TP53 mutation is associated with improved disease control in patients with advanced RAS wild-type colorectal adenocarcinoma treated with cetuximab and pembrolizumab.

Introduction
Colorectal adenocarcinoma (CRC) is the 2nd most common cause of cancer-related death in the United States with an expected 5-year overall survival (OS) rate of less than 15% for patients with metastatic disease 1. This dismal prognosis, despite the significant improvements in systemic therapy, including use of biomarker-based targeted therapy in select patients, stresses the need for better therapeutics 2–4. Immune checkpoint inhibitors targeting the programmed death-1 (PD1)/PD-Ligand1 (PD-L1) and the cytotoxic T-lymphocyte antigen-4 (CTLA4) pathways have revolutionized the treatment of microsatellite instable (MSI-H) CRC but they are not active in the most common microsatellite stable (MSS) CRC 5–7. Contrary to MSI-H CRC, the immune tumor microenvironment (TME) in MSS CRC is characterized by a relative abundance of regulatory T-cells (Treg), TH17 T-cells, myeloid-derived suppressor cells (MDSC), and M2 protumor tumor-associated macrophages (TAM) and tumor-associated neutrophils (TAN) resulting in a cytokine and chemokine milieu favoring immune suppression and tumor progression 8. Furthermore, intratumoral infiltration with exhausted CD8+ T-cells expressing more than one inhibitory immune checkpoints (i.e., PD1 and TIM3) and overall exclusion of CD8+ T-cells from the tumor core are common in MSS CRC 8.
In our effort to improve outcomes with immune checkpoint inhibitors for MSS CRC, we conducted a clinical trial testing the combination of the anti-PD1 monoclonal antibody (mAb) pembrolizumab plus the anti-epidermal growth factor receptor (EGFR) mAb cetuximab in patients with pre-treated, advanced CRC without oncogenic RAS mutations (RAS wild-type, RASwt) based on clinical, standard care sequencing 9. We observed that while this regimen did not improve outcomes compared to historic results, it did induce favorable changes in the TME including an increase in intratumoral cytotoxic CD3+CD8+ T-cells (CTL). The non-tumor cells in the TME expressed several exhaustion markers with striking differences between the primary and metastatic tumor sites. Our data also suggested that the clinical efficacy of anti-EGFR-based therapies may require simultaneous targeting of alternative T-cell exhaustion pathways and additional augmentation of innate immunity.
Loss of function genetic alterations in the TP53 gene (p53mt) are present in most CRC and drive relapse and metastasis 10–12. Importantly, p53mt CRC is less enriched in CD8+ T-cells, natural killer (NK) cells, and immune cytolytic activity (antitumor immune) signatures compared to TP53 wild-type (p53wt) CRC 13. Additionally, exosomes released from CRC cells increase pro-tumor immune cell populations within the TME through delivery of several long (lncRNA) and short non-coding RNAs (miRNA) 14–18. This correlates with the lack of benefit from immunotherapy in the p53mt enriched MSS CRC 19, 20. While the abnormal p53 protein can accumulate intracellularly leading to p53-specific humoral immune responses in CRC, the effect of this in OS is unclear 21.
We hypothesized that p53mt is a negative predictive biomarker for treatment with cetuximab plus pembrolizumab in patients with RASwt CRC. Unexpectedly, we observed improved progression-free survival (PFS) and greater decrease in tumor size among p53mt compared to p53wt tumors. In this paper, we describe these findings along with a comprehensive molecular analysis in a subgroup of patients incorporating genetic, genomic and proteomic data. Additionally, we compared these results with a comprehensive bioinformatics analysis of data from The Cancer Genome Atlas (TCGA).

Materials and Methods

Design
The methods and results of the clinical trial NCT02713373 (Roswell Park study I-274515) have been previously described 9. In summary, 44 patients with pre-treated, advanced RASwt CRC were treated with cetuximab plus pembrolizumab in 21-day cycles. The RAS status was determined using standard care testing. The study was negative for the primary endpoints with overall response rate (ORR) of 2.6% and a 6-month PFS of 31%; the median PFS was 4.1 months. Tumor biopsy was performed at baseline and before cycle 4 (optional biopsy). Baseline archival tissue samples were allowed if a biopsy was not feasible or posed excess risk. Blood samples for correlative analyses were collected at baseline and prior to cycles 2 and 4. Multispectral immunofluorescence staining of tumor tissue and flow cytometry on single cell suspensions of tumor tissue and peripheral blood mononuclear cells (PBMC) were performed as previously published 9.
For this post hoc exploratory analysis, we included genomic data beyond the RAS status that were collected as part of standard care therapy. As multiple patients on trial did not have extended panel next-generation sequencing (NGS) performed as standard of care, we performed DNA and RNA sequencing on tumor samples as part of a separate Roswell Park de-identified biospecimen and data research protocol (BDR, BDR-147321). Additionally, we performed whole transcriptome sequencing. Eighteen patients had tissue available for dual DNA/RNA extraction and sequencing. Of these, 10 had available matched on-treatment tumor samples. Subsequently, we amended the clinical trial I-274515 protocol (Amendment 6) to link these genomic/genetic data with clinical and other correlative outcomes from the clinical trial. When patients had both clinical and research DNA sequencing that included testing for TP53, the clinical sequencing data were entered in the study I-274515 database. Patients who had only clinical DNA sequencing (tumor or cell-free DNA) that was collected after study enrollment were excluded from the biomarker analysis cohort.

DNA/RNA extraction from tumor samples
The purification of total RNA and genomic DNA from formalin-fixed paraffine-embedded (FFPE) tumor samples was prepared using the Truxtrac FFPE total NA Plus Kit (Covaris). Samples were provided as either FFPE slides, cores or curls. The FFPE samples were emulsified in Tissue Lysis Buffer in the presence of Proteinase K on the Covaris E220 sonicator. This was followed by an incubation at 56 °C for a short duration. The RNA-containing supernatant was separated from the DNA-containing tissue by a centrifugation step. RNA was then de-crosslinked, DNAse digested, and purified over an RNA-specific spin column (Covaris). Sequentially, DNA was released from the DNA-containing tissue by the Covaris E220 sonicator. The resuspension was then subjected to an additional Proteinase K digestion, following a de-crosslinking step at 80 °C. DNA was then purified over a DNA-specific spin column (Covaris). Quantitative assessment of the purified total RNA and Genomic DNA was then accomplished by using Qubit RNA and DNA quantitation kits (Thermofisher). The RNA is further evaluated qualitatively by a Tapestation 4200 (Agilent technologies).

DNA sequencing (research)
Next generation sequencing libraries were prepared with the VariantPlex Solid Tumor v1.2 (ArcherDx), using 100ng DNA. Following manufacturer’s instructions, samples were end-repaired, followed by ligation of index adapter sequences and molecular barcodes. Two rounds of Anchored Multiplex PCR using gene specific primers from the desired panel generate target amplicons with complete adapter sequences for Illumina sequencing were performed. Final libraries were validated for appropriate size on a 4200 TapeStation D1000 Screentape (Agilent Technologies, Inc.). The sequencing libraries were quantitated using KAPA Biosystems qPCR kit (Roche), and were pooled together in an equimolar fashion, following the manufacturer’s instructions. They were then denatured and diluted to 2.4pM with 20% PhiX control library added. The resulting pool was then loaded into the appropriate NextSeq Reagent cartridge and sequenced on NextSeq500 following the manufacturer’s recommended protocol (Illumina Inc.). The sequencing coverage and quality statistics for each sample and the list of genes targeted are summarized in Supplementary Table 1A.

Whole transcriptome analysis
Whole transcriptome analysis of FFPE samples was performed using SureSelect XT HS2 RNA Target Enrichment System (Agilent Inc). 200ng of RNA was fragmented and converted to cDNA followed by end repair, duplex UMI adaptor ligation, and 12 cycles of PCR to complete the dual-index adapters. Unique dual-indexed, molecular-barcoded libraries were purified with AMPureXP beads (Beckman Coulter) and validated for appropriate size (225–275bp) on a Tapestation 4200 DNA1000 screentape (Agilent Inc.). 200ng of purified library was then hybridized to the SureSelectXT Human All Exon V6+UTR Capture library (Agilent Inc.). The hybridized regions were then bound to streptavidin magnetic beads and washed to remove any non-specific bound products. Eluted library underwent a second 13 cycle PCR amplification to generate enough material for sequencing. Final libraries were purified, measured by Tapestation 4200 DNA1000 screentape (250–350bp), and quantitated using KAPA qPCR (KAPA Biosystems). Individual libraries were pooled in equimolar 2nM final concentration. The resulting pool is then loaded into an appropriate NovaSeq flow cells for 101 cycle paired-end sequencing and sequenced on a NovaSeq6000 following the manufacturer’s recommended protocol (Illumina Inc.) to achieve an average of 50 million paired-end reads per sample. The sequencing coverage and quality statistics for each sample are summarized in Supplementary Table 1B.

Plasma Cytokine Analysis
The CodePlex Human Adaptive Immune Panel kit (Bruker Cellular Analysis) for the cytokines GM-CSF, Granzyme B, IFN-γ, IL-2, IL-4, IL-5, IL-6, IL-7, IL-8, IL-9, IL-10, IL-13, IL-15, IL-17A, IP-10, MCP-1, MIP-1α, MIP-1β, Perforin, sCD137, TNF-α, and TNF-β was used to analyze cytokine concentrations. Patient plasma samples were allowed to thaw until room temperature. While waiting for samples to thaw, the chips from the CodePlex kit were allowed to thaw for 1 hour and remaining kit components were prepared as recommended by the product protocol from the manufacturer. Samples were loaded in duplicate onto the chips before being loaded into the Isolight instrument. Automated analysis of the raw data to obtain concentrations in pg/mL was performed using the IsoSpeak software (Bruker Cellular Analysis) and then exported to an excel file for further analysis.

Statistical Considerations
Summary statistical analyses were provided for the demographic features of the patients included in this study. The differences of the demographics and outcomes between the two TP53 mutation subgroups were compared by Non-parametric Kruskal-Wallis Test and Fisher’s Exact Test. Log-rank Test/Cox Proportional Hazards Analysis was performed for OS and PFS. The OS and PFS were estimated by the Kaplan-Meier method. SAS version 9.4 (SAS Institute, Cary, NC) was used for statistical analyses. All tests are two-sided and performed at a nominal significance level of 0.05.

Bioinformatics Analysis
Samples were either obtained from the Roswell Park Comprehensive Cancer Center Genomics Shared Resource or from the TCGA Firebrowse. Pre-processing and quality control (QC) steps were carried out by the Roswell Park Bioinformatics Shared Resource using an established pipeline following commonly adopted practices for RNA sequencing data analysis. Principal Component Analysis to assess sample similarity was carried out using the ‘PCAtools’ package in R. Differential Gene Expression (DGE) analysis was then carried out with DESeq2 22. Differentially expressed genes were shown in volcano plots, using the ‘EnhancedVolcano’ package in R, and key transcripts were displayed as faceted boxplots using the ‘gplots’ package in R. Further, differential expression rank order was used for subsequent Gene Set Enrichment Analysis (GSEA) 23, performed using the ‘clusterProfiler’ package in R on collections available through the Molecular Signatures Database (MSigDB) 24. Heatmaps were used to visualize differential expression and Consensus Molecular Subtype (CMS) signatures 25 and were constructed using the ‘pheatmap’ package in R. Finally, immune deconvolution was conducted for 6 different algorithms using ‘TIMER’ 26. All analyses were performed using R statistical software, version 4.1.1. The quality assessment of the data is presented in the Supplementary Figure 1.

Results

Demographics
Thirty-five patients had extended NGS results that included TP53 sequencing (standard care and/or research, Table 1) and constitute the biomarker analysis cohort. Twenty-seven patients had a p53mt tumor. There were no differences in baseline demographics in the biomarker analysis cohort vs. overall study population (Supplementary Table 2). All patients had prior therapy with fluoropyrimidine-based therapy. Only 3 patients had prior treatment with anti-EGFR mAb (all with p53mt tumor). The only difference in baseline characteristics was higher proportion of men in the p53mt group.

DNA alterations
Figure 1 shows the mutations detected with NGS. The most common alterations beyond TP53 mutations were mutations in APC (46%), AKT1 (43%), VHL (43%), CDKN2A (37%), ERBB4 (34%), RET (31%). We also identified 3 patients with KRAS mutant tumors in research testing, most likely related to intratumoral heterogeneity and/or higher sensitivity of the research panel.

Efficacy analysis
Patients with p53mt tumor had superior PFS compared to patients with p53wt tumors (Figure 2A). The median PFS was 5.5 (3.9–6.4) and 2 (1.8–4.3) months in patients with p53mt and p53wt tumors, respectively (p<0.001). Additionally, more patients with p53mt tumor had decrease in the size of target lesions (61% vs. 0% in p53wt tumors, p=0.005). The median percent decrease in target lesions was 10% (range 60% decrease - 30% increase) in p53mt tumors vs. 30% increase in p53wt tumors (range 10–40% increase, p=0.001). Five patients (18%) with p53mt tumors had decrease >50% in CEA vs. none in p53wt (p=0.063). There was no difference in OS between the 2 groups (median 13.8 vs. 14.5 months, p= 0.734, Figure 2B). Since the standard care DNA sequencing was performed with a variety of commercial panels that may have not included the same genes in their reports, we did not formally evaluate the effect of co-mutation profile in PFS and OS. With this limitation, the 3 patients with KRAS mutant tumors on research testing did not attain significant clinical benefit from study treatment (Figure 1).

Cetuximab/Pembrolizumab has a divergent effect in peripheral blood PD-1+ CTLs of patients with p53mt vs. p53wt CRC
We have previously shown that treatment with cetuximab and pembrolizumab can induce changes in many immune cell populations in peripheral blood over time, some of which significantly correlate with treatment outcomes 9. We compared baseline and on-treatment changes in the same peripheral blood immune cell populations stratified by the TP53 status (Supplementary Table 3). There were no statistically significant differences in CTLs or other immune cell populations at baseline or on-treatment between p53mt and p53wt tumors. Notable discordant changes were observed between the two groups, where the PD-1+ CTLs (overall, CD45RO−, and CD45RO+) decreased in patients with p53mt tumors (66%, 71%, and 64%, respectively) while increased in patients with p53wt tumors (58%, 255%, and 32%, respectively) in on-treatment versus baseline samples. Treg (CD4+CD25+FoxP3+
T-cells) also decreased on-treatment in the p53mt tumors (approximately 4% decrease vs. 38% increase in p53wt tumors). While GARP+ Treg decreased on-treatment in both groups, CTLA4+ Treg increased by 16% in the p53mt group and decrease by 1% in the p53wt group.

Effects of Cetuximab/Pembrolizumab on the immune TME of p53mt vs. p53wt CRC
Our previously published data indicated an increase in the numbers of tumor-infiltrating CTL on-treatment 9. Transcriptomic analysis revealed several differentially expressed genes in 10 matched pre- and post-treatment samples (Figure 3A). GSEA showed uniform upregulation in several metabolic gene sets and most immune gene sets including NK-mediated immunity and IL-12 pathway while the IL6 pathway was downregulated (Figure 3B). To understand the different patient outcomes based on the TP53 status, in the present analysis we examined differences in the TME between p53mt and p53wt tumors at baseline and on-treatment using different methods. We first repeated the transcriptomic analysis based on the TP53 status of the tumor. Only 2 tumor samples were p53mt, limiting statistical significance. Overall, when comparing pre- and post- treatment transcriptional profiles from both the p53mt and p53wt patients, there were no overlapping pathways, despite the therapeutic intervention, suggesting p53 status may dictate differences in transcriptional therapeutic response (data not shown). Despite the lack of transcriptional overlap, we did identify functional overlap with treatment in both groups in proliferative, immune, and metabolic pathways (Figure 4). The immune deconvolution analysis in 18 baseline tumor samples (88% of those being p53mt) showed increased myeloid (mature and progenitor), B-cell and NK cell populations in p53wt tumors (Figure 5). M1 macrophage, monocyte, NK, common myeloid and granulocyte-monocyte progenitor cells decreased on-treatment in p53wt tumors while M2 macrophages increased in p53mt tumors (Supplementary Figure 2). The overall immune deconvolution score from the TIMER algorithm increased in p53mt vs. decreased p53wt tumors (Supplementary Figure 2). Given the limited number of samples, no significant differences were found; however, given trends, we expect to see significant differences with a larger patient population.
Finally, we repeated these analyses in the TCGA dataset. This dataset comprised of 176 RASwt CRC samples, 123 p53mt and 53 p53wt. Several immune and metabolic pathways overlapped between the TCGA and I-274515 datasets (Supplementary Figure 3). There was enrichment of lymphocyte- and leukocyte-mediated immunity as well as of lipid and amino acid metabolism. The immune deconvolution of the TCGA data revealed transcriptional alterations to B-cells, dendritic cells and central memory cell populations in p53mt vs. p53wt tumors (Supplementary Figure 4).
We then evaluated differences in immune cells populations in the TME of p53mt vs. p53wt tumors on-treatment with flow cytometry (Supplementary Table 4). There were no significant differences. CTLs increased while PD1+ CTLs (total and CD45RO+) decreased in both groups on-treatment. A few notable discordant trends were evident on-treatment. When comparing baseline with on-treatment tumor biopsies, T-cells, CD45RO+ and PD1+ CD45RO− CTLs increased in p53mt tumors (14%, 10%, and 35%, respectively) while they decreased in p53wt tumors (29%, 7%, and 57%, respectively). Finally, the NK (CD3−CD56+CD16+) cells increased by 230% in p53mt tumors vs. 146% increase in p53wt tumors. Subsequently, we evaluated the expression of PD-L1, T-cell exhaustion markers (TIM3, CTLA4, LAG3) and the activation marker OX40 using multispectral immunofluorescence staining (Supplementary Table 5). We characterized several immune cell phenotypes based on the expression of activation (OX40) and exhaustion markers (CTLA4, TIM3, LAG3) 9. The number of PD-L1+ tumor cells was higher in p53mt vs. p53wt tumors (p=0.02) while the number of OX40−/AE1_AE3−/PD-L1− non-tumor cells, positive for LAG3, CTLA4 or TIM3) was higher in p53wt tumors (p=0.03) at baseline. On-treatment analyses were only feasible in a very small subset of patients (1 with p53mt and 4 with p53wt tumor); the PD-L1+ tumor cells decreased by 80% in the sole p53mt tumor while the median number increased by 102% in the p53wt tumors. Additionally, there were discordant changes in the expression of other immune checkpoints between the 2 groups. None of the on-treatment comparisons reached statistical significance.
Finally, since the CMS may be prognostic in CRC patients treated with anti-EGFR mAb 27, we evaluated the CMS in baseline tumor specimens. Most tumors could not be assigned to a specific subgroup; 2 tumors were CMS2 and 4 had mixed characteristics (Supplementary Figure 5). In this small dataset, we could not identify any prognostic implications of CMS (data not shown).

Plasma cytokine profile in patients with p53mt vs. p53wt CRC
We performed additional analyses evaluating plasma cytokines as prognostic and predictive biomarkers of response. The median concentration of each cytokine was used as cutoff for high- vs. low-biomarker (Supplementary Table 6). High baseline IL-2 and IL-6 were predictive for worse OS (p=0.02 and 0.05, respectively). High baseline IL-8 was predictive for worse PFS (p=0.04). Increase in IL-13 on-treatment was predictive for better OS and PFS (p=0.02). High perforin before cycle 4 correlated with improved PFS (p=0.002). The plasma levels of GM-CSF, INFγ, IL-8, IL-15, IL-17A, TNFα and TNFβ were significantly lower in patients with p53mt vs. p53wt CRC at baseline (Supplementary Table 7). There were no statistically significant changes on-treatment of any of the tested cytokine levels on-treatment between the 2 groups. Several cytokines (IL-13, MCP-1/CCL2, MIP-1α/CCL3) increased on-treatment in patients with p53mt tumors while they decreased in patients with p53wt tumors; the opposite trend was evident in IL-10 concentration on-treatment.

Discussion
In our study, patients with p53mt CRC had improved PFS with cetuximab and pembrolizumab treatment compared to patients with p53wt CRC. To our knowledge, this is the first report that the TP53 status of the tumor may have an impact on the short-term outcomes of immune checkpoint inhibitor-based combinations in MSS CRC. Taking into consideration the poor prognosis of patients with p53mt CRC 28 and the fact that only the PFS and not the OS was affected by the tumor TP53 status in our study, this should be further investigated as a possible positive predictive biomarker for patient selection for immunotherapy-based trials in MSS CRC.
There is growing evidence that the status of p53 can regulate immune responses in cancer 29. p53 is involved in the CTL function by a) upregulation of antigen presentation via the major histocompatibility complex I pathway 30, 31, b) PD-L1 downregulation via miRNA-34 32, and c) induction of Fas-dependent apoptosis 33. Furthermore, p53mt may lead to p53-specific humoral immune responses in CRC 21. The effects of p53mt on anticancer immunity and response to immune checkpoint inhibitors may be disease-dependent 13. Additionally, we have previously shown that p53mt CRC have increased tumor mutation burden (TMB) 34. In patients with localized CRC, higher TMB correlates with increased numbers of tumor-infiltrating lymphocytes but patients with MSS p53mt tumor or with mutations in TP53 pathway have lower numbers of tumor-infiltrating lymphocytes 35. TMB, when sufficiently high (>10 mutations/Mb), is a biomarker for CRC patient selection for anti-PD1 therapy but the antitumor activity is limited (PFS 3 months, ORR 11%) 36. This is most likely related to differences in predicted immunogenic neoantigens and the T-cell clonality, activation status, and interactions with other cells in the TME like macrophages 37. Additionally, tumors from patients treated with anti-EGFR mAbs can acquire secondary mutations both outside (passenger mutations) and within the EGFR/RAS/BRAF/MAPK axis more frequently compared to tumors that are treated without these agents 38. Passenger mutations are a surrogate of adaptive mutability, the transient downregulation of mismatch repair and homologous recombination DNA repair genes and concomitant upregulation of error-prone polymerases in EGFR-resistant cells leading to DNA damage, increased mutability, and triggering MSI 39. It is plausible that p53mt tumors are more prone to adaptive mutability when treated with EGFR mAbs leading to improved outcomes when these agents are used in combination with immune checkpoint inhibitors.
Furthermore, the number of PD-L1+ tumor cells at baseline was significantly higher in p53mt vs. p53wt tumors in our study. Additionally, the number of OX40−/PD-L1− non-tumor cells positive for at least one exhaustion marker beyond PD1 (LAG3, CTLA4 or TIM3) was significantly higher in the TME of p53wt tumors. The immune deconvolution analysis in a subgroup of patients with mostly p53mt CRC showed difference in the size of several immune cell populations at baseline and changes on-treatment. Our data indicate that it is plausible that, while overall the tumor-infiltrating CTLs increase with treatment 9, the experience and the exhaustion/activation status of these T-cells may be different in p53mt vs. p53wt tumors in response to combined EGFR/PD1 inhibition as is the status of the immune checkpoint pathways PD1/PD-L1, TIM3, LAG3, CTLA4, and OX40. In peripheral blood, PD1+ CTLs decreased on-treatment in patients with p53mt tumors; the opposite was true in patients with p53wt tumors. Changes in PD1+ CTLs in the TME was more uniform between the 2 groups. The mechanisms behind the differences in immune cell populations in the TME and peripheral blood in response to treatment with cetuximab and pembrolizumab depending on the tumor’s TP53 status cannot be fully characterized from this dataset but is most likely related to altered interactions between tumor, stromal and immune cells. In support, the GSEA revealed functional dysregulation at the baseline and on-treatment samples that varied between the two TP53 groups, with alterations to immune, epigenetic, and metabolic pathways. Furthermore, the cetuximab mechanism of action (at least partially) involves antigen-dependent cell cytotoxicity (ADCC) mediated via the NK cells. In our study – along with significant increase in intratumoral CTL – we observed increased intratumoral NK cell infiltration 9 that was more pronounced in patients with p53mt tumors, though it did not reach statistical significance. PD1 is a negative immune checkpoint for both T-cells and NK cells 40. Thus, pembrolizumab may synergize with cetuximab by improving NK cell responses, an effect possibly more relevant in p53mt RASwt tumors. Finally, the CMS subtype is associated with clinical benefit from anti-EGFR therapy, with the best outcomes seen in CMS2 (canonical) followed by CMS4 (mesenchymal) and then CMS3 (metabolic) 27. Interestingly, this effect appears to be dependent on the chemotherapy backbone 41. Additionally, the presence of both TP53 and APC mutations, enriched in CMS2, is associated with increased anti-EGFR sensitivity 42 and CMS1 tumors are hypermutated and immunogenic 43. Most of the baseline tumor samples in our study could not be assigned to a specific subtype while 4 samples had a mixed phenotype (features of CMS1, CMS2, and CMS3). The mixed phenotype has been associated with worse prognosis 44, a finding that we could not replicate in our study.
The pathogenesis and progression of CRC is heavily influenced by cytokines and chemokines secreted by tumor, stromal and immune cells 45. In line with its role in CRC pathogenesis 45, elevated baseline plasma IL-6 was a negative prognostic biomarker in patients treated on study. Interestingly, systemic IL-15 and INFγ, both associated with the antitumor effects of CTL and NK cells 8, were lower at baseline in patients with p53mt CRC. Lower systemic levels of IL-17A at baseline, a TH17-related cytokine with multiple tumor-promoting immune effects 8, in patients with p53mt tumors may have contributed to improved CD8+ T-cell/NK cell infiltration on-treatment despite the lower baseline systemic IL-15 levels. Non-significant trends (increase) were noted in the concentration of the CD8+ T-cell/NK cell-attracting chemokine MIP-1α/CCL3 (antitumor), the TH2 cytokine IL-13, and the monocyte chemoattractant MCP-1/CCL2 (tumor-promoting) on-treatment in patients with p53mt tumors. Additionally, in the whole study population, on-treatment increase in IL-13 correlated significantly with improved OS and PFS. Furthermore, we observed lower baseline levels of IL-8 in p53mt tumors while high baseline IL-8 was predictive of worse PFS, indicating an antitumor effect potentially related to a smaller baseline tumor-promoting N2 TAN population 8. While the immune deconvolution analysis revealed trends for smaller myeloid cell populations at baseline in p53mt tumors, this hypothesis cannot be fully evaluated within the context of the clinical trial data. Finally, the systemic levels of the immunosuppressive IL-10 appear to increase with increasing tumor burden 46. Accordingly, in our study there was a non-significant difference in the change of IL-10 concentration on-treatment where IL-10 decreased in patients with p53mt tumors (8%) but increased in patients with p53wt tumors (25%).
Our study has several limitations. First, the number of patients with available tumor tissue at baseline and on-treatment was small, limiting the statistical strength of observed changes in the TME. The number of p53wt CRC samples with available whole transcriptome sequencing data on study was also small. To better understand the transcriptional differences associated with the TP53 mutation status, we conducted a robust bioinformatics analysis in CRC samples from TCGA dataset, specifically comparing p53wt to p53wt in the presence of wildtype RAS. This allowed for the assessment of the impact of p53 status, within a larger cohort. Additionally, the status of different immune cell populations and cytokines in peripheral blood and associated dynamic changes on-treatment may not be fully reflective of changes in the TME. In support, while PD1+ CD45RO+ CTLs decreased in peripheral blood of patients with p53mt CRC, they increased in p53mt CRC tumor tissue. Our previous analysis also revealed that the primary tumors and metastatic sites may not have a similar immune TME composition 9. Furthermore, the observed changes may be solely a pharmacodynamic effect of the anti-PD1 and/ or anti-EGFR therapy rather than biologically linked to treatment outcomes. Additionally, the transcriptomic analysis was feasible only in a small subset of patients, most of whom had p53mt tumors. The latter can limit the generalizability of the results. The mapping quality was also low (<80%) with some rRNA contamination, a common finding in FFPE samples. This may have negatively affected the gene expression estimates. Additionally, since the standard care sequencing was performed with different panels reporting only a limited set of clinically relevant genes, we decided not to evaluate the effect of co-mutation profile in time-to-event outcomes. Finally, the inclusion of patients with both right- and left-sided CRC 9 in this study introduces an additional level of heterogeneity regarding the possible effects of immune checkpoint inhibitors and anti-EGFR mAbs in the TME, plasma cytokines, and antitumor efficacy 47–50.
In conclusion, the TP53 status of the tumor was associated with improved PFS in a secondary analysis in a trial of cetuximab plus pembrolizumab in advanced CRC. Future studies evaluating immune-oncology agents in patients with MSS, RASwt CRC should include TP53 as an integrated biomarker and evaluate its performance as a predictive biomarker.

Supplementary Material
Supplementary Material