Genomic profiles and their relationships with clinical characteristics and immune features in cervical cancer

Highlights

• •AR (78%) is the gene with the highest mutation frequency among cervical cancer patients in our cohort.
• •TMB level has significantly difference between HPV+ and HPV- cervical cancer.
• •Prognostic predictive indicators related to CESC patients, including BMI, BARD1, CEP290, and SLX4 mutation genes.
• •MATH may be an important indicator for gauging treatment response in cervical cancer.

Abstract

Objective

This study aimed to investigate the genomic alteration profiles of cervical cancer patients, examine the correlation between mutation patterns and clinical and immune attributes, and discover novel targets for treatment of individuals with cervical cancer.

Methods

We performed targeted next-generation sequencing of tumor tissues and blood samples obtained from 45 cervical cancer patients to analyze somatic alterations, mutation patterns, and HLA alleles comprehensively. Additionally, we used flow cytometry to assess expression levels of immune checkpoint genes.

Results

Notably, genes such as AR (78%), KMT2D (76%), and NOTCH1 (62%) exhibited higher mutation frequencies. Moreover, the tumor mutation burden (TMB) was significantly greater in HPV-positive cervical cancer patients than in HPV-negative patients (P=0.029). BMI (P=0.047) and mutations in BARD1 (P=0.034), CEP290 (P=4E-04), and SLX4 (P=0.0128) were identified as predictors of shorter overall survival in cervical cancer patients. Furthermore, the present study revealed significant upregulation of PD-1 (P=0.027) and Tim-3 (P=0.048) in the high mutant-allele tumor heterogeneity (MATH) cohort. In the elderly cervical cancer patient population, HLA-A03:01 emerged as a high-risk allele (OR=3.2, P<0.0001); HLA-C07:02 (OR=0.073, P=0.02) and HLA-B*07:02 (OR=0.257, P=0.037) were associated with a reduced risk among patients with low TMB.

Conclusions

This study offers insights into the mutation characteristics of cervical cancer patients and identifies potential therapeutic.

Introduction

Cervical cancer is one of the most prevalent gynecological malignancies worldwide, negatively affecting women's lives and health [1]. This insidious disease has numerous high-risk factors, including HPV infection, sexual activity at an early age, and multiple sexual partners, among others [2]. Presently, the primary treatment modality for advanced or recurrent cervical cancer involves a combination of surgical intervention and radiotherapy/chemotherapy. Regrettably, even with these aggressive treatments, the overall survival rate remains distressingly low for patients in advanced stages or with recurrent disease [3]. In recent years, the advent of targeted therapy and immunotherapy drugs has led to novel treatment modalities for cervical cancer. Leveraging next-generation sequencing (NGS) permits a thorough examination of genetic alterations, offering insights into the intricate molecular landscape of tumor development and facilitating tailoring of personalized treatment strategies. Notably, Huang et al. [4] discerned four frequently mutated genes (FAT1, MLL3, MLL2, and FADD) via whole-genome sequencing of 102 pairs of cervical cancer tissues and their normal counterparts. Furthermore, tumor mutation characteristics and immune-related biomarkers, including tumor mutation burden (TMB), microsatellite instability (MSI), and tumor-infiltrating lymphocytes (TILs), hold promise as indicators for assessing the prognosis and treatment response of cancer patients [5]. Research findings have consistently demonstrated that colorectal cancer patients with mismatch repair deficiency (dMMR) exhibit heightened sensitivity to both radiotherapy and chemotherapy [6]. Additionally, a high tumor mutation burden (TMB) is a promising predictive marker for favorable responses to immune checkpoint inhibitors, particularly in cervical squamous cell carcinoma patients [7]. Notably, based on the outcomes of the KEYNOTE-158 clinical trial, pembrolizumab has received approval for treatment of recurrent or metastatic cervical cancer patients, specifically those with a positive PD-L1 status [8]. The FDA has also extended its approval to include use of pembrolizumab for patients with PD-L1 expression, a TMB ≥10 mutations per megabase (Mb), and high microsatellite instability (MSI-H) [9,10].

Tumor development is intricately linked to genetic mutations, yet research on the distinct pathways involved in carcinogenesis in the context of HPV infection is relatively rare. While mutations associated with tumor development in HPV-infected cancers share some similarities with those observed in HPV-negative tumors, they are not entirely identical. Notably, research has revealed that the mutation frequencies of two renowned tumor-suppressor genes, TP53 and PTEN, are markedly lower in HPV-positive cervical cancer patients than in their HPV-negative counterparts [11]. Moreover, the degree of immune cell infiltration varies across different HPV infection groups and is characterized by heightened infiltration of B cells and CD8+Tcm in HPV-positive cervical cancer patients [11]. Elucidating the intricate interplay between HPV infection, genomic alterations, and the immune microenvironment is pivotal for comprehending the pathogenesis of cervical cancer more comprehensively.

The primary objective of this study was to conduct a comprehensive analysis of the mutational landscape and clinical attributes of cervical cancer patients. To achieve this goal, we retrospectively gathered clinical data for a cohort of 45 cervical cancer patients. By employing next-generation sequencing technology, we conducted targeted sequencing of both tumor tissues and corresponding blood samples. The sequencing panel included 764 well-validated genes associated with tumorigenesis. Our analysis encompassed delineation of genomic alterations, mutation profiles, and immunological characteristics within the cervical cancer cohort. Furthermore, we meticulously examined the correlations between mutation patterns, clinical parameters, and immune attributes among cervical cancer patients. This concerted effort aimed to provide compelling evidence elucidating the molecular mechanisms underpinning cervical cancer development and, in turn, to offer fresh perspectives for precision-oriented cervical cancer treatment strategies.

Materials and methods

Clinical samples

We carefully selected cervical cancer patients who underwent treatment at the Affiliated Tumor Hospital of Xinjiang Medical University between February 2017 and October 2019. All these patients received a definitive pathological diagnosis of cervical cancer; 45 individuals were included in our study. Formalin-fixed paraffin-embedded (FFPE) specimens from patients with cervical cancer, along with matched peripheral blood samples collected concurrently, were acquired from these 45 patients. It is important to note that this study was conducted in strict adherence to the ethical principles outlined in the Helsinki Declaration and received full approval from the Ethics Committee of the Affiliated Tumor Hospital of Xinjiang Medical University (Approval No: K-2,021,005). The inclusion criteria for participation in this study were as follows: (1) confirmed diagnosis of cervical cancer through histopathological examination at our hospital; (2) received treatment at our hospital; and (3) willing to provide signed informed consent and furnish complete clinical data, along with a commitment to cooperate with follow-up procedures prior to enrollment.

The exclusion criteria were as follows: (1) history of prior surgery, radiotherapy, chemotherapy, or other forms of treatment; (2) diagnosed with malignancies other than cervical cancer; (3) presented with severe immune system disorders or hematological diseases; and (4) severe heart, liver, lung, or other significant medical conditions.

Data acquisition

Somatic mutation data for cervical squamous cell carcinoma patients were retrieved from the TCGA database (available at https://portal.gdc.cancer.gov) and subsequently preprocessed with the R package ‘maftools’.

DNA extraction, library construction, and targeted next-generation sequencing

All samples used in this study were processed at the second-generation sequencing laboratory (Zhenhe Technology Co., Ltd., Beijing). DNA was extracted from both tumor tissues and plasma-free DNA using DNA extraction kits. Subsequently, the DNA concentration was quantified, and the integrity of the genomic DNA was assessed. The genomic DNA was then randomly fragmented into pieces approximately 200–300 base pairs in size using an ultrasonic disruptor. After confirming that the size of the fragmented DNA met the specified requirements, end repair and A-tailing procedures were carried out on the fragmented DNA, followed by the addition of molecular index adapters. The resulting DNA was subsequently purified, and a 1 μL sample of the purified supernatant was used to determine the DNA concentration. The number of amplification cycles was determined based on the DNA concentration. Following amplification, an equivalent quantity of magnetic beads was added for purification, effectively removing fragments of varying sizes. An additional 1 μL of the purified supernatant was utilized to assess the DNA concentration, while the remaining 50 μL of clear liquid was transferred to a new centrifuge tube and stored at −20°C either as a library or for direct use in subsequent hybridization experiments. Gene capture probes were introduced to the DNA library samples using a specialized kit, facilitating hybridization of target DNA, which was then incubated at 47 °C for a duration of 64–72 h. Following purification, another 1μL of the purified supernatant was used to determine the library concentration. Finally, the library was subjected to sequencing using the NovaSeq platform in accordance with Illumina's standard protocol.

Sequencing data and mutation analysis

The raw sequencing data, derived from 90 samples (comprising 45 paired tissue and peripheral blood samples), were aligned to human reference genome hg19 (GRCh37) using BWA-MEM with default parameters. All resulting mappings were consolidated into a single concatenated reference sequence. Subsequently, Fastp software (v0.20.0) was used to eliminate reads with quality scores (Q20) less than 90%, and filtering was applied based on the minimum Phred quality score (MapQ) set by the read mappings, removing reads with inadequate mapping quality (MapQ<5). To increase alignment accuracy, Genome Analysis Toolkit (GATK, v4.1.8) was utilized to perform local realignment of reads around insertions and deletions near the alignment results and rectify alignment errors introduced by insertions and deletions. Following this, base quality recalibration was executed. By utilizing the recalibrated BAM files, the Mutect2 module of GATK was used to analyze mutations within the paired blood and tumor samples, with the goal of discerning somatic mutations specific to the tumor tissue. Subsequently, the FilterMutectCalls module was used to refine the mutation sites by screening and excluding mutations identified as contaminants, germline mutations, or those originating from the blood control sample. The mutational signatures of 45 cervical cancer samples were analyzed using the ‘deconstructSigs’ package.

Calculation of the APOBEC enrichment score

Members of the APOBEC family primarily deaminate cytidines located in a TCW motif, where W represents either an A or T nucleotide. The APOBEC mutagenesis signature consists of two mutation patterns within this motif: TCW to TTW and TCW to TGW. The formula for computing the APOBEC enrichment score is as follows [12]:

$$\text{APOBEC}\text{Enrichmen}{\mathrm{t}}_{\text{TCW}}=\frac{\text{Mu}{\mathrm{t}}_{\text{TCW}}/\text{Co}{\mathrm{n}}_{\text{TCW}}}{\text{Mu}{\mathrm{t}}_{\mathrm{C}}/\text{Co}{\mathrm{n}}_{\mathrm{C}}}$$

Mut_TCW represents the total number of mutated cytidines within the TCW motif in a 41-base pair window. Mut_C represents the total number of mutated cytidines in the 41-base pair window. Con_TCW and Con_C represent the total number of TCW motifs and cytidines, respectively, in the 41-base pair window. One-sided Fisher's exact test was performed to calculate statistical significance.

Driver genes and GO and KEGG enrichment analyses

Following sample clustering analysis, group-level statistics were calculated. OncoDock FML software, which employs gene function algorithms, was used to calculate driver genes within each group as well as across all samples independently. Driver genes meeting the criterion of a false discovery rate (FDR) cutoff<0.01 were carefully selected. The identified genes were then subjected to Gene Ontology (GO) analysis using the ‘enrichGO’ function within the R package ‘clusterProfiler’, and Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis was performed using the ‘enrichKEGG’ function. Visual representation of the GO and KEGG results was created using the R packages ‘enrichplot’ and ‘ggplot2’. Additionally, differences in signaling pathways between the various groups were thoroughly analyzed and compared.

Genomic characterization

The number of nonsynonymous or missense mutations identified in the tumor samples (denoted as Nm) was tallied. Analysis of the aligned BAM files allowed us to determine the total bases within the exonic regions of the sequencing data (designated Ne). To compute the TMB (tumor mutation burden) value, we divided Nm by Ne. Typically, TMB values are expressed as mutations per megabase (mut/Mb).

For comparison, we employed the blood leukocytes as normal samples. By using the alignment files, we calculated the number of reads that completely covered microsatellite loci and assessed the length of the repeat units. The repeat unit type and count were determined for each locus. To establish a baseline for normal samples, we calculated the number of repeat unit types across all normal samples and derived the mean (M) and standard deviation (SD). Subsequently, we determined the number of repeat unit types (N) for each microsatellite locus within the test sample. Loci were categorized as unstable if N was greater than or equal to M plus n times the standard deviation (SD). Ultimately, we counted the total number of unstable loci to determine the MSI (microsatellite instability) status of the sample.

The MAF (mutation-allele fraction) for each variant site was computed based on sequencing data. Subsequently, the MAD (median absolute deviation) value is derived from these MAF values. The absolute difference between each MAF value and the median was calculated by multiplying the median absolute difference by a constant (1.4826). The MAD value is obtained by dividing the result by the median of the MAF values and multiplying it by 100. The formula for calculating the score is as follows:

$$\text{MATH}\text{score}=\frac{1.4826*\text{median}\text{absolute}\text{deviation}\left(\text{MAD}\right)}{\text{the}\text{median}\text{of}\text{the}\text{MAF}}*100$$

Distinguishing high and low MATH groups based on quartiles.

Detection of HPV

Fluorescence PCR was employed to analyze HPV types in 45 patients using a Human Papillomavirus (HPV) nucleic acid typing detection kit (manufactured by Guangdong Hybribio Biotech Co., Ltd., Guangdong, China). This kit encompasses a panel of 15 high-risk HPV types (HPV16, 18, 31, 33, 35, 39, 45, 51, 52, 53, 56, 58, 59, 66, and 68) and six low-risk HPV types (HPV6, 11, 42, 43, 44, and CP8304). The specific procedures were conducted in strict adherence to the manufacturer's protocol.

HLA genotype detection

Genomic DNA samples for HLA typing were meticulously extracted from peripheral blood samples that had been preserved with EDTA anticoagulant via a salt extraction method. To capture and enrich the target fragments from the library, HLA-specific primers were used, followed by high-throughput sequencing of the enriched HLA target fragments. The sequencing depth was designed to comprehensively encompass the genomic regions associated with the HLA-A, HLA-B, and HLA-C alleles. Subsequently, a comprehensive analysis of the sequencing results was performed using a combination of NextGENe software and the GenBank database.

Flow cytometry

Peripheral blood samples (5mL each) were collected from 45 cervical cancer patients, from which peripheral blood mononuclear cells (PBMCs) were isolated via density gradient centrifugation using fresh EDTA-containing anticoagulated blood. Tissue samples were obtained from these same patients via biopsy forceps and subsequently cut into fragments measuring 1–3 mm in size using sterile surgical scissors. These tissue fragments were then placed in tubes containing RPMI 1640. Next, the tissue was processed to yield a single-cell suspension, which was washed twice with R10. Flow cytometry was used to assess expression of PD-1 and Tim-3 on CD4+ and CD8+T cells in both peripheral blood and tumor tissue samples collected from 45 cervical cancer patients. The flow cytometry data were analyzed using BD FACS Diva flow cytometry analysis software.

Statistics

All statistical analyses were carried out using SPSS 18.0 software. Groupwise comparisons of rates were conducted using the chi-square test, while comparisons of means between samples were analyzed through the t-test and one-way analysis of variance (ANOVA). To evaluate the correlation between continuous variables, the Pearson correlation coefficient was used. A P value<0.05 was considered to indicate statistical significance.

Results

Patient cohort

A total of 45 FFPE samples were collected from patients diagnosed with advanced cervical squamous cell carcinoma, as detailed in Table 1. The median age of the patients was 56 years, and 22 individuals received concurrent chemoradiotherapy as part of their treatment regimen. Among these patients, 40 (88.9%) tested positive for HPV infection, with 34 (85%) cases being associated with high-risk HPV types. Among the high-risk types, 32 (94%) had HPV16, 1 (3%) had HPV18, and 1 (3%) had dual infections involving both HPV16 and HPV18.

Table 1.

Basic characteristics of cervical cancer patients.

Clinical characteristics	Cohort(n = 45)	Proportion (%)
Age (years)
<56 years	25	55.6
56 years	20	44.4
range	37-78
BMI
<24	17	37.8
24	28	62.2
Abortion
No	19	42.2
Yes	26	57.8
Pathology type
PDSCC	14	31.1
MDSCC	24	53.3
Other	7	15.6
FIGO stage
II	16	35.6
III	28	62.2
IV	1	2.2
Histology type
Florid	14	31.1
Nodular	22	48.9
Cavity	7	15.6
Other	2	4.4
HPV infection
No	4	8.9
Yes	40	88.9
CEA (ug/L)
<5	32	71.1
5	13	28.9
SCC (ug/L)
<1.5	9	20
1.5	36	80
Pelvic lymph node
negative	27	60
positive	18	40
Concurrent chemoradiotherapy
No	23	51.1
Yes	22	48.9
TMB
TMB-High	25	55.6
TMB-Low	20	44.4
MSI status
MSI-High	2	4.4
MSS	42	93.3
MATH status
MATH-High	12	26.7
MATH-Low	33	73.3

Next-generation sequencing revealed that 25 patients (55.6%) exhibited a high TMB (TMB-High) and 20 patients (44.4%) a low TMB (TMB-Low). Two patients displayed high microsatellite instability (MSI-High), whereas 42 patients demonstrated microsatellite stability (MSS). Another patient had an indeterminate MSI status. Notably, both tumors identified as MSI-High were also categorized as TMB-High and displayed elevated tumor heterogeneity, as indicated by Mutant-Allele Tumor Heterogeneity (MATH) analysis. Among the patient cohort, 12 individuals exhibited high MATH scores, signifying pronounced tumor heterogeneity; 33 individuals exhibited low MATH scores.

Genomic profiles of cervical cancer

Somatic single-nucleotide variations (SNVs) and insertions/deletions (INDELs) were analyzed across 45 samples, followed by a comprehensive statistical assessment. Initially, a total of 12,601 SNV sites were identified; these sites were subsequently refined by eliminating sites within 3′ and 5′ untranslated regions (UTRs), flanking regions, intergenic regions (IGRs), intronic regions, RNA-coding regions, silent synonymous mutations, and splice regions. This curation resulted in a final set of 3894 sites, as visualized in Fig. 1A. The prevailing point mutation observed was C>A, with notable involvement of genes such as AR (78%), KMT2D (76%), NOTCH1 (62%), ARID1B (58%), and ARID1A (56%). AR predominantly exhibited missense mutations, in-frame deletions, frameshift deletions, and mixed mutations, while KMT2D displayed missense mutations, in-frame deletions, and mixed mutations as its primary mutation patterns. On the other hand, NOTCH1 was characterized by missense mutations and mixed mutations, as illustrated in Fig. 1B. For further insights, bioinformatics analysis was performed on the sequencing results for 45 cervical cancer patients, revealing interacting genes. A total of 105 mutually exclusive or coexistent mutated gene pairs were examined, 34 of which exhibited significant differences (P<0.05), as demonstrated in Fig. 1C. Notable interactions were detected between genes such as NOTCH3 and GNAS, IRS2 and KMT2D, and IRS2 and GNAS (P<0.01). The three pathways with the highest mutation frequencies were RTK-RAS (95.6%), PI3K (86.7%), and NOTCH (86.7%), as shown in Fig. 1D. Notably, the most frequently mutated site across all the samples was p.E545K of PIK3CA (46.67%), a well-recognized hotspot mutation site, as shown in Fig. 1E. To further analyze the influence of APOBEC (apolipoprotein B mRNA editing enzyme, catalytic polypeptide-like) enrichment, we employed the R package ‘maftools’ and compared the APOBEC-enriched group (comprising 5 patients) to the non-APOBEC-enriched group (comprising 40 patients) (Fig. 1F). Interestingly, three genes (FAT1, PLCG2, and PMS2) exhibited significantly greater mutation frequencies in the APOBEC-enriched group than in the non-APOBEC-enriched group (P < 0.01), as denoted in Fig. 1G.

Fig. 1.

Genomic features of cervical cancer samples. A: Description of somatic mutations in 45 cervical cancer patients. B: Distribution of mutated genes in the 45 samples. C: Analysis of mutual exclusivity and co-occurrence of mutated genes in the 45 samples. D: Signaling pathways enriched with mutated genes. E: Most frequently mutated site in PIK3CA identified in this cohort. F: Enrichment analysis of APOBEC in cervical cancer. G: Genes with significant differences in mutation frequency between the APOBEC-enriched and non-APOBEC-enriched groups in the cohort.

A comprehensive analysis of driver genes was conducted on the 45 cervical cancer samples, and genes displaying a false discovery rate (FDR) of less than 0.05 were carefully identified. In total, 14 driver genes, namely, MPL, IARS, ATG3, FLCN, HAUS6, KIT, CEP290, JUN, HLA-A, E2F3, MAPK, TNFRSF14, HLA-B, and PIK3CA, were discerned, as illustrated in Fig. 2A. To determine the functional significance of these driver genes, we performed Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses. The results are presented in Fig. 2B, C. These driver genes are prominently enriched in pathways linked to tumor development, the immune response, and viral infection. Fig. 2D shows the contribution proportion of known SBS signatures in each sample.

Fig. 2.

Analysis of driver genes and mutational signatures. A: Analysis of 14 driver genes identified using the oncodrive function of the MutSigCV and maftools packages. B, C: Enrichment analysis of Gene Ontology and KEGG pathway enrichment data. The size of the dots represents the number of genes in the pathway, and the corresponding x-axis represents the percentage of genes in the pathway compared to the total number of genes. The color indicates the significance of the gene enrichment. D: SBS signatures of cervical cancer patients in the cohort.

Comparison of HPV-positive and -negative cervical cancer patients

In Supplementary Table S1, we present a comparison of clinical and pathological characteristics, as well as mutational features, among a cohort of 45 cervical cancer patients. Within this patient group, 40 patients were HPV positive and 4 HPV negative; 1 patient remained untested for HPV status. Notably, no discernible differences in clinical characteristics were observed between the positive and negative groups. However, in the HPV-negative subgroup, both the tumor mutation burden (TMB) and variant-allele heterogeneity were found to be notably low, indicative of genetic stability, and microsatellites also exhibited stability. Intriguingly, compared with that in the HPV-negative group, the TMB in the HPV-positive group was significantly greater (P=0.029).

Subsequently, we conducted a thorough analysis to determine genetic differences between HPV-positive and HPV-negative cervical cancer patients. Among the HPV-positive patients, the five genes with the highest mutation frequencies were AR (80%), KMT2D (78%), NOTCH1 (65%), ARID1A (60%), and ARID1B (60%), as depicted in Supplementary Fig. S1A. The predominant pathways observed within this group included RTK-RAS (38/40), PI3K (36/40), NOTCH (36/40), Hippo (28/40), and Wnt (27/40), as illustrated in Supplementary Fig. S1B.

In the HPV-negative group, the five genes with the highest mutation frequencies were AXIN1 (75%), GNAS (75%), AR (50%), ASXL2 (50%), and ATG3 (50%), as indicated in Supplementary Fig. S1C. Similarly, the primary pathway enrichment in this group was RTK-RAS (4/4), followed by PI3K (3/4), NOTCH (3/4), Wnt (3/4), and the cell cycle (3/4), as shown in Supplementary Fig. S1D. Notably, there was significant overlap in mutation pathways between the two groups, indicating similarities in the underlying genetic alterations.

Prognostic factors for cervical cancer

Relationships between clinical and pathological characteristics, mutation features, overall survival (OS) and progression-free interval (PFI) in cervical cancer patients were explored through univariate and multivariate Cox regression analyses. The results of the univariate Cox regression analysis showed that BMI was associated with shorter overall survival (OS) and that carcinoembryonic antigen (CEA) was associated with both shorter OS and a shorter progression-free interval (PFI) (Table 2). Factors with P< 0.1in the univariate Cox regression analysis were included in the subsequent multivariate regression analysis. The results showed that BMI was an independent predictor of shorter OS in cervical cancer patients (P=0.047) (Table 2).

Table 2.

Clinical and mutational features with prognostic relevance.

	OS				PFI
	Univariate		Multivariate		Univariate		Multivariate
	HR (95 % CI)	P	HR (95 % CI)	P	HR (95 % CI)	P	HR (95 % CI)	P
Age	0.786	0.673			0.617	0.495
56 vs. <56	(0.257-2.405)				(0.154-2.468)
BMI	0.231	0.015	0.291	0.047	0.255	0.054	0.300	0.095
24 vs. <24	(0.071-0.754)		(0.071-0.754)		(0.064-1.022)		(0.073-1.234)
FIGO stage	53.887	0.087	381084.664	0.956	7.838	0.056	5.596	0.128
II vs. III-IV	(0.563-5159.931)		(0.000-5.525E+201)		(0.950-64.654)		(0.609-51.443)
HPV infection	0.441	0.289			0.600	0.631
Yes vs. No	(0.097–2.003)				(0.075–4.824)
CEA	5.469	0.004	2.564	0.134	5.163	0.019	3.139	0.147
≥5 vs. <5	(1.730–17.285)		(0.748–8.787)		(1.305–20.420)		(0.668–14.740)
SCC	0.726	0.626			27.002	0.386
51. vs. <1.5	(0.200-2.639)				(0.016-46499.983)
pelvic lymph node	1.325	0.614			2.019	0.297
Positive vs. Negative	(0.444-3.956)				(0.539-7.562)
Concurrent chemoradiotherapy	0.453	0.188			0.280	0.113
Yes vs. No	(0.139-1.471)				(0.058-1.351)
TMB	0.572	0.316			0.792	0.728
High vs. Low	(0.192-1.705)				(0.212-2.952)
MSI	0.046	0.594			0.046	0.659
High vs. MSS	(0.000-3825.189)				(0.000-41255.759)
MATH	0.400	0.233			0.254	0.197
High vs. Low	(0.088-1.807)				(0.032-2.040)

We further analyzed the prognostic significance of genetic alterations on overall survival (OS) and progression-free interval (PFI) in cervical cancer patients. Initially, we performed a univariate Cox regression analysis to identify mutation-associated genes linked to prognosis. The detailed findings are presented in Supplementary Table S2, where we identified a total of 17 mutation-related genes that were significantly correlated with patient survival.

Subsequently, a rigorous multivariate Cox regression analysis was conducted, revealing that BARD1 (HR=2.12E-03, P=0.036), CEP290 (HR=9.53E+02, P<0.001), and SLX4 (HR=1.34E+01, P=0.046) were found to be independent predictive factors influencing the prognosis of cervical cancer patients.

Relationships between mutation characteristics and the immune landscape in cervical cancer patients

In our study, we conducted a comprehensive analysis to examine associations between tumor mutational burden (TMB), microsatellite instability (MSI), mutant-allele tumor heterogeneity (MATH) and the immune microenvironment.

To assess the immune microenvironment, we employed flow cytometry to determine proportions of CD4+T cells and CD8+T cells among PBMCs and tumor-infiltrating lymphocytes (TILs). Additionally, we evaluated expression levels of the immune checkpoint markers PD-1 and Tim-3.

The outcomes of our investigation indicated no statistically significant differences in immune cell infiltration or in the expression of PD-1 or Tim-3 within PBMCs across various groups categorized by TMB, MSI, or MATH status, as shown in Fig. 3B, D, and F. For a more comprehensive depiction of immune cell infiltration and expression of PD-1 and Tim-3, please refer to Fig. 3A, C, and E. However, a notable finding emerged when we assessed TILs. Specifically, our results revealed significantly greater expression levels of the inhibitory receptors PD-1 and Tim-3 in the high-MATH subgroup than in the low-MATH subgroup (P=0.027; P=0.048), shedding light on a potential link between MATH status and immune checkpoint regulation.

Fig. 3.

Relationships between mutation characteristics and immune cell infiltration and immune checkpoint expression. Relationships between TMB (A), MSI (C), and MATH (E) and numbers of CD4+T cells, CD8+T cells, PD-1, and Tim-3+ T cells in TILs and between TMB (B), MSI (D), and MATH (F) and numbers of CD4+T cells, CD8+T cells, PD-1 and Tim-3+T cells in PBMCs.

HLA allele analysis

We proceeded to conduct a detailed analysis of HLA typing variations among distinct mutation characteristic groups. HLA class I allele typing outcomes for the 45 patients revealed a total of 114 genotypes for the HLA-A, HLA-B, and HLA-C loci. The most prevalent allele detected was HLA-A03:01 (29.6%) (Fig. 4A), followed by HLA-C07:02 (11.7%) (Fig. 4C) and HLA-B*07:02 (10.9%) (Fig. 4B).

Fig. 4.

Analysis of HLA class I allele genes. A: Genotype distribution of HLA-A (A), HLA-B (B), and HLA-C (C) alleles in cervical cancer samples. D: Forest plot showing that HLA-A03:01 is a significantly positively correlating factor in elderly cervical cancer patients. E: HLA-C07:02 and HLA-B*07:02 were found to be protective factors in low-TMB cervical cancer patients.

We then delved into the relationship between the 114 HLA genotypes and the clinical and mutational characteristics of cervical cancer patients. Fig. 4D shows a significant difference between the older and younger age groups for HLA-A03:01, with a greater risk association observed in older cervical cancer patients (OR=3.2, 95%CI=1.835–5.869; P<0.0001). Conversely, HLA-C07:02 and HLA-B*07:02 were linked to a lower risk of CC in patients with a low TMB (Fig. 4E) (OR=0.073, 95%CI=0.009–0.586, P=0.02; OR=0.257, 95%CI=0.067–0.984, P=0.037).

Differences in mutational features between our cohort and the TCGA cohort

According to the TCGA database, the primary type of point mutation detected among 289 cervical cancer patients is predominantly C>T (Supplementary Fig. S2A). The notably mutated genes include TTN (29%), PIK3CA (28%), KMT2C (19%), MUC16 (17%), and KMT2D (13%), among others (Supplementary Fig. S2B). The dominant point mutation type in our cohort of cervical cancer patients was C>A, and missense mutations were the most common mutation types in both cohorts. Within our cohort's top 20 genes according to mutation frequency, the mutation frequencies of KMT2D (76%vs.13 %), KMT2C (49%vs.19 %), and PIK3CA (47%vs.28 %) were significantly greater than those in the TCGA-CESC cohort. We conducted an in-depth analysis of the different mutated genes, including AR, ARID1B, and KMT2D, between our study cohort and the TCGA-CESC cohort; the results are shown in Supplementary Fig. S3.

In the TCGA database, the mutated genes among cervical cancer patients were enriched in the RTK/RAS (161/289), NOTCH (152/289), and PI3K (136/289) signaling pathways (Supplementary Fig. S2C). This discovery aligns with the findings for our study cohort. Further bioinformatics analysis of TCGA-CESC mutation data revealed substantial interactions involving PIK3CA and MUC5B, PIK3CA and MDN1, and MUC16 and MDN1 (P<0.01), as illustrated in Supplementary Fig. S2D.

Discussion

This study used samples from 45 cervical cancer patients for second-generation targeted sequencing. In this comprehensive analysis, we identified a total of 3016 somatic mutation sites, with the most frequently mutated site being PIK3CA p.E545K, a well-known hotspot mutation site in tumors, consistent with previous research [13]. PIK3CA mutations correlate with uncontrolled tumor growth, proliferation, and invasion [14]. Surprisingly, AR (78%) exhibited the highest mutation frequency in this study, which has not been previously reported. Targeted sequencing of 419 cervical cancer patients from 18 centers in 7 European Union countries indicated that 7% of patients harbored amplifications, mutations, or frameshift deletions in the AR gene [15]. Currently, there is no specific evidence indicating a significant correlation between AR gene mutations and cervical cancer development. Thus, we hypothesize that this mutation might be region-specific and influenced by the unique genetic background of the region. We identified a total of 14 driver genes that were predominantly associated with viral infection, the immune response, and tumor-related pathways, underscoring their pivotal roles in cervical cancer. These driver genes might serve as targets for targeted therapies.

Our study revealed that the mutated genes within our cohort are predominantly involved in the RTK-RAS (95.6%), PI3K (86.7%), NOTCH (86.7%), Wnt (68.9%), and Hippo (66.7%) pathways, suggesting their potential contributions to cervical cancer development and progression [16], [17], [18], [19], [20]. Notably, these frequently mutated pathways identified within our cohort were consistent with those exhibiting high mutation frequencies within the TCGA database.

We identified the prognostic significance of BARD1, CEP290, and SLX4 gene mutations, which are linked to shorter overall survival. To validate our results, we conducted an analysis using somatic mutation and survival data from the TCGA database. The results of the K‒M survival analysis were consistent with our results, indicating an association between mutations in BARD1, CEP290, and SLX4 and poor prognosis in cervical cancer patients. However, there are only 1 and 2 patients with mutations in BARD1 and CEP290, respectively, in the TCGA database. Therefore, additional samples are needed to confirm our findings. Genetic and epigenetic changes in BARD1 might be poor prognostic factors for breast and ovarian cancers [21]. Furthermore, mutations in BARD1 have been associated with cervical cancer susceptibility [22]. However, research on the prognosis of cervical cancer patients with BARD1 mutations is still lacking. CEP290 mutations have been associated with various ciliopathies [23], but their role in tumors has not yet been well explored. SLX4 mutations have been identified in gastric-type endocervical adenocarcinoma and are associated with DNA damage repair [24]. The specific relationships between BARD1, CEP290 and SLX4 mutations and cervical cancer necessitate further exploration.

By comparing mutational characteristics between HPV-positive and HPV-negative cervical cancer patients, we observed that HPV-positive cervical cancer patients exhibited a significantly greater tumor mutational burden (TMB) than HPV-negative patients (P = 0.029). The TMB is often used as a surrogate marker for neoantigens, and high-TMB tumors tend to feature more tumor-infiltrating T cells and stronger antitumor immune responses [25]. MSI status has been shown to predict the response to immune checkpoint inhibitors (ICIs) [26] and has been identified as a marker for pembrolizumab treatment. While our study revealed that both of the identified MSI-H patients were HPV positive, no statistically significant difference in MSI status was noted between the HPV-positive and -negative patients. Our results indicated that the enriched pathways of the mutated genes generally resembled one another in the HPV-positive and HPV-negative groups. However, these results may be influenced by the small sample size, rendering them insufficient to reach statistical significance. A study focused on gynecologic tumors demonstrated that HPV-driven reproductive tract tumors, including cervical cancer tumors, often exhibit increased PD-L1 expression, MSI-H status, and high TMB [27]. Although various HPV-positive tumors have improved survival rates and treatment responses compared to HPV-negative tumors [28], [29], [30], the predictive value of HPV for immune therapy has yet to be established.

Genomic characteristics and the immune landscape play pivotal roles in the diagnosis and treatment of cervical cancer, as they help to identify vital biomarkers facilitating personalized therapy. Tumor heterogeneity, as assessed by the mutant-allele tumor heterogeneity (MATH) score, represents variability in behavior among individual patients or even within individual tumors. However, the relationship between MATH and immune cells remains a subject of debate. This study examined differences in expression of the immune checkpoint molecules PD-1 and Tim-3 between the high-MATH group and the low-MATH group. We discovered that PD-1 and Tim-3 levels were greater in tumor-infiltrating lymphocytes (TILs) in the high-MATH subgroup (P = 0.027, P = 0.048). However, the association between MATH and the immune landscape in cervical cancer remains underexplored. Several studies suggest that high MATH is associated with reduced immune cell infiltration and decreased activation of the immune response, a relationship that has been established in breast cancer [31]. MATH also has implications for treatment. For instance, patients with head and neck squamous cell carcinoma featuring high MATH tend to benefit from postoperative adjuvant radiotherapy, while the addition of adjuvant chemotherapy does not significantly enhance treatment outcomes [32]. In conclusion, our data suggest that cervical tumors exhibiting high MATH correlate with reduced antitumor immune activity. The role of MATH in cervical cancer prognosis and treatment response requires further investigation.

Studies have indicated that HLA-I allele polymorphisms contribute to the presentation of a diverse array of tumor antigens to T cells, a critical factor influencing responses to immune checkpoint inhibitors (ICIs) in cancer patients [33]. HLA-C07:02 serves as an inhibitory ligand for the killer cell immunoglobulin-like receptor (KIR) 2DL2/3, dampening the cytotoxicity mediated by natural killer (NK) cells [34]. Chen et al. [35] established an association between HLA-C07:02 and the risk of cervical intraepithelial neoplasia grade 3 (CIN3). Our results suggest a positive correlation between tumor mutation burden (TMB) in cervical cancer patients and the prevalence of the HLA-C07:02 allele. However, further research is essential for elucidating the relationship between HLA-C07:02 and cervical cancer.

This study has certain limitations, notably, its small sample size and its single-center nature, potentially leading to a lack of statistical significance. Additionally, all cervical cancer tissues we examined were from patients with squamous cell carcinoma, and no comparative analysis was performed between different histological subtypes.

In summary, our study provides a comprehensive analysis of the mutation characteristics of cervical cancer patients. The notably elevated frequency of AR gene mutations serves as a unique regional genetic feature. A total of 12 driver genes were identified that are primarily associated with viral infection, the immune response, and tumor-related pathways, emphasizing their crucial roles in cervical cancer. These driver genes hold promise as potential targets for targeted therapies. By comparing mutational characteristics, we observed that HPV-positive cervical cancer patients exhibited a significantly greater tumor mutational burden (TMB) than HPV-negative patients. Additionally, high MATH was associated with elevated expression levels of the immune checkpoint receptors PD-1 and Tim-3. Frequencies of the HLA-C07:02 and HLA-B07:02 alleles were significantly greater in the high-TMB group. Our research findings will help to elucidate the molecular mechanisms involved in cervical cancer and offer fresh insights into the identification of effective therapeutic targets for this disease.

Data availability statement

The sequence data that support the findings of this study have been deposited in NCBI Sequence Read Archive (SRA) with the accession codes “PRJNA1077449”.

Ethics statement

This study was approved by the Ethics Committee of the Affiliated Tumor Hospital of Xinjiang Medical University on March 4,2021 (No: K-2021005).

Fundings

This study was supported by the Key Research and Development Program of Xinjiang Uygur Autonomous Region of China (2022B03019-5), Shanghai Cooperation Organization Science and Technology Partnership Program and International Science and Technology Cooperation Program (2020E01056) and Postgraduate research innovation project of Xinjiang Uygur Autonomous Region (XJ2023G180).

CRediT authorship contribution statement

Zinan Lu: Writing – review & editing, Writing – original draft, Software, Funding acquisition, Formal analysis, Data curation, Conceptualization. Peiwen Fan: Project administration, Methodology, Investigation, Formal analysis, Conceptualization. Wen Huo: Visualization, Validation, Software, Resources, Methodology, Data curation. Yaning Feng: Validation, Software, Project administration, Funding acquisition, Conceptualization. Ruozheng Wang: Visualization, Supervision, Resources, Project administration, Investigation, Funding acquisition.

Declaration of competing interest

The authors declare that they have no competing interests.