2 large studies independently found prior recurrence and tumor multiplicity to be predictors for future recurrence, and high grade and stage to be predictors for progression after BCG treatment.

The association between high tumor grade and stage with subsequent disease progression following BCG treatment is likely a reflection of their poor prognostic value. Independent of treatment modality, progression rate in high-grade T1 bladder cancer was found to be 21%. In addition, understaging is a well-described phenomenon in bladder cancer, occurring in up to 50% of patients with presumed non–muscle-invasive high-grade disease. Although effective for NMIBC, BCG is thought to be futile against MIBC. Thus, treatment failure in a subset of patients with high-grade NMIBC may be a byproduct of understaging.

In addition, the presence of CIS has been extensively examined as a predictor of BCG failure. In 2 cohorts treated only with induction BCG, concomitant CIS was found to be a predictor of shorter progression-free survival (PFS) and cancer-specific survival (CSS). ^, The effect was especially pronounced in patients with T1 disease treated with induction BCG, in whom the presence of CIS led to a 58% progression rate to T2 disease or higher. This finding was subsequently corroborated in a cohort treated with induction and maintenance therapy.

Contrastingly, the association between prior recurrence and tumor multiplicity with response to BCG is more controversial, with different studies yielding divergent results. However, many studies indicating no association were underpowered or did not use maintenance therapy according to current standards. A recent review by an expert panel deemed these tumor characteristics to be important for predicting treatment response.

Aside from the traditional clinicopathologic features, expression levels of molecular biomarkers also correlate with response to therapy. Of these, p53, a cell cycle regulator, has been the most extensively studied. Several groups have linked immunohistochemistry (IHC) p53 overexpression to disease progression after intravesical BCG. However, it is unknown whether p53 overexpression on IHC correlated with loss of function. In addition, variations in p53 quantification methods and arbitrary thresholds make it difficult to compare results across studies. For instance, several different anti-p53 antibodies are commercially available. These antibodies differ with regard to the epitope each recognizes, and thus can lead to different results. Overall, it is generally thought that the available information does not justify the use of p53 in clinical decision making.

A multitude of other molecules have been examined as potential predictors for BCG response, including cell cycle regulators (Retinoblastoma protein), apoptosis inhibitors (survivin, bcl-2), cell adhesion molecules (E-cadherin, ezrin), and markers of proliferation (Ki-67). However, these studies all have similar shortfalls to those investigating p53: inconsistent diagnostic standards, inadequate study populations, and the lack of validation. Given the immense phenotypic heterogeneity within bladder cancer as well as the complex mechanism of action of BCG, it is unlikely that single biomarkers can accurately predict treatment success.

One analysis accounting for some of the genotypic heterogeneity in bladder cancer is the UroVysion fluorescence in situ hybridization (FISH) test, which detects increased copy numbers of chromosomes 3, 7, and 17, and loss of 9p21, all putative chromosomal abnormalities found in bladder cancer. A positive FISH result after BCG induction confers increased risk of recurrence (3-fold to 5-fold) and progression (5-fold to 13-fold), depending on timing of FISH positivity. By virtue of its ability to anticipate tumor formation, FISH is a valuable clinical tool for predicting failure after BCG. Because many patients who have a positive FISH test have no visible tumor at the time of assessment but subsequently develop recurrence in 6 to 24 months, this phenomenon has been dubbed molecular failure and such patients would be ideal candidates to enroll into clinical trials for combination immunotherapeutic agents.

The use of next-generation sequencing for the comprehensive molecular characterization of bladder cancer has not only shed light on tumor biology but also provided clues for molecular mechanisms of treatment success and failure. With regard to chemotherapy for MIBC, therapy-driven clonal evolution leading to chemoresistance has been shown. Furthermore, somatic mutations in DNA damage repair (DDR) genes also seem to confer cisplatin-based chemosensitivity, and molecular subtyping of MIBC has been linked to different phenotypic responses after neoadjuvant chemotherapy.

In NMIBC, polymorphisms impairing cellular DDR have also been associated with better outcomes after BCG. This association is thought to stem from the higher mutational burden and neoantigen load, which ultimately provokes a stronger immunogenic response. In a recent study of patients with high-risk NMIBC, a higher total mutation burden was found in patients who responded to intravesical therapy compared with those who did not. In contrast, initial efforts in NMIBC molecular subtyping did not correlate with BCG response. However, others have identified ARID1A mutations to be significantly associated with an increased risk of recurrence after BCG treatment (hazard ratio, 3.14; P = .002). Further elucidation of the NMIBC molecular landscape and refinement of classifications may lead to insights on tumor biology and therapeutic efficacy.

Host factors

From the CUETO study, female gender and age both emerged as poor prognosticators for BCG response. Palou and colleagues subsequently corroborated the higher recurrence and progression rates in women treated with BCG. The urinary cytokine profile observed in women after BCG treatment is different than that found in men. Because stimulation through the androgen axis is implicated in the development and progression of bladder cancer, it is conceivable that differences in the hormonal milieu may also play a role in the responsiveness to immunotherapy.

Poorer outcomes seen in elderly patients intuitively stem from a weaker BCG response caused by their waning immune systems. ^, At first glance, patients more than 80 years old had the poorest recurrence-free survival (RFS) in a subset analysis of the phase 2 combination BCG/interferon (IFN)-α trial. In another study, although initial response rates were equivalent in patients more than and less than 70 years old, more older patients recurred on long-term follow-up. Analysis of the prospective EORTC 30911 study recapitulated the poor prognostic effect of age on RFS, PFS, and CSS in BCG-treated patients ; however, even in older patients (>70 yeas old), BCG was more effective than epirubicin. Thus, old age seems to be more prognostic than it is predictive, and applies to patients undergoing all intravesical therapies.

Besides age, other host factors potentially affecting the immunogenic response have recently been interrogated. Variant polymorphisms in genes encoding several cytokines (interleukin [IL]-6, IL-17, IL-2, and tumor necrosis factor [TNF]-α), chemokines (MCP-1) as well as effector molecules (TNF-related apoptosis-inducing ligand [TRAIL] receptor) were associated with increased recurrence after BCG. ^, Adding these genomic signatures to key clinicopathologic features, a risk score was constructed and shown to achieve an area under the curve of 82% for predicting treatment response. In addition, genes involved in detoxification ( hGPX1 ), nucleotide excision repair, and regulation of macrophage susceptibility to intracellular mycobacterial growth ( NRMAP1 ) were also shown to be influential on BCG treatment outcomes.

Polymorphisms impairing cellular DDR may lead to higher mutational burden and levels of neoantigens that ultimately provoke a stronger immune response following BCG. Both DDR polymorphisms and high mutational burden have been linked to response to intravesical BCG. ^, Overall, associations between gene polymorphisms and BCG response warrant further exploration. Moreover, because most such studies were performed in homogenous ethnic and/or geographic populations, whether these associations can be extended to the global population remains to be seen.

Emerging evidence suggests that the unique microbial ecosystems may profoundly affect response to different cancer treatments. In its simplest form, many members of the gastrointestinal microbiota are known to influence the metabolism, pharmacokinetics, and toxicity of drugs. For instance, Mycoplasma hyorhinis can metabolize and inactivate the chemotherapy agent gemcitabine, leading to drug resistance. Although the mechanism is not yet completely understood, the efficacy of immune checkpoint blockade is also influenced by the gastrointestinal microbiome. ^, In animal models, eradication of certain commensal organisms using antibiotics led to treatment resistance to cytotoxic T lymphocyte–associated protein 4 (CTLA-4) and PD-L1 (programmed death-ligand 1) blockade. ^, In contrast, oral administration of Bifidobacterium potentiated the effect of PD-L1 therapy. In humans, a study of a cohort of patients with renal cell carcinoma, urothelial carcinoma, and advanced non-–small cell lung cancer found that patients treated with antibiotics before or shortly after the administration of immune checkpoint blockade had significantly shorter PFS and overall survival. Moreover, Akkermansia muciniphila was found to be the critical organism, whereby oral supplementation of the bacteria to antibiotic-treated mice restored responsiveness to immunotherapy.

In light of the newly discovered relationship between the microbiome and anticancer therapy responsiveness, it is conceivable that response to BCG can also be influenced by interactions with the commensal bacteria in the genitourinary tract. There is already evidence that bacillus binding, thought to be mediated by interactions between the bacterial cell wall and fibronectin in the urothelium, can be affected by the presence of Lactobacillus iners . Whether the intravesical microbiome can influence the outcomes of BCG therapy in other ways remains to be seen.

Mechanistic factors

Despite decades of research, many questions regarding BCG’s mechanism of action remain unanswered. A strategy to understand its mechanism has been to compare the immune response engendered in responders versus nonresponders and attempt to gain insight from the differences seen. Early efforts attempted to correlate differences in clinical immune response before and after BCG with therapeutic efficacy. Because skin reactivity to purified protein derivative (PPD) is the gold standard to detect antituberculin immunity, some investigators postulated that it can be used to characterize pretreatment BCG-specific immunity and predict improved antitumor response. Although earlier studies did not find a clear correlation, ^, 1 recent study by Biot and colleagues found RFS to be significantly improved in patients with positive PPD. Results from this study have prompted the launch of SWOG 1602, a randomized phase III clinical trial examining the effect of priming and boost response to BCG.

In addition, side effects of BCG treatment have also been used to predict response. It has been reported that patients who developed fevers during treatment have significantly lower recurrence rates. Subsequently, an analysis of the EORTC 30911 results also indicated improved response rates in patients with significant side effects. However, this observation could also be caused by the responders continuing to receive longer treatment courses of BCG, and thus reporting additional side effects. When limited to patients with and without symptoms within 6 months of treatment, no difference was found in RFS.

Investigating further, many studies have interrogated the difference between the cellular and cytokine response after BCG treatment. Higher levels of leukocyturia following BCG induction were associated with improved response. Specifically, polymorphonuclear cells, through the production of TRAIL, were implicated as the effector cells of cytotoxicity.

In addition, molecules associated with antigen presentation, such as heat shock protein (HSP) 90, are thought to play an integral part in facilitating effector cell recruitment and are necessary for treatment success. A greater increase in major histocompatibility complex (MHC) class I expression, especially on the tumor cells, has also been found to predict higher RFS after treatment. ^, Intriguingly, MHC II and intercellular adhesion molecule 1 (ICAM1), which are typically restricted to immune cells, have also been found on cancer cells in BCG responders. ^, Note that the expression of these molecules is thought to be influenced by interferon, which is upregulated during BCG treatment.

Besides tumor cells, antigen-presenting cells (APCs) may also contribute to the recruitment and activation of effector cells on exposure to BCG. However, whether the presence of APCs portends treatment success is unclear. Although immature dendritic cells (DCs) have been detected more frequently in the urine of BCG responders, high levels of mature, tumor-infiltrating DCs have also been shown to predict treatment failure. Similarly, antagonistic effects from different subsets of macrophages may differentially tip the balance toward treatment success or failure. Although those participating in T-helper 1 (Th1) response lead to tumor cell killing (M1), others involved in Th2 response are thought to stimulate cancer growth (M2). Along these lines, tumor-infiltrating cluster of differentiation (CD) 68+ tumor-associated macrophages were found to predict higher recurrence rates after BCG, indicating their involvement in the inflammatory circuit promoting tumor progression. ^,

In addition, it is unknown which cells are the key effectors in the BCG-elicited antitumor immune response. Early studies indicated that BCG-induced activity was lost in mice lacking lymphocytes, suggesting that these cells are necessary for the antitumor response. This theory was supported by IHC studies showing increased numbers of CD4+ T cells after BCG treatment. ^, Moreover, the number of CD4+ T cells and CD4/CD8 ratio in the pretreated tumors were found to be predictive of treatment response. In other studies, natural killer (NK) cells were found to be crucial for BCG-induced cytotoxicity. In a small study, NK cell and tumor interactions were found to be stronger in BCG responders than in non-responders. Without knowledge of the putative cytotoxic agent, it is difficult to pinpoint the cytotoxic response leading to treatment success.

Despite this lack of knowledge, insights can be obtained from the posttreatment immunogenic response. To this end, cytokine and chemokine profiles engendered after BCG instillation have been correlated with treatment success. Generation of the aforementioned Th1 polarized immune response is essential for an effective antitumor treatment. It is hypothesized that BCG is effective only when the tumor microenvironment (TME) converts from Th2 to Th1, and has no effect on the microenvironments already polarized to Th1. Quantification of eosinophil infiltration and degranulation (Th2 polarized) and the ratio of GATA-3+ (high in Th2-polarized TME) to T-bet+ lymphocytes (high in Th1-polarized TME) predicted BCG failure. The predictive value of GATA-3+/T-bet+ ratio has subsequently been validated.

Urinary cytokines levels may also be used to predict treatment success. IL-2, a canonical Th1 cytokine secreted by CD4+ T cells thought to stimulate proliferation and maturation of several downstream effectors, is detected in larger quantities in the urine collected from BCG responder than from non-responders. Furthermore, IL-2 levels peak earlier than IL-10 levels (surrogate for Th2 response) in responders, pushing the TME toward a Th1 response. ^, High IL-10 levels indicating a robust Th2 response, did not preclude treatment success. ^, ^, Other investigators found that it was the relative levels of the counteracting cytokines, rather than their absolute values, that predicted treatment outcomes. For instance, the ratio between IL-6 and IL-10 had 83% sensitivity and 76% specificity in predicting recurrence after BCG. ^,

Several other cytokines have been identified to be differentially expressed in the urine of responders and non-responders. Higher levels of IL-8 after BCG therapy significantly correlated with longer CSS. Other cytokines identified to be potential predictors of response include IL-18, TNF-α, IL-12, and TRAIL. The fact that a multitude of cytokines are related to treatment success is reflective of the complexity of the immune response resulting from BCG instillation. The authors prospectively tested the hypothesis that a panel of urinary cytokines can accurately assess the multifaceted immune response generated by intravesical BCG. In a prospective study of 125 patients, urine was collected at various time points and multiple cytokines assessed. Various time point and ratio combinations were studied using computational analysis. After extensive modeling, the inducible levels of cytokines at the last (sixth) BCG instillation, calculated as the difference between preinstillation levels and postinstillation levels (4 hours after BCG), was most predictive of response. The number of cytokines required was then reduced to the minimum required to retain predictive power. A nomogram (CyPRIT [cytokine panel for response to intravesical therapy]) using a panel of 9 cytokines (IL-2, IL-6, IL-8, IL-18, IL-1ra, TRAIL, IFN-g, IL-12[p70], and TNF-α) had an accuracy of 85.5% in predicting response to BCG (95% confidence interval, 77.9%–93.1%) ( Fig. 1 ). Efforts to validate the use of CyPRIT are currently underway.