Clinicopathological and molecular characteristics of early-onset vs late-onset colorectal cancer according to tumor location

The incidence of early-onset colorectal cancer (EOCRC) is rapidly increasing worldwide in decade when screening of colorectal cancer (CRC) is more prevalent. The clinicopathological and molecular characteristics of EOCRC have not yet been clarified. This study aims to evaluate clinicopathological and molecular features among EOCRC and late-onset colorectal cancer (LOCRC) patients according to different tumor locations. We identified CRC patients from a prospectively maintained CRC database between January 2015 and December 2018. The clinicopathological and molecular characteristics including dMMR, mutation of PIK3CA, BRAF and KRAS were compared between EOCRC and LOCRC. The relationships according to different tumor locations were assessed. Totally 4468 patients were analyzed in this study. Compared to LOCRC patients, EOCRC patients were more likely to have status of dMMR (OR, 2.52; P < 0.001), regardless of tumor location. EOCRC patients were more likely to be detected with mutation of PIK3CA (OR, 1.24; P = 0.041), which only tended to exist in the left-side colon (OR, 1.51; P = 0.06), but not in the right-side colon or rectum. No significant difference was found for BRAF or KRAS mutation, but mutation of KRAS was more frequently found in left-side colon (OR, 1.34; P = 0.04) among EOCRC patients. Status of dMMR, mutation of PIK3CA, BRAF and KRAS was different between EOCRC and LOCRC patients according to different tumor locations, which implied that EOCRC might be a unique subgroup of CRC patients. Further investigations of molecular and genetic differences should be performed to help define new diagnosing and therapeutical strategies for EOCRC patients.


Introduction
Colorectal cancer (CRC) is the third most common and the second most lethal cancer worldwide [1], which has been found to have a higher incidence with increasing age. Due to benefits of CRC screening, CRC incidence has declined overall or been stable in developed countries in the last decades. In contrast, the incidence of CRC among individuals younger than 50 years of age (early-onset CRC, EOCRC) is on a significant rise [2,3]. It is estimated that around 11% of colon cancers and 18% of rectal cancers occur in individuals younger than 50 years of age [4]. Previous studies suggested that EOCRC patients might be a specific subgroup that tended to have a higher prevalence of low-grade tumor differentiation, to exhibit an advanced stage, and to present with or develop to metastatic disease [5]. However, whether EOCRC patients have a distinct biological behavior remains controversial.
Status of mismatch repair (MMR) genes and mutation of PIK3CA, BRAF and KRAS were important biomarkers to evaluate CRC. Microsatellite instability (MSI) tumors, whose carcinogenetic pathway is also known as the "Mutator Phenotype pathway", represent 10-15% of all CRCs [6]. The underlying pathogenesis may relate to presence of germ-line mutations in MMR genes, which results in MSI or deficiencies of mismatch repair genes (dMMR) [7]. dMMR could be found in 15% of sporadic CRC, mainly due to an epigenetic inactivation of MLH1 [8]. Palomba G et al. reported an overall mutation rate were 35.6% for KRAS, 2.1% for BRAF, and 14.3% for PIK3CA gene in CRC [9]. A comprehensive genetic characterization of EOCRC in a large cohort is lacking, thus the difference of dMMR status, mutation of KRAS, BRAF, PIK3CA genes between EOCRC and LOCRC patients still needs further research, as well as their difference according to different tumor locations. To fill this knowledge gap, we attempted to investigate the clinicopathological and molecular characteristics between EOCRC and LOCRC patients, and their difference according to different tumor locations, therefore stratifying CRC patients by holding up the age of onset and tumor location.

Patients' selection
This study was in accordance with the Helsinki Declaration and approved by the institutional review board of the Sixth Affiliated Hospital of Sun Yat-sen University, Guangzhou, China. Patients diagnosed with CRC who underwent surgery for colorectal adenocarcinoma from January 2015 to December 2018 were identified at the Sixth Affiliated Hospital of Sun Yat-sen University from a prospectively maintained CRC database. Exclusion criteria were: (1) patients younger than 18 or older than 90 years; (2) recurrent CRC; (3) patients with multiprimary cancer; (4) patients with substantial missing data. EOCRC patients were defined as those diagnosed with CRC under age of 50 years old in this study. Data of clinicopathological and molecular variables were collected including age, gender, body mass index, and information of tumor (location, stage, differentiation and molecular testing). After manual review, there were total 4468 patients included in this study (947 EOCRC cases and 3521 LOCRC cases). Flow chart of patient selection was shown in Fig. 1.

Definition of variables
Tumor staging was performed according to the 8th Edition of American Joint Committee on Cancer TNM staging system. Tumor differentiation was categorized as well differentiated, moderately differentiated, poorly differentiated, undifferentiated carcinoma, mucinous carcinoma, or signet-ring cell tumor. The primary tumor location site was categorized as right-side colon if the tumor was located above the splenic flexure (including cecum, ascending colon and transverse colon), and left-side colon if it was located at or below the splenic flexure and above rectum (including descending colon and sigmoid colon), and rectum. Status of MMR were examined using immunohistochemical (IHC) staining for MLH1, MSH2, MSH6, and PMS2 protein. Briefly, formalin-fixed paraffin-embedded tissues were serially cut into 4 um sections on silanized glass slides. Slides were deparaffinized with dimethylbenzene and rehydrated through graded alcohols before retrieving antigen by incubation in sodium citrate buffer. Endogenous peroxidase was blocked with hydrogen peroxide solution and then incubated with anti MLH1, MSH2, MSH6, and PMS2 at 4 °C overnight. Slides were stained with diaminobenzidine in an Envision System and counterstained hematoxylin. Cases with complete nuclear loss of MMR expression in invasive tumor cells but with retained expression in inflammatory cells and/or adjacent normal tissue as positive controls were considered MMR deficiency.
Mutation of PIK3CA, BRAF and KRAS genes were examined using polymerase chain reaction (PCR), including mutation of exon9 and exon20 of PIK3CA, BRAF V600E and exon2, exon3 and exon4 of KRAS. Sequences were listed in Table 1.

Statistical analysis
SPSS were used for data analysis. Descriptive statistics for measurements are expressed as mean value ± standard deviation, and t-tests or Wilcoxon rank sum tests were used to compare according to the data type. Survival curves were constructed using the Kaplan-Meier method. The relationships according to different tumor locations were assessed using binary logistic regression models. Results were considered statistically significant with two-sided p-value < 0.05.

Clinicopathological characteristics
A total of 4468 patients with CRC were included in this study, among whom 947 patients (21.2%) were EOCRC and 3521 patients (78.8%) were LOCRC. 2730 (61.1%) patients were male and 1738 (39.1%) were female, with a gender ratio of 1.53. Patients with EOCRC were significantly likely to have family history (P<0.001), while patients with LOCRC tended to with early stage (stage I or II, P=0.010). About one half (47.9%) of tumors were located in rectum. Data of clinicopathological characteristics of patients were shown in Table 2.  TGG TCT TGT TGT TGG CT  GGT ATG GTA AAA ACA TGC TGAGA  PIK3CA 20  CAT TTG CTC CAA ACT GAC CA  TTC CTA TGC AAT CGG TCT TTG  BRAF  V600E ATC CCA CCT CCT AAA ATA ATC AAG A TAA CCT TTT TGA AAG GGG GC  KRAS  2  TCA AGT CCT TTG CCC ATT TT  TGC ATG GCA TTA GCA AAG AC  KRAS  3  TTG TGG ACA GGT TTT GAA AGA  AGA AGC AAT GCC CTC TCA AG  KRAS  4  GTG TGA CAT GTT CTA ATA TAG TCA  GAA TGG TCC TGC ACC AGT  ; P < 0.001, respectively), but not in rectum. As mentioned above, EOCRC patients were more likely to detected with mutation of PIK3CA, which only trended to exist in left-sided colon (OR, 1.51; CI 0.98-2.33; P = 0.06), but not in right-sided colon or rectum. Overall, no significant difference was found for BRAF or KRAS mutation between EOCRC and LOCRC, but mutation of KRAS was more frequently in EOCRC than LOCRC in left-sided colon (OR, 1.34; CI 1.02-1.77; P = 0.04), especially mutation at exon 2 (OR, 1.60; CI 1.21-2.12; P = 0.001), but not in right-sided colon or rectum (Table 4).

Discussion
Mounting evidence indicates that EOCRC has unique molecular profiles, which might influence disease outcomes and response to therapy [10]. With regard to MMR status, recent studies found that dMMR was more prevalent in EOCRC were detected more frequently in EOCRC overall, but only in colon including both left side and right side rather than in rectum. Mutation of PIK3CA, BRAF and KRAS is mainly focus on CRC patients with recurrence or metastasis. They are important potential biomarkers for prognosis as well as targeted therapies [13,14]. Though their influence on therapeutic strategy in non-recurrent or non-metastatic CRC was unknown, frequency of such mutation and the association with clinicopathological variables were more and more reported in recent years [15][16][17].
In a meta-analysis of PIK3CA by Shuofei Yang, et al. forty-four studies enrolling 17,621 patients were eligible and the rate of PIK3CA mutation was 12.9% and was associated with proximal tumor location [18]. In the present study, mutations of PIK3CA was detected in 12.2% patients overall, and more likely in EOCRC than LOCRC, but not associated with tumor location significantly.
As previously reported, mutation of KRAS is an early event in the carcinogenesis of CRC and its incidence in CRC patients is about 30% to 50% [19]. In our study, the incidence of KRAS mutation was 46.8%, which was similar to those in studies of Eastern and Western populations [15]. Most previous studies reported a higher incidence of KRAS mutations in EOCRC patients, which may due to KRAS mutations usually associate with more advanced tumor stage in CRC patients [20]. Pilozzi E et al. [19] reported that the EOCRC with left-sided tumors showed KRAS mutation in 43% of cases; in contrast, in the LOCRC, KRAS mutation was observed in 14% of cases. Gunal et al. [20] showed a very high incidence of KRAS mutations in patients younger than 40 years compared with those older than 40 years (66.7% and 36%, respectively). In our study, there was no significant difference of KRAS mutations incidence between EOCRC and LOCRC patients, but mutation of KRAS was more frequently in EOCRC than LOCRC in left-sided colon (OR, 1.34; CI 1.02-1.77; P = 0.04), but not in right-sided colon or rectum.
Frequency of BRAF mutation was 10.8% in meta-analysis with about 12 thousand patients, and was found associated with advanced TNM stage, poor differentiation [21]. In our study, we didn't find a significant difference for BRAF mutation in age and tumor location.