Results

model for survival analysis with the coxph function implemented in the R package. PFS

in this study was defined as the period from the date of diagnosis to the date of

radiographically confirmed progression of disease or death during first-line therapy. OS

in this study was defined as the time from diagnosis to death. PFI was specifically used

in TCGA and GSE39582 datasets to evaluate clinical outcomes^96,98. PFI was defined as

the time from the date of diagnosis to the date of the first evidence of disease

progression, locoregional recurrence, new primary tumor, distant metastasis, or death.

Only variables with p < 0.1 were included in the multivariate analysis. We used

Benjamini and Hochberg multiple-testing correction method to adjust multiple tests.

The statistical analyses and visualization in this study were performed in an R

environment (version 3.6.1).

in Table 8. Briefly, the median age of patients was 51 years old; among these patients,

79.6% had low-grade tumors, 63% had right-sided primary tumors, and 42% had

peritoneal metastases. No patients had concomitant BRAF and KRAS mutations. Two

patients did not receive anti-cancer therapy after surgery and thus were not included in

the PFS analysis; their median PFS was 5.1 months, and their median OS was 16.8

months (Figure 3).

After profiling the transcriptome of all tumors, the association between immune

cell subtypes and patient survivals revealed that CD4+ T cell and B cell signatures were

significantly correlated with longer PFS (Hazard ratio, HR= 0. 61, p = 0.005 for CD4+

T cell and HR = 0.72, p = 0.039 for B cell), and no immune cell types were significantly

associated with patient OS (Table 9). In the analyses of the association between immune

gene expression and patient survival, we found some checkpoints responsible for

linking B cell and T cell signaling, such as ICOS (inducible T cell costimulatory) (PFS,

HR = 0.0.71, p = 0.035), ICOSLG (inducible T cell costimulatory ligand) (OS, HR =

0.69, p = 0.034) and CD40LG (CD40 ligand) (PFS, HR = 0.62, p = 0.010), were

associated with longer survival time. In contrast, there was no association between

CD40, CD80, or CD86 and prognosis (Table 10). Besides, there were prognostic

implications in the expression of complement genes. The key findings include that

tumors with higher expression of CR1, CD59, C1QA, and MBL2 were significantly

associated with shorter OS than those with lower expression of these genes. Tumors

with higher expression of C1S, C1R, and SERPING1 were significantly associated with

shorter PFS than tumors with lower expression of them (Table 11). Thus, the

complement pathway may predict patients’ prognosis in those with de novo metastatic

CRC with MSS and BRAF V600E mutation. We first assessed the expression correlation

of complement genes because the complement system involves various genes and

pathways and showed that they could be grouped into three co-expression modules

(Figure 4A). Subsequently, we grouped the patients into two groups according to the

clustering of genes in each module. Patients with high expression of complement genes

(Cluster 2) (Figure 4B) in their tumors had significantly shorter PFS (p = 0.00066) and a

trend for shorter OS (p = 0.087) when module 2 (most genes belong to the classical

complement pathway) was used for stratification (Figure 4C). Module 1 and Module 3

did not have sufficiently clear gene data for stratification (Figure 5).

To develop a better prognostic panel, we generated a complement score, which was

calculated by measuring the average gene expression value in module 2 (Figure 6A).

We used the median value of complement scores as the threshold to determine tumors

with low and high scores. The survival analyses showed that patients with a high

complement score had significantly shorter PFS (p = 0.029) and OS (p = 0.007) than

those with a low complement score (Figure 6B). Furthermore, this complement score

did not correlate with specific metastatic sites (Figure 7). As expected, complement

scores generated based on the genes in Modules 1 and 3 were not correlated with PFS

and OS (Figure 8). In the multivariate analysis, after adjusting for multiple tests, a high

complement score was associated with significantly shorter OS (HR: 2.44, 95%

confidence interval (CI): 1.26–4.70, adjusted p = 0.008) (Table 12). To validate the

results of our study, we searched for external CRC cohorts in open datasets; six patients

with metastatic BRAF V600E mutant CRC were included in both GSE39582 and TCGA

datasets. Thus, Stage 1-4 BRAF V600E mutant CRC (N= 49 in GSE39582 dataset and

N= 55 in the TCGA dataset) were used as the alternatives. The results showed a

significantly shorter PFI (p = 0.03) and a trend for shorter OS (p = 0.37) in patients with

a high complement score than in those with a low complement score in the GSE39582

cohort (Figure 6C). In the TCGA colon and rectal adenocarcinoma cohort, patients with

a high complement score correlated with a non-significant shorter OS time than those

with a low complement score (p = 0.056), and there was no significant difference in PFI

between these two groups (Figure 9A). On the other hand, the TCGA cohort included

31 Stage 4 CRC with genotypes of MSS and BRAF wild type. In these tumors, a high

complement score was not associated with a poor prognosis (Figure 9B). The

correlation between a complement score and BM1 or BM2 was also analyzed in the

GSE39582 and TCGA datasets. The results revealed that the complement scores in

BM1 CRC were significantly higher than that in BM2 CRC (Figure 6D, Figure 9C).

In immunohistochemical staining, the density of intra-tumoral C4d-expression

cells was semi-quantified (N=53). One patient was excluded from the analysis because

of no available tissue sections for C4d staining. C4d staining was categorized into three

groups. C4d-IHC 0 referred to low expression, C4d-IHC 1 to intermediate expression,

and C4d-IHC 2 to high expression (Figure 10A). We evaluated the association between

complement signatures and C4d expression in our cohort and showed that the

complement scores in tumors with high (p = 9.6e^⁻4) and intermediate (p = 5.5e^⁻4)

expressions of C4d were significantly higher than the scores in tumors with low

expression (Figure 10B).

Recently, many studies indicate that in situ C1q produced by TAMs may

contribute to intra-tumoral complement pathway activation^92,93. This study evaluated the

correlation between tumor-infiltrating immune cells and a complement score^93,94. We

used two independent methods to estimate immune cell abundance, and the result

showed that macrophage abundance was strongly associated with the complement score

(Figure 11A). Next, we analyzed the association between a complement score and the

M1 or M2 signature and demonstrated that only the signature of M2 TAM (ρ = 0.78, p <

10e^⁻16; Figure 11B) but not the signature of M1 TAM was strongly correlated to the

complement score.