**6. Predictive biomarkers of therapy dynamics**

#### **6.1 Genomic biomarkers**

#### *6.1.1 Tumor mutational burden*

Tumor mutational burden (TMB) refers to the frequency of non-synonymous mutations and is directly related to the neoantigen load. A high frequency of

mutations generally results in a high rate of neoantigen production, thereby increasing the probability of an immune response [140]. Therefore, TMB has been investigated and validated as a predictive biomarker for ICI response by numerous studies.

The association between TMB and a response to ICI has been extensively studied in NSCLC patients, however, with variable outcomes. After whole-exome sequencing (WES), a high mutational burden (>178 mutations per sample) observed in NSCLC patients treated with Pembrolizumab correlated to better ORR (68%) compared to patients with a low mutational burden (0%). Therefore, low TMB was correlated with poor efficacy in patients and is considered a marker of primary resistance to ICI treatment [140]. Similarly, a study with 4064 NSCLC patients showed that a high TMB had a significantly higher OS compared to a low TMB [141]. Numerous other studies have also shown a similar association between TMB and ICI response [142–144]. In contrast to these observations, a study whereby NSCLC patients were treated with Pembrolizumab and chemotherapy showed that TMB with >175 mutations per exome was not able to predict a response [145]. It is important to note that some tumors with a low TMB are still capable of responding to ICI. This highlights that, although TMB is a good indicator of ICI response, it is not the only determinant factor. On a broader scale, the correlation between TMB and response to ICI has been demonstrated across 27 tumor types [146]. The KEYNOTE-158 study with 750 participants showed that TMBhigh tumors were associated with better overall response rates (28%) and progressionfree survival (24%) compared to TMB-low tumors (7% and 14%, respectively). Interestingly, 12.5% of the TMB-high cohort were also mismatch repair deficient and were even more likely to respond to ICIs [147]. These studies provided compelling evidence for the use of TMB as a biomarker to determine benefit from ICIs.

Despite the association between TMB and ICI response, there are challenges that complicate the use of TMB as a biomarker in the clinic. TMB is typically measured using whole-genome sequencing (WGS), whole-exome sequencing (WES), or targeted next-generation sequencing (NGS). WES has been the standard method of choice but is resource-intensive and time-consuming and is most often used in a research setting. Therefore, in a drive for a more feasible detection method, multiple NGS panel assays were developed which targets specific sites of the genome [148]. The current challenge is the standardization of the method in terms of the regions that are targeted and sequencing depth [149]. The definition of TMB and sampling methods also limit its use. Variations in cancer types means there is no standard cut point in the definition for a high TMB or low TMB, and each tumor type may have its own optimal threshold to predict a response [150]. In addition, the sampling methods are invasive, and single biopsies can often lead to misclassification of the TMB due to tumor and intratumor heterogeneity. A study showed that 20% of NSCLC and 52% of urothelial cancers were misrepresented as a high TMB. Further multi-sample analysis revealed a low TMB [151]. Lastly, it would be useful to test the effect of TMB on a protein level for neoantigens, since only a subset of mutated genes result in potent neoantigens that are able to elicit an immune response [152]. Although numerous studies have provided supportive evidence for TMB as a predictive biomarker for ICI response, assessment of combination or multiple biomarkers in conjunction with TMB may have a stronger predictive value.

#### *6.1.2 Mismatch repair deficiency and microsatellite instability*

Mismatch repair genes such MLH-1, MSH-2, MSH-6, and PMS-2 are responsible for DNA repair. Loss of function in these genes is referred to as mismatch repair

deficiency (MMR-D). It leads to the accumulation of mutations during replication at a significantly higher rate than normal as well as the development of microsatellite instability (MSI) [153]. MMR-D/MSI is especially common in pancreatic, endometrial, cervical, prostate, and gastrointestinal cancers, including colorectal, gastric, and small intestinal cancer [154]. These tumors are particularly rich in frameshift mutations resulting in a high neoantigen load. Additionally, these tumors have also been found to contain a high level of infiltrating immune cells. These factors frequently enhance the immune response. Therefore, MMR-D can be used as a predictive biomarker for determining ICI response.

Clinical trials have shown that Pembrolizumab has durable outcomes in patients with MMR-D/MSI tumors. A study evaluating the efficacy of Pembrolizumab in colorectal cancer patients with and without MMR-D as well as MMR-D non-colorectal cancer patients showed promising results. For colorectal cancer with MMR-D, an overall response rate of 40% was observed whereas, for non-colorectal cancers with MMR-D, an overall response rate of 71% was observed. In contrast, patients without MMR-D exhibited an ORR of 0%. These results demonstrated that MMR-D patients produce a more favorable response to ICI treatment and are ideal candidates. This study led to the recommendation for MMR-D testing in metastatic colorectal cancer. In 2017, the FDA approved Pembrolizumab for patients with solid MMR-D/MSI tumors. This represents the first FDA approval for cancer treatment based on a genetic biomarker alone [155].

#### *6.1.3 IFN pathway profiles*

Activated CD8+ T cells secrete IFN-γ following binding to the MHC–peptide complex. IFN-γ is a cytokine that activates immune cells and stimulates an immune response. In the tumor cell, JAK/STAT signaling is activated by IFN-γ which results in the release of chemokines to promote an anticancer response. Moreover, IFN-γ triggers the upregulation of MHC-1 and PD-L1 expression promoting antigen presentation in APCs. IFN-γ expression was found to predict a positive response to PD-1 immune checkpoint inhibitors in melanomas and NSCLC. Conversely, mutations in IFN pathway genes such as IFNGR1/IFNGR2, JAK1/JAK2, STAT, and IRF1 have been associated with poor outcomes and resistance in patients receiving ICI therapy [156, 157]. In melanomas and MMR-D colorectal cancers, the loss of function in JAK1 and JAK2 have also been identified as mechanisms of both primary and secondary resistance to ICIs [158, 159].

A study including NSCLC and melanoma patients treated with Nivolumab and Pembrolizumab, respectively, indicated that increased expression of IFN-γ correlated with improved OS and PFS [160]. Similarly, another study investigating a four-gene IFN-γ signature (IFN-γ, CD274, LAG3, and CXCL9) in NSCLC patients treated with Durvalumab revealed that a positive signature for the gene set was associated with higher ORRs, PFS, and OS in comparison with signature-low patients [161]. It has also become increasingly common to assess IFN-γ in combination with other biomarkers such as TMB. A study in melanoma patients assessed both inflammatory gene profiles and the TMB. Patients treated with Pembrolizumab exhibiting high levels of both biomarkers had an ORR of 54% compared to an ORR of 14% in patients with low expression levels [162]. Furthermore, in melanoma patients treated with neoadjuvant Ipilimumab and Nivolumab, tumors with high IFN gene signatures and TMB displayed a 100% response rate, while tumors with low expression profiles of both had a 37% response rate [163, 164]. Similar results have been observed for NSCLC and

renal cell carcinoma [165]. These studies demonstrate the emerging role of inflammatory gene expression profiles as a predictive biomarker for ICI response. Challenges associated with the use of such gene panels arise from the replication of results due to intratumor heterogeneity and sampling methods, once again highlighting the limitations of single region sampling.

#### **6.2 Tumor-immune microenvironment biomarkers**

#### *6.2.1 PD-L1*

ICIs that target PD-1 or PD-L1 aim to disrupt the PD-1/PD-L1 axis, allowing cells to mount an antitumor response by preventing T cell downregulation [166]. Consequently, PD-L1 expression is one of the most extensively studied predictive biomarkers for response to ICI therapy. In the KEYNOTE-001 study, patients with PD-L1 expression of more than 50% had an ORR of 45% and improved PFS and OS. In comparison, patients who displayed 1–49% PD-L1 expression had an ORR of only 17% [167]. This study ultimately led to the approval of Pembrolizumab in NSCLC patients who display more than 50% PD-L1 and established the expression of PD-L1 as a companion predictive biomarker for patient selection. Positive correlations have also been seen for gastric cancer, colorectal cancer, and hepatocellular carcinoma [17, 168, 169]. Subsequent trials for PD-L1 as a predictive biomarker led to approvals by the FDA for urothelial, triple-negative breast cancer (TNBC), head and neck, gastric, esophageal cancers, and cervical cancer at various cut points.

PD-L1 expression has significant spatial and temporal heterogeneity. Expression varies between sites of the same tumor and between metastatic sites. Given this, the use of PD-L1 as a predictive biomarker has limitations. Detection is usually carried out using immunohistochemistry, but it is not adequately standardized. Even in the same cancer type, there are variations in thresholds. There are five main PD-L1 diagnostic antibodies that are available for detection. These antibodies have only been validated in the context of its companion drug trial: Pembrolizumab (Dako 22c3), Nivolumab (Dako 28–8), Durvalumab (Ventana SP263), Avelumab (Dako 73–10), and Atezolizumab (Ventana SP142). Variations in detection between assays have been noted. Dako 73–10 scores more cells as positive and Ventana SP142 scores more as negative leading to misinterpretations [170]. Detection of PD-L1 is frequently observed in patients who respond to anti-PD-1/ PD-L1 immunotherapies. However, [43] reported that even when NSCLC tumors displayed more than 50% PD-L1 staining, approximately half of the subset of patients still had primary resistance to Pembrolizumab. This study suggested that PD-L1 expression alone may be insufficient at predicting resistance. As with TMB, it is critical to note that PD-L1 does not preclude response to treatment. In the study mentioned earlier, although PD-L1 positive patients had a higher response rate, 15% of PD-L1-negative patients still responded [171].

#### *6.2.2 Tumor infiltrating lymphocytes*

Tumor-infiltrating lymphocytes (TILs) encompass lymphatic cell populations that invade the tumor tissue. TILs may promote an antitumor response (CD4+), exert cytotoxic antitumor activity (CD8+), or even limit a response (FOXP3+ Treg). These cells have therefore been associated with prognosis and response to ICI in many

*Current Advances in Immune Checkpoint Therapy DOI: http://dx.doi.org/10.5772/intechopen.107315*

cancer types, including NSCLC, TNBC, colorectal cancer, and melanoma. The density, location as well as phenotype of TILs give an indication of the response. In melanoma patients treated with Pembrolizumab, the spatiotemporal dynamics of TILs showed that the presence of CD4+ and CD8+ T cells at the infiltrative margin of the tumor was associated with patients who respond to treatment. The high density of cells allowed for increased infiltration into the tumor parenchyma of responders [172]. Another study revealed that responders had high levels of stromal TILs (50%) in comparison with non-responders (15%) for TNBC patients treated with Pembrolizumab [173]. An investigation into the temporal dynamics of TILs showed that an increase in TILs at 3 weeks, compared to the baseline reading, was correlated with response in melanoma patients treated with Ipilimumab [174]. Furthermore, the phenotype of TILs may also be used as a prognostic biomarker. A study showed that CD69+ CD103+ tumor resident CD8+ T cells were associated with improved survival in melanoma [175]. In contrast, FOXP3 tregs have been associated with poor survival in numerous cancer types [176]. The prognostic value of TILs has also been demonstrated by combining detection with PD-L1 expression to allow for better accuracy in determining response. Patients who exhibited high CD8+ TILs and low PD-L1 had an OS of approximately 93% in comparison with patients with low CD8+ TILs and high PD-L1 (61%). The authors suggested that CD8+ TIL combined with PD-L1 expression was better at predicting response than each biomarker alone [177].

### **6.3 Blood-based biomarkers**

### *6.3.1 Circulating tumor DNA and tumor cells*

The noninvasive nature of blood biopsies reduces patient suffering and provides certain advantages such as overcoming the heterogeneity issues of single sample tissue biopsies. It also allows multiple sampling throughout the disease progression and acquisition of real-time data. Therefore, there it is imperative to develop reliable blood-based biomarkers [178]. Emerging studies have linked circulating DNA (ctDNA) and circulating tumor cells (CTCs) found in the peripheral blood with response to ICI. In a study with melanoma patients, detectable baseline ctDNA that persist during treatment correlated with a poor response of only 6%. However, when ctDNA was initially detectable and became undetectable at 12 weeks, the response rate was 77% and when ctDNA was undetectable at both the baseline and 12 weeks, the response rate was 72% [179]. Thus, ctDNA may serve as a biomarker of response. Studies went further to assess TMB from the ctDNA. In NSCLC, it was shown that blood TMB correlated with tissue TMB and was associated with ICI response [180]. CTCs have also been suggested as prognostic biomarkers. In NSCLC patients, blood sampled before and after treatment with Nivolumab showed that high levels of CTCs before treatment was associated with an increased risk of disease progression and death [181].

#### *6.3.2 Soluble biomarkers*

Some indicators such as neutrophil-to-lymphocyte ratio (NLR), lactate dehydrogenase (LDH), and various cytokines (IL-6 and IL-8) have been studied as biomarkers of response to ICI in a variety of tumors [182]. Neutrophils that express PD-L1

attenuate the antitumor response by binding to PD-1 T cells. Therefore, NLR has been suggested to have a predictive role for response to ICIs in melanoma and NSCLC. A study of melanoma patients treated with Ipilimumab demonstrated that patients with an NLR > 3 had a poor OS and PFS [183]. Similar results were shown in another study where an NLR >5 was also associated with a lower OS and PFS [184]. In advanced solid tumors, the OS of high NLR patients was 8.5 months, while the OS of patients with a low NLR was 19.4 months [185]. Changes in LDH during ICI treatment correlates with patient response. A study showed that patients who displayed an elevated baseline serum LDH value had a shorter OS at 12 months (44%) compared to patients with normal LDH values (71%). Moreover, a 10% increase from the baseline level during ICI treatment also indicated poor ICI efficacy [186]. Lower levels of the cytokine IL-6 at the baseline and on treatment have been correlated with improved response, while higher levels of IL-6 correlate with a shorter OS [187]. Additionally, in NSCLC and melanoma, it was reported that lower levels of IL-8 were associated with improved treatment responses, while higher baseline IL-8 levels were associated with poorer OS [188].
