We collected 145 FFPE CRC tumor blocks from 70 patients. Seven patients were excluded due to insufficient primary tumor tissue. Clinical and demographic characteristics of the 63 remaining study patients are summarized in and detailed in Table S1. The majority of patients had stage IV disease at presentation, and all had liver-only metastases. In two-thirds of cases, the paired specimens were synchronous. Of metachronous pairs, colectomy preceded metastectomy in all but two cases. No patient received EGFR-targeted monoclonal antibodies prior to surgery, but four patients were subsequently treated with cetuximab.
Clinical and demographic characteristics for the cohort of 63 patients
Of the 63 patients comprising our cohort, 48 (76%) died over the course of the study. The median follow-up time for the entire cohort was 28 months (Table S1). The median OS of patients with unresectable liver metastases was 18 months, compared to 58 months when liver metastases were resected with curative intent (HR = 7.68, 95% CI (3.81, 14.61), P = 9.0 × 10−10). The median OS for the entire study cohort was 33 months (Table S1 and Fig. S1). Biopsies of unresectable liver metastases were obtained during a palliative procedure now infrequently used at UCSF.
Concordance of PTEN expression in paired CRC primaries and liver metastases
Loss of PTEN expression was observed in 7 of 62 assessable CRC primaries (12.3%) and 6 of 58 assessable liver metastases (10.3%) ( and ). Uninterpretable results due to poor internal control staining were more common in liver metastases than in primary CRC. This was sometimes secondary to poor tissue morphology, including diffuse necrosis, and was not associated with exposure to chemotherapy (Table S1; OR = 2.87, 95% CI (0.29, 144.3), P = 0.656). Uninterpretable results were distinguished from equivocal staining, which was defined as diffuse, very weak staining of tumor cells that could represent either a weak positive or a nonspecific negative result (). In two of four cases where multiple, synchronous liver lesions from the same patient were available, staining was equivocal in some metastases and positive in others (reported as positive, ). The percent of cells staining positive for PTEN followed a bimodal distribution, with the majority of tumors exhibiting PTEN expression in 0% or 100% of cells (Table S1B).
Figure 1 Concordance of PTEN expression, PIK3CA, and KRAS/NRAS/BRAF mutations in 63 pairs of primary and metastatic colorectal carcinomas. The top rows represent the primary tumors (P) and the bottom rows represent the liver metastases (M). Patient number is listed (more ...)
Figure 2 Immunohistochemical staining for PTEN expression. (A) PTEN positive primary colon adenocarcinoma and (B) corresponding PTEN positive liver metastasis, both from patient 54; (C) PTEN negative primary colon adenocarcinoma and (D) corresponding PTEN negative (more ...)
After exclusion of cases with uninterpretable or equivocal PTEN staining, 50 CRC primary/liver metastasis pairs remained (). PTEN expression (positive or negative) differed in only one pair, corresponding to 98% concordance (95% CI for true proportion of concordance, [89.3%, 99.9%]) (). Patient 7, with discordant PTEN staining, had surgery to remove the CRC primary and biopsy of liver metastases within 3 months, without intervening chemotherapy. Two metastases were sampled: staining was positive in one lesion and equivocal in the other, with mostly necrotic tissue and few tumor cells (not shown).
Concordance of PTEN, PIK3CA, RAS, and BRAF in paired CRC primaries and liver metastases.1
If only patients with uninterpretable PTEN staining were excluded, 57 CRC primary/liver metastasis pairs remained (). PTEN expression (positive, negative or equivocal) differed in eight pairs (), corresponding to 86.0% (49/57) concordance (95% CI for proportion of concordance, [74.2%, 93.7%]). This concordance is the most conservative estimate for these data, as it assumes that every equivocal PTEN measurement disagrees with the PTEN status of the matched sample.
PTEN expression is not associated with mutation of PIK3CA or KRAS/NRAS/BRAF
Sanger sequencing with PNA clamping of wild-type alleles and the Sequenom platform were used to assay for mutations in PIK3CA, KRAS, NRAS, and BRAF (). The frequency of PIK3CA mutations was 8% in colorectal primaries (5/62) and 11% (7/62) in liver metastases, with 96.7% concordance (59/61 matched pairs; 95% CI for proportion of concordance, [88.7%, 99.6%]) (). In two cases, a subpopulation of tumor cells with a mutation in exon 9 of PIK3CA was detected only in the metastasis (Fig. S2). The frequency of MAPK pathway mutations in KRAS, NRAS or BRAF, was 53.2% (33/62) in primaries and 51.6% (32/62) in metastases. MAPK pathway mutations were 98.4% concordant in matched primaries and metastases (60/61; 95% CI for proportion of concordance, [91.2%, 100.0%]) ().
Overall, 60.3% of tumors exhibited at least one aberration in the PI3K or the MAPK pathway and 16% of tumors contained aberrations in both pathways (38/63 and 10/63 patients, respectively; ). Similarly, 57% of samples in which PTEN was not expressed (4/7) exhibited a mutation in either KRAS or BRAF. Concomitant absence of PTEN staining and the presence of mutant PIK3CA were never observed in the same tumor, but this result was not statistically significant. In fact, there was no statistically significant association between PTEN expression and mutation of PIK3CA or KRAS/NRAS/BRAF in CRC primaries or liver metastases (Tables S2 and S3).
Association of PTEN expression with OS
The median OS of patients with tumors that did not express PTEN was 9 months from diagnosis (), compared to 49 months for patients with PTEN positive matched primaries and metastases (HR = 6.25, 95% CI [1.98, 15.42], P = 0.0017) ( and Table S1). Results were similar for OS by PTEN status among all assessable CRC primaries (HR = 5.63, 95% CI [1.84, 12.86], P = 0.0022) (). Of the seven patients whose tumors exhibited loss of PTEN expression, 5 (71.4%) had unresectable liver metastases (95% CI for proportion of unresectable liver metastases [29%, 96.3%], P = 0.453) (). Patient 6, who underwent liver resection and has survived over a year, carries a diagnosis of hereditary nonpolyposis CRC. Patient 7, with PTEN staining in the metastasis, but not the primary, survived 9 months (Table S1).
Association of molecular markers with resectable liver metastases and survival
Figure 3 Kaplan–Meier estimates of overall survival related to: (A) PTEN expression in matched colorectal primaries and liver metastases; (B) PTEN expression in colorectal primaries; (C), number of aberrations in PTEN, PIK3CA and/or KRAS/NRAS/BRAF in matched (more ...)
Forty-eight patients had complete biomarker data (PTEN expression; PIK3CA, KRAS, NRAS, and BRAF mutation status) from their CRC primary and liver metastasis (). In these patients, there was a decreasing linear trend in the odds of liver metastases being resected as the number of biomarker aberrations increased (OR = 0.45; 95% CI [0.18, 1.01], P = 0.052) (). There was an increasing trend in the risk of death as the number of biomarker aberrations increased (HR = 1.86, 95% CI [1.17, 2.89], P = 0.0083) ( and ).
When the 56 patients with complete biomarker data from their CRC primary were considered, results were similar to those described above. In these 56 patients, there was again a decreasing linear trend in the odds of liver metastases being resected as the number of biomarker aberrations increased (OR = 0.42; 95% CI [0.18, 0.93], P = 0.033). There was an increasing trend in the risk of death as the number of biomarker aberrations increased (HR = 1.96, 95% CI [1.24, 2.99], P = 0.0032) ().
In the subcohort of patients with unresectable liver metastases, PTEN expression remained associated with OS (HR = 3.81, 95% CI [1.15, 11.21], P = 0.030) (Fig. S3A). Identification of any aberration in the PI3K or MAPK pathway was not significantly related to survival of patients with unresectable or resected liver metastases (Fig. S3B and C).
Variables significantly associated with OS in a univariate Cox PH regression model (Table S4) were considered in multivariate model construction (Table S5). Although the number of aberrations was a significant predictor of OS in the univariate analysis, we excluded it from the multivariate analysis because of its strong association with PTEN expression (P = 0.007). After adjusting for age, liver resection status, and the presence of a PIK3CA or RAS/BRAF mutation, PTEN loss remained significantly associated with an increased risk of death (HR = 6.31, 95% CI [2.03, 17.93], P = 0.0023) (Table S5).