Colorectal cancer staging criteria do not rely on examination of neuronal tissue. The authors previously demonstrated that perineural invasion is an independent prognostic factor of outcomes in colorectal cancer. For the current study, they hypothesized that neurogenesis occurs in colorectal cancer and portends an aggressive tumor phenotype.
In total, samples from 236 patients with colorectal cancer were used to create a tissue array and database. Tissue array slides were immunostained for protein gene product 9.5 (PGP9.5) to identify nerve tissue. The correlation between markers of neurogenesis and oncologic outcomes was determined. The effect of colorectal cancer cells on stimulating neurogenesis in vitro was evaluated using a dorsal root ganglia coculture model.
Patients whose tumors exhibited high degrees of neurogenesis had 50% reductions in 5-year overall survival and disease-free survival compared with patients whose tumors contained no detectable neurogenesis (P = .002 and P = .006, respectively). Patients with stage II disease and high degrees of neurogenesis had greater reductions in 5-year overall survival and disease-free survival compared with lymph node-negative patients with no neurogenesis (P = .002 and P = .008, respectively). Patients with stage II disease and high degrees of neurogenesis had lower 5-year overall survival and disease-free survival compared with patients who had stage III disease with no neurogenesis (P = .01 and P = .008, respectively). Colorectal cancer cells stimulated neurogenesis and exhibited evidence of neuroepithelial interactions between nerves and tumor cells in vitro.
Tumor-stromal interactions are critical in cancer development. This field encompasses angiogenesis, inflammation, matrix remodeling, and now perineural invasion (PNI). PNI is a poor prognostic factor in malignancies like prostate, head and neck, and pancreatic cancer1-7; furthermore, reports from our group and others indicate that PNI is an independent prognostic factor of poor outcomes in colorectal cancer.8-11 However, the role of neurogenesis in colorectal cancer is unclear.
According to guidelines, patients with lymph node-negative colorectal cancer may not receive adjuvant therapy.12 Although the majority of these patients have more indolent disease processes, there is a subset of patients who have aggressive phenotypes and die of recurrent cancer.13 It is difficult to prospectively identify patients with lymph node-negative who potentially may benefit from the adjuvant therapies currently received by lymph node-positive patients. The identification of independent prognostic factors would facilitate the selection of patients with lymph node-negative disease who potentially may benefit from adjuvant therapy.
We hypothesize that neuroepithelial interactions are important in colorectal cancer progression. Therefore, we evaluated whether neurogenesis occurs in colorectal cancer and its role as an independent prognostic factor in patient outcomes. We also investigated the role of neurogenesis in potential therapy stratification of patients with lymph node-negative disease.
MATERIALS AND METHODS
In Vitro Studies
Tissue culture and cell lines
The primary human colon adenocarcinoma cell line KM12 and its metastatic KM12L4 cell line counterpart, which were characterized previously by Morikawa et al,14, 15 were selected for this study. These colon carcinoma cell lines, which were isolated from a patient with Dukes B colon cancer and expanded into a nude mouse model splenic injection model of colon cancer liver metastases, were purchased from Dr. Isaiah J. Fidler at the University of Texas MD Anderson Cancer Center. The KM12 cell line demonstrates low metastatic potential in in vivo metastasis models, whereas KM12L4 cells are highly metastatic. The HCT116 cell line (human colon carcinoma cell line) also was selected.
Two new in vitro neurogenesis assays
Dorsal root ganglia assay.
A dorsal root ganglia (DRG) assay was used to evaluate neurogenesis in vitro as described previously.16-20 Briefly, DRG from the cervical, thoracic, and lumbar areas of 8-week-old Balb/C mice were dissected under sterile conditions. EHS Matrigel (BD Biosciences, San Jose, Calif) was used as an extracellular matrix. A single DRG and 1 × 105 cells (either KM12 or KM12L4) were cocultured in Matrigel. KM12 and KM12L4 cells only and DRG only were cultured in Matrigel as controls. The cocultures were photographed at a magnification of ×40 using inverted microscopy (Nikon, Instrument Inc., Houston, Tex). Active neurogenesis in the form of new dendritic projections from DRG was quantified using the Optimas 6.1 image-analysis system (Optimas Corp., Bothell, Mass) to obtain images from different experimental days. Neuroepithelial interactions between DRG and colon cancer cells also were evaluated by confocal microscopy.
Neuronal promoter assay.
To confirm that the projections observed in the DRG assay were nerve projections, we used a novel neuronal promoter assay. Most neurogenesis model assays rely on the measurement of the total number or length of neurites derived from modified neuronal cells. Primary neuronal cultures are extremely difficult and not practical. Hence, we have developed a novel system for measuring neurogenesis derived from true neuronal cells. The ganglia were derived from Mapttm1(GFP)Klt/J mice from Jax (JAX Mice stock no. 004779; The Jackson Laboratory, Bar Harbor, Me). In these mice, a knock-in of the green fluorescent protein (GFP) coding sequence into the first exon disrupts expression of the microtubule-associated protein tau (Mapt) gene and produces a cytoplasmic GFP fused to the first 31 amino acids. The fluorescent signal is detected exclusively throughout the developing central and peripheral nervous systems. GFP expression persists into adulthood. Ganglia derived from these mice were cocultured either with the HCT116 cell line or alone for controls. By using confocal microscopy, the individual neurons and neurite sprouting from each DRG were confirmed as neuronal projections and quantified as described above. All studies were done in triplicate and were reproduced twice.
Human Tissue Studies
Creation of colorectal tissue array and database
The records of 236 consecutive patients with colorectal adenocarcinomas who underwent primary resection of their tumors at the Michael E. DeBakey Veterans Affairs (VA) Medical Center between January 1995 and June 2000 were reviewed. Data pertaining to demographics, staging, treatment modalities, pathology characteristics, and oncologic clinical outcomes (overall survival and disease-free survival) were reviewed for each patient and entered into a comprehensive colorectal cancer database. The observation time in this patient cohort was the interval between the date of surgical resection and the date of last contact (death or last follow-up). The mean duration of follow-up was 6.6 years (range, 4.6-10.3 years). Cancer recurrences, deaths, and disease-free periods were tracked by trained cancer registrars and recorded in the institution's cancer registry, which was updated annually. The determination of cancer-related versus cancer-unrelated death was made by the tumor registrar who tracked the individual patient. If the cancer registrar was unable to make a determination regarding disease status or cause of death, then the patient's electronic medical record, which included clinic visit notes, imaging studies, carcinoembryonic antigen determinations, and surveillance endoscopy reports, was reviewed by a surgical oncologist who determined the disease status or cause of death. If a patient was lost to VA follow-up, then the patient or their next of kin was contacted by telephone. All patient information and pathology material was collected under a protocol that was approved by the Institutional Review Board of Baylor College of Medicine. Either a medical or surgical oncologist who participates in our institution's multidisciplinary tumor board established tumor stage postoperatively. Stages 0 to IV were defined according to the 2002 American Joint Commission on Cancer (AJCC) TNM classification system. The level of resection was determined for each patient based on the operative and pathology reports: R0 indicated negative gross and pathologic margins, R1 indicated grossly negative margins with positive microscopic margins, and R2 indicated grossly positive margins.
Specimens from each patient in the database were used to create a tissue array. Original hematoxylin and eosin-stained slides from resected colorectal adenocarcinoma specimens were reviewed by a pathologist and mapped. Areas of highest tumor grade and normal colon were circled. Normal colon specimens came from blocks of tissue distant from the lesion. The tissue array was built manually using 2-mm punch-biopsy cores that were transferred from mapped donor blocks and placed into recipient paraffin array blocks. From each mapped and sampled area, cores were taken in duplicate. Internal controls that contained 10 different types of tissue within each core also were placed at spaced intervals within each recipient block. A database was built for every block produced and included the coordinates of each core and the area and patient of origin. The collection of human tissue specimens was performed under a protocol that was approved by the Baylor College of Medicine Institutional Review Board, and patient identifiers were removed before the release of tissue blocks from the Pathology Department.
Immunostaining for nerves
Immunohistochemically stained specimens were created from 5-μm slices of paraffin-embedded tissue array blocks. Antigen retrieval was accomplished with 0.01 M citrate buffer, pH 6.0; then, the specimens were microwaved at 900 W for 3 minutes followed by 90 W for 20 minutes. Nonspecific protein binding was blocked for 30 minutes in a blocking solution of 4 g bovine serum albumin; 20 mL 0.5 M Tris-buffered saline TBS, pH 7.6; 180 mL distilled water; and 2 mL 10% sodium azide. After washing, the sections were incubated with primary rabbit antihuman protein gene product 9.5 (PGP9.5) (1:1000 dilution; Chemicon, Temecula, Calif) overnight at 4°C in a humidified slide box. After washing, the slides were incubated with biotinylated secondary antibody (Vectastain avidin-biotin complex [ABC] kit; Vector Laboratories, Burlingame, Calif) for 30 minutes at room temperature. The immunostaining was observed with the ABC immunoperoxidase system (Vectastain ABC kit; Vector Laboratories). The slides were counterstained with hematoxylin for 90 seconds, dehydrated, and mounted in permanent media.
Evaluation of neurogenesis in colorectal cancer
A pathologist (G.E.A.) read all immunostained whole mount slides from each patient. Each slide was marked at points of the greatest nerve concentration (hotspots) in tumor and normal colon areas. Because the depth of the serial section was such that certain regions could not be observed in cross-section or because, for any particular slide, there was insufficient nerve concentration to warrant mapping a hotspot, not every region had a hotspot on each slide. The net result of this was that each slide featured an average of 3 of 9 possible hotspots.
A single individual created digital images of the hotspot marked on the mapped slides using a Nikon microscope (Nikon Instruments, Inc., Melville, NY) and an Ikegami digital camera (Ikegami Electronics USA, Inc., Maywood NJ). The images were analyzed using the Optimas 6 Image Analysis Suite (Optimas Corp.) using the brown color of the positively stained nerves as a threshold, and the cross-sectional area of the neural hotspots was calculated and entered into a databases.
Impact of neurogenesis as a prognostic indicator of outcomes
Evaluation of nerve density and diameter.
Two previously validated surrogate markers8 of neurogenesis in colorectal cancer tissue were evaluated: nerve density and nerve diameter. Nerve density was evaluated by counting the number of nerves per high-power field (hpf). Nerve density results were grouped into 3 categories: 1) no nerves per hpf, 2) 1 to 20 nerves per hpf, or 3) >20 nerves per hpf. Nerve diameter (the average of the 3 largest nerves identified) also was measured in all specimens. Patient tumor specimens were classified as having nerve diameters <120 μm or >120 μm. Two observers who were blinded to patient outcomes performed all measurements.
Quantification of neurogenesis in colorectal cancer.
To determine whether neurogenesis was present in colorectal cancer tissues, we used a tissue microarray that was constructed with tissues from resected colorectal adenocarcinoma specimens. Normal and adenocarcinoma samples were represented twice per patient in 2-mm cores. Sections were stained with antibodies against PGP9.5. The cores were digitized with the Olympus BLISS HD virtual microscopy system (Olympus Corporation, Center Valley, Pa) at ×20 magnification on the subepithelial plexi. The images revealed representative examples of normal colon and adenocarcinoma with the nerves of the subepithelial plexi staining in brown. The subepithelial neural density quantification was done with the Optimas 6 Image Analysis Suite (Optimas Corp.). We also investigated the specimens for evidence of neuroepithelial interactions between the stroma and cancer cells of the tissue specimens.
Overall survival was censored at the time of last follow-up for patients who remained alive and at the time of death for patients who died from causes unrelated to cancer. Disease-free survival was censored at the time of last follow-up for patients who remained disease-free and at the time of death for patients who died from causes unrelated to cancer. The influence of neurogenesis (nerve density and nerve diameter) on overall and disease-free survival was evaluated with the Kaplan-Meier method. To examine cancer-specific overall survival and disease-free survival, we included only patients who underwent curative R0 resection to eliminate the effect of residual disease in patients who underwent R1 and R2. The association of nerve density and diameter with various clinical characteristics was assessed using the Fisher exact test. The effect of nerve density on recurrence and cancer-specific survival was analyzed using a Cox regression model for multivariate analysis controlling for age, the number of resected lymph nodes, tumor grade, AJCC stage, and whether or not the patient received adjuvant therapy. The association between the time to recurrence/death and nerve density and diameter was tested using a log-rank test. Differences between means were compared using the Student t test and analyses of variance. Analyses were performed with Prism software (version 4; GraphPad, San Diego, Calif), Stata statistical software (release 10; StataCorp LP, College Station, Tex), and R (a language and environment for statistical computing; R Development Core Team, R Foundation for Statistical Computing, Vienna, Austria). Results were considered significant at P < .05.
Evidence of Neurogenesis in Colorectal Cancer
When DRG were cultured alone using the DRG assay, active neurogenesis in the form of new dendritic projections from DRG was negligible. In contrast, active neurogenesis was observed from DRG when they were cocultured with the low metastatic KM12 colon adenocarcinoma cells (5% increase) (Fig. 1). This active neuronal response was much more pronounced in response to the highly metastatic KM12L4 cells (15% increase; P < .05) (Fig. 1).
By using our new neurogenesis assay, as described above, we observed a 2-fold increase in neurogenesis when DRG alone were compared with DRG plus HCT116 cells (P < .017). The median area in DRG alone was 43,055 pixels (70,541 μm2) (Fig. 2A); and, in DRG plus HCT116 cells, the median area was 70,661 pixels (11,5771 μm2; P < .01) (Fig. 2B).
Three-dimensional models were constructed of both normal tissues (Fig. 3A) and adenocarcinoma tissues (Fig. 3B) that were stained for PGP9.5. The images were quantified by plotting the surface plot of the nerve area on the x-axis and the intensity of the stain on the y-axis. There was a 3-fold increase in the cancerous specimen (1790 units vs 4839 units). Furthermore, low nerve densities were observed in the stalk and stroma of normal colonic tissue (Fig. 4B), but higher nerve densities were observed in colonic stroma of adenocarcinoma tissue (Fig. 4C,D).
Evidence of Neuroepithelial Interactions
Neuroepithelial interactions were observed in the form of active migration of tumor cells along DRG dendritic projections. The highly metastatic colon cancer KM12L4 cells exhibited higher degrees of neurotropism when cocultured with DRG than the lower metastatic potential KM12 colon cancer cells. The average tumor cell migration distance over 6 days of coculture was 110 μm for the KM12L4 cells versus 87.5 μm for the KM12 cells. This difference did not reach statistical significance, but there was a trend toward greater tumor cell migration among the phenotypically more metastatic cell line.
To evaluate the tumor stromal interaction, we examined normal and adenocarcinoma tissues that were stained for PGP9.5. Figure 4 provides images of nerve growth in specimens (Fig. 4B-D) and illustrates that the nerve growth has polarity as it grows in the stromal tumor cells (Fig, 4E-G).
Impact of Neurogenesis on Biologic Tumor Behavior
Patient and tumor characteristics
Ninety-nine percent of our patients were men, and their mean age was 66 years (range, 38-93 years). Their tumors were located most commonly in the right colon (27%), rectosigmoid (35%), and rectum (26%). The array contained a representative sample of all stages of colorectal cancer, including stage 0 disease in 8% of patients, stage I disease in 19%, stage II disease in 29%, stage III disease in 24%, and stage IV disease in 19%. Seventy-eight percent of the patients with colorectal cancer who were included in our tissue array underwent complete surgical resections (R0), 6% underwent R1 resection, and 16% underwent R2 resection. Sixty-seven percent of patients with stage III and IV disease who were eligible for adjuvant therapy after an R0 resection received adjuvant chemotherapy and/or radiation. Of the 116 patients who had lymph node-negative disease, 13% received adjuvant chemotherapy and/or radiation; of that 13%, 60% were patients who had tumors in the rectum or rectosigmoid.
On univariate analysis, the effect of clinical and pathologic variables on 5-year disease-free and overall survival confirmed the significance of increasing tumor stage, lymph node status, the presence of metastasis, AJCC disease stage, resection level, and tumor grade as poor prognostic factors in colorectal cancer (data not shown; P < .001). In our patient population, there was no significant difference in outcomes relating to the number of lymph nodes analyzed.
Correlation of neurogenesis and colorectal cancer
Overall, 63% of patients with colorectal cancer exhibited some degree of neurogenesis, and 37% had no nerves in their primary tumors. Of the patients who exhibited neurogenesis, 88% had lower nerve density (1-20 nerves per hpf), and 11% had a high degree of neurogenesis (>20 nerves per hpf). The percentage of tumors that exhibited neurogenesis increased from carcinoma in situ lesions up to stage III lesions and included 40% of stage 0 lesions, 53% of stage I lesions, 65% of stage II lesions, 72% of stage III lesions, and 70% of stage IV lesions (P = .03). Furthermore, of the tumors that exhibited neurogenesis, higher degrees of neurogenesis (>20 nerves per hpf) were more likely to be observed in lymph node-positive patients (22% of patients with stage III disease) compared with lymph node-negative patients (9% of patients with stage I disease and 11% of patients with stage II disease). Tumor grade was not correlated with neurogenesis (P = .09).
No significant differences in the degree of neurogenesis were observed among primary colon tumors in different colonic locations (data not shown; P = .2). However, rectal tumors exhibited significantly more neurogenesis than colon tumors (73% of rectal tumors vs 60% of colon tumors; P = .02). Similarly, of the tumors that had neurogenesis, far more rectal tumors (20%) exhibited a high degree of neurogenesis (>20 nerves per hpf) compared with only 8% of colon lesions.
Impact of Colorectal Cancer Neurogenesis on Patient Outcomes
We evaluated 2 surrogate markers of neurogenesis in colorectal cancer: nerve density and nerve diameter. Overall, patients who had high nerve density (>20 nerves per hpf) had significantly decreased cancer-specific overall survival (5-year survival rate, 40% vs 86% for patients without neurogenesis; P = .002) (Fig. 5A) and an increased rate of recurrence (5-year disease-free survival rate, 41% vs 80% for patients without neurogenesis; data not shown; P = .006). The median survival for patients who had a high degree of neurogenesis was 2.8 years compared with 9.7 years for patients without neurogenesis. Patients who had tumors with a high degree of neurogenesis, as exhibited by large nerve diameters (>120 μm), had a similar decrease in cancer-specific overall survival (5-year survival rate, 30% vs 66% for patients with smaller nerves; P = .03) (Fig. 5B) and had an increase in cancer recurrence (5-year disease-free survival rate, 22% vs 64% for patients with smaller nerves; data not shown; P = .04).
Multivariate analysis using the Cox multiple regression method confirmed that a high degree of neurogenesis was a significant predictor of decreased cancer-specific overall and disease-free survival (Fig. 5C) when the analysis was controlled for age, the number of lymph nodes resected, tumor grade, AJCC stage, and whether or not the patient received adjuvant therapy. In our data analysis, neurogenesis emerged as a much more powerful, independent predictor of outcomes compared with lymph node status. In fact, the only prognostic factor that was more powerful than neurogenesis in our study was the presence of metastatic disease.
Impact of Neurogenesis on Lymph Node-Negative Colorectal Cancer
Patients who had stage II (lymph node-negative) disease with tumors that exhibited a high degree of neurogenesis had significantly worse cancer-specific overall and disease-free survival compared with patients who had either intermediate or no neurogenesis evident in their tumors. Lymph node-negative patients with a high degree of neurogenesis had a significantly lower 5-year overall survival rate compared with lymph node-negative patients without neurogenesis (25% vs 85%; P = .002) (Fig. 6A), and they also had a shorter disease-free survival (25% vs 75%; P = .008) (Fig. 6B).
Furthermore, when we compared patients who had stage II disease and a high degree of neurogenesis (who routinely received no adjuvant therapy) with patients who had stage III disease without neurogenesis (who routinely received adjuvant therapy), this subset of lymph node-negative patients actually had a shorter cancer-specific overall survival (5-year survival rate, 25% vs 77%; P = .01) (Fig. 6C) and disease-free survival (5-year disease-free survival rate, 25% vs 70%; P = .008) (Fig. 6D).
Impact of Neurogenesis on Lymph Node-Positive Colorectal Cancer
Patients who had stage III disease and tumors that exhibited a high degree of neurogenesis tended to have shorter 5-year cancer-specific overall survival compared with lymph node-positive patients who did not have neurogenesis (30% vs 77%; P = .1) (Fig. 6E), and they had shorter 5-year cancer-specific disease-free survival (31% vs 70%; P = .2) (Fig. 6F). Eighty percent of patients with the highest degree of neurogenesis actually received adjuvant therapy, which was well above our patient cohort average of 67%.
The role of nerves and neuroepithelial interactions in solid malignancies remains a poorly understood phenomenon. PNI remains the best-characterized interaction between nerves and tumors6, 21 and is reported mostly in literature on prostate cancer, head and neck cancer, and pancreatic cancer.1-7, 22, 23 Recent reports from our group and others have demonstrated that PNI is an independent prognostic factor of adverse outcomes in patients with colorectal cancer.8-11, 24 In our investigations of PNI, we noticed significant differences in nerve tissue density among different colorectal cancer primary tumors, and we hypothesized that active neurogenesis may occur in some colorectal cancers.
Ayala and colleagues recently reported active neurogenesis in prostate cancer.19 Nerve density in prostate tissues was higher in cancer specimens and premalignant specimens compared with normal prostate tissues. Only recently, a few investigators reported the presence of nerves in tumors.25-28 Now, we report the observation of neurogenesis in colorectal cancer.
Overall, we observed that 66% of all patients with colorectal cancer exhibited some degree of neurogenesis in their primary tumors. It is noteworthy that patients with rectal cancer have a higher incidence and intensity of neurogenesis than patients with colon cancer. This may be related to the fact that the rectum is surrounded by a rich autonomic nerve plexus in the pelvis. Because it is well known that, stage for stage, patients who have rectal cancer have decreased survival compared with patients who have colon cancer, this observed increase in neurogenesis among patients with rectal cancer may help explain the difference in survival.29
Patients who have a high degree of neurogenesis have significantly lower cancer-specific overall and disease-free survival compared with patients who have no evidence of neurogenesis (Fig. 5A,B). Strikingly, our multivariate analysis confirmed that a high degree of neurogenesis was an independent prognostic factor for poor outcomes in colorectal cancer (Fig. 5C). In fact, with the exception of the presence of metastasis, neurogenesis was the most powerful predictor of a poor outcome in patients with colorectal cancer in our series.
Currently, therapy stratification for patients with colorectal cancer is based primarily on lymph node status: Lymph node-positive patients routinely receive adjuvant therapy, and lymph node-negative patients largely do not.12 Our data indicate that neurogenesis acts as an independent predictor of a poor outcome in patients with lymph node-negative colorectal cancer. We observed that lymph node-negative patients who had high degrees of neurogenesis had significantly worse survival than lymph node-positive patients without neurogenesis. This suggests that this subset of lymph node-negative patients with a high degree of neurogenesis may benefit from available adjuvant therapy.
We also observed that a high degree of neurogenesis in primary tumors accurately predicted poor outcomes in lymph node-positive patients (Fig. 6E,F). In this subset of patients who have lymph node-positive disease with a high degree of neurogenesis, the patients die of recurrent colorectal cancer despite receiving up-to-date adjuvant therapy regimens. This emphasizes the need to understand the molecular mechanisms of neurogenesis and to develop targeted therapies based on this novel phenomenon. The first step in identifying molecular mediators of neurogenesis is to develop an effective experimental model of neurogenesis in colorectal cancer. With this in mind, we modified the mouse DRG model to study the neuroepithelial interactions in colorectal cancer. By using this model, we established that colorectal cancer cells induced neurogenesis from DRG in vitro to an extent that is related directly the metastatic potential of the cancer cells. Furthermore, this model exhibited a reciprocal neuroepithelial interaction with tumor cell migration along neuronal projections from the DRG, mimicking the process of neurogenesis and PNI observed in colorectal cancer specimens. More studies will be necessary to elucidate the molecular mediators of these neuroepithelial interactions.
In summary, we report the phenomenon of neurogenesis in colorectal cancer and its association with aggressive stages of colorectal cancer. Furthermore, we have demonstrated that neurogenesis is indicative of poor survival and recurrence and is an independent prognostic factor for poor outcomes in colorectal cancer. Thus, neurogenesis offers the potential for better therapy stratification of patients with lymph node-negative colorectal cancer. Further research will be necessary to elucidate the mechanisms of this phenomenon and to develop targeted therapies against neurogenesis in colorectal cancer.