Inhibition of axonal outgrowth in the tumor environment: Involvement of class 3 semaphorins

Abstract

That tumors lack innervation is dogma in the field of pathology, but the molecular determinants of this phenomenon remain elusive. We studied the effects of conditioned media from Colon 26 and B16 mouse tumor cell lines on the axonal outgrowth and cellular differentiation of embryonic Institute of Cancer Research (ICR) mouse dorsal root ganglion cells. Tumor‐conditioned media suppressed dorsal root ganglion axonal extension but had no effect on neuronal or glial differentiation. We found that the tumor cells expressed most of the class 3 semaphorins – axon guidance molecules. Blocking the activity of class 3 semaphorins with the soluble receptor neuropilin‐1 significantly counteracted the tumor‐induced inhibition of axonal extension. Together, these results suggest a role for tumor‐secreted class 3 semaphorins in selectively inhibiting axonal outgrowth of dorsal root ganglion neurons. (Cancer Sci 2007; 98: 1192–1197)

Abbreviations:

DMEM

Dulbecco's modified Eagle's medium

DRG

dorsal root ganglion

FBS

fetal bovine serum

GFAP

glial fibrillary acidic protein

LLC

Lewis lung carcinoma

NGF

nerve growth factor

Npn

neuropilin

PBS

phosphate‐buffered saline

PBST

PBS–Triton X‐100

PI

propidium iodide

RT

room temperature

RT‐PCR

reverse transcription–polymerase chain reaction

Sema3

semaphorin‐3

VEGF

vascular endothelial growth factor.

The lack of innervation in tumors is a generally accepted fact,^{( 1 )} but the molecular determinants of this phenomenon are poorly understood. Furthermore, certain tumors such as esophageal cancer were shown to be innervated by peptidergic nerve fibers and to promote process extension in DRG neurons.^{( 2 )} Notably, sarcomas can secrete NGF, which promotes proliferation and axonal outgrowth of DRG neurons, the observation of which led to the initial discovery of NGF.^{( 3 )} Thus, the effects of tumors on axonal growth are unclear and need further analysis.

Peripheral nerves are known to associate with blood vessels,^{( 4 )} reflecting their need for oxygen and nutrients and their control of vasoconstriction and vasodilation.^{( 5 )} In mutant embryos containing disorganized nerves, blood vessel branching is altered to follow the nerve,^{( 6 )} suggesting that local signals supplied by nerve fibers may provide a template that determines blood vessel patterning. These data indicate a functional relationship between axonal outgrowth and angiogenesis under physiological conditions, and make the question of tumor innervation even more intriguing.

DRG neurons innervate most of the internal organs, thus their processes have the highest potential for interacting with the cells of a tumor growing in the body. We therefore investigated whether molecular cues secreted from tumor cells affect axonal outgrowth and additionally alter the differentiation capacity of immature neural cells in the DRG. Here, we report that supernatant isolated from two tumor cell line cultures inhibited process extension of DRG neurons, and present evidence implicating secreted class 3 semaphorins as mediators of this tumor‐induced axonal inhibition.

Materials and Methods

DRG culture.  Embryos at embryonic day 12.5 were dissected from pregnant Institute of Cancer Research (ICR) mice (SCL, Shizuoka, Japan). DRG were extracted free of surrounding tissues and placed one per well in 24‐well, poly‐l‐lysine‐coated culture plates (Iwaki, Tokyo, Japan). Colon 26 (C26; mouse colon cancer) and B16 (mouse melanoma) tumor cell lines were cultured in DMEM (Sigma‐Aldrich, St Louis, MO, USA) supplemented with 10% FBS (Sigma‐Aldrich). Culture supernatants were collected when cells reached approximately 80% confluence. Four culture media were prepared: (i) control medium, DMEM + 10% FBS + 50 ng/mL human recombinant NGF (PeproTech, Rocky Hill, NJ, USA); (ii) tumor‐conditioned medium, equal amounts of control medium with 100 ng/mL NGF and tumor cell line supernatant were mixed; (iii) rescue medium, tumor‐conditioned medium with 50 µg/mL recombinant fusion protein of the extracellular domain of murine Npn1 with the Fc fragment of human IgG; and (iv) control rescue medium, tumor‐conditioned medium with 50 µg/mL recombinant fusion protein of the CD4 glycoprotein with the Fc fragment of human IgG.^{( 7} , ^{8 )} DRG were incubated in one of the above media for 48 h and then fixed and stained. Culture media were replaced with fresh media after 24 h of incubation.

Immunocytochemistry.  The immunohistochemical procedures on culture plates were basically the same as reported previously.^{( 9 )} Briefly, DRG on culture plates were fixed in 4% paraformaldehyde in PBS (pH 7.5) for 10 min at RT and rinsed with PBST. Non‐specific binding of secondary antibody was blocked with 5% normal goat serum and 1% bovine serum albumin in PBST (blocking serum) for 30 min at RT. Cultures were incubated with mouse anti‐β‐III tubulin primary antibody (1:200; Covance, Richmond, CA, USA) or rabbit anti‐GFAP antibody (1:200; DAKO, Kyoto, Japan) overnight at 4°C in blocking serum, rinsed three times for 10 min at RT with PBST and then incubated with goat antimouse IgG conjugated to Alexa Fluor 488 (1:100; Invitrogen, Carlsbad, CA, USA) in blocking serum for 1 h at RT. Cultures were rinsed again with PBST, and PI was added for 1 min at RT for visualization of nuclei.

RT‐PCR.  Total RNA was isolated from C26, B16, LLC, cloneM‐3 (mouse melanoma) and MM102‐TC (mouse mammary gland carcinoma) cells using the Isogen (Wako, Osaka, Japan) isolation kit according to the manufacturer's instructions, and were reverse‐transcribed with the SuperScript (Invitrogen) RT‐PCR system as reported previously.^{( 10 )} The class 3 semaphorin Npn1 and Npn2, and β‐actin gene sequences were amplified using ExTaq (TaKaRa, Tokyo, Japan) DNA polymerase with: Sema3A forward, 5′‐CGGGACTTCGCTATCTTCAG‐3′ and reverse, 5′‐GGGACCATCTCTGTGAGCAT‐3′; Sema3B forward, 5′‐AACCCATGCTTCAACTGGAC‐3′ and reverse, 5′‐CTGGAGGTGGAGAAGACAGC‐3′; Sema3C forward, 5′‐TGGCCACTCTTGCTCTAGGT‐3′ and reverse, 5′‐GCCTTCAGCTTGCCATAGTC‐3′; Sema3D forward, 5′‐AGCACCGACCTTCAAGAGAA‐3′ and reverse, 5′‐GTGCATATCTGGAGCAAGCA‐3′; Sema3E forward, 5′‐TTGGACAGCAATTTGTTGGA‐3′ and reverse, 5′‐AGCCAATCAGCTGCAAGAAT‐3′; Sema‐3F forward, 5′‐TGCTTGTCACTGCCTTCATC‐3′ and reverse, 5′‐TACAGGTGTGTTCGGTTCCA‐3′; Sema3G forward, 5′‐TCTTTGGCACAGAGCACAAC‐3′ and reverse, 5′‐CCTGCACCATACACGTTCAC‐3′; Npn1 forward, 5′‐CTCCCGCCTGAACTACCCTGAAAAT‐3′ and reverse, 5′‐CCACTTGGAGCCATTCATTGGTGTA‐3′; Npn2 forward, 5′‐TGAATCTCCAGGGTTTCCAG‐3′ and reverse, 5′‐GTCCACCTCCCATCAGAGAA‐3′; and β‐actin forward, 5′‐CCTAAGGCCAACCGTGAAAAG‐3′ and reverse, 5′‐TCTTCATGGTGCTAGGAGCAG‐3′ primer pairs. The PCR products were fractioned by electrophoresis and the positive bands were visualized in a FAS III (Toyobo, Osaka, Japan) ultraviolet transilluminator.

Western blotting.  C26 and B16 cells were trypsinized and collected. Procedures for the preparation of cell lysates and western blotting were basically the same as those reported previously.^{( 11 )} Briefly, total protein was extracted using Nonidet (N)P‐40 Tris buffer with proteinase inhibitor cocktail added. Proteins were fractioned by sodium dodecylsulfate–polyacrylamide gel electrophoresis and transferred to a membrane. The membrane was incubated for 30 min at RT with a generic protein to cover any remaining sticky places. Primary antibodies against mouse Sema3A (1:1000; Abcam, Cambridge, UK), Sema3C (1:1000; R&D Systems, Minneapolis, MN, USA) or glyceraldehyde‐3‐phosphate dehydrogenase (1:1000; Chemicon, Temecula, CA, USA) were added for 1 h at RT followed by horseradish peroxidase (HRP)‐conjugated goat antimouse IgG (1:1000; DAKO). The positive bands were visualized by ECL detection reagents (Amersham Biosciences, Piscataway, NJ, USA) in a LAS‐3000 imaging system (Fujifilm, Tokyo, Japan).

Image analysis.  Single and double labeling were visualized by confocal laser scanning microscopy (LSM 510; Carl Zeiss, Göttingen, Germany). Alexa Fluor 488 was assigned to the green channel and PI to the red channel. For single‐labeling experiments, an average of 16 z‐axis scans was used to generate the final images. In double‐labeling experiments, each fluorochrome was scanned separately and sequentially to minimize the probability of signal transfer among channels. Measurements of DRG halo diameter were carried out using Zeiss LSM software version 3.2 on six samples from each experimental group. Within each experimental group, at least 100 PI‐positive cells were investigated for costaining with β‐III tubilin or GFAP. Diameters and percentages of positive cells were averaged to obtain a mean density for each marker per experimental group.

Statistical analysis.  Diameters of DRG halos were compared using Student's paired t‐test. Percentages of cells expressing β‐III tubulin or GFAP were compared using non‐parametric tests (Mann–Whitney U‐test and Kruskal–Wallis test). Data are expressed as mean ± SEM. Differences were considered significant when P < 0.05.

Results

Tumor‐conditioned medium inhibited DRG axonal outgrowth.  DRG cells cultured under normal conditions (control medium) grew steadily, and within 48 h of culturing extended radial processes in all directions (Fig. 1a,b), in agreement with previous studies using similar culture media (DMEM supplemented with NGF).^{( 12} , ^{13 )} The removal of NGF from the culture medium resulted in extensive cell death (data not shown); therefore, all culture media were supplemented with NGF. In contrast, the DRG cells cultured with C26 (Fig. 1c) or B16 (Suppl. Fig. 1) tumor‐conditioned medium showed limited axonal outgrowth and their axons lacked the straightness of control axons (Fig. 1d). Quantitative measurement of DRG halo diameters revealed that DRG cells grown in tumor‐conditioned medium had significantly smaller halos than those grown in control medium (Fig. 1e).

Figure 1.

Tumor‐conditioned medium inhibits dorsal root ganglion (DRG) axonal outgrowth. Confocal images of DRG primary cultures exposed to (a,b) control medium or (c,d) colon 26 tumor‐conditioned medium. Cultured cells were stained with anti‐β‐III tubulin antibody. (b,d) High‐power views of areas indicated by boxes in (a) and (c), respectively. Scale bar = 500 µm. (e) Statistical evaluation of the DRG halo diameter. Data show mean ± SEM from five random fields. *P < 0.001 versus control. Shown are representative data from one of three independent experiments.

Tumor‐conditioned medium did not alter DRG cellular differentiation.  Tumor‐conditioned medium affected axonal outgrowth; however, tumor‐conditioned medium may also be involved in other cellular processes such as differentiation of DRG cells. Therefore, we stained DRG cells cultured in control or tumor‐conditioned media with antibodies against the neuronal marker β‐III tubulin or the astrocyte marker GFAP; the nuclei of all cells were counterstained with the DNA‐binding dye PI (Fig. 2a–l). We then quantitatively evaluated the percentages of β‐III tubulin‐ or GFAP‐positive cells among the total population of PI‐stained cells under high magnification (Fig. 2m,n). We evaluated β‐III tubulin‐positive neuronal cells adjacent to the center of the DRG halo because the density of neuronal cells in the center was too high to clearly distinguish individual cells, but we evaluated GFAP‐labeled astrocyte lineage cells in the periphery of the DRG halo where the incidence of astrocytes was high. The β‐III tubulin‐positive cells in both control (Fig. 2a–c) and tumor‐conditioned media (Fig. 2d–f) had similar appearance, and quantitative analysis of the percentage of β‐III tubulin‐positive cells revealed that there was no significant difference between the control and tumor conditions (Fig. 2m). Likewise, the GFAP‐positive cells in both control (Fig. 2g–i) and tumor‐conditioned media (Fig. 2j–l) had similar appearances, and there were no significant differences in the percentage of astroglial cells in the two conditions (Fig. 2n). Together, these results indicate that suppression of DRG axonal outgrowth in tumor‐conditioned medium did not result from an altered differentiation of the DRG cells into neuronal and astrocyte lineages.

Figure 2.

Tumor‐conditioned medium does not affect neuronal or glial differentiation of dorsal root ganglion (DRG) cells. Neuronal or glial cell development of DRG cells in primary cultures exposed to (a–c, g–i) control medium or (d–f, j–l) colon 26 tumor‐conditioned medium was analyzed by staining for (a,d) the neuronal marker β‐III tubulin or (g,j) the astrocyte marker glial fibrillary acidic protein (GFAP). Cells in (a,d,g,j) were counterstained with propidium iodide (PI) (b,e,h,k, respectively). The images in (c,f,i,l) are merged images of (a,b), (d,e), (g,h) and (j,k). Scale bar = 20 µm. (m,n) Percentage of β‐III tubulin‐positive neuronal cells (m) or GFAP‐positive glial cells (n) among the total PI‐positive cells in DRG primary cultures exposed to control or colon 26 tumor‐conditioned medium. Data are mean ± SEM from five random fields. Shown are representative data from one of three independent experiments.

Effects of tumor‐conditioned medium on DRG were dependent on class3 semaphorins.  Based on the above results, we searched for molecular cues in the tumor‐conditioned medium that potentially inhibit axonal outgrowth from DRG cells. Due to its well‐known growth cone repulsive effects,^{( 14} , ¹⁵ , ¹⁶ , ^{17 )} and because it is the only secreted class of the semaphorin family of proteins in vertebrates,^{( 18 )} we examined class 3 semaphorin expression in C26 and B16 tumor cells. We found that both C26 and B16 cells expressed Sema3A, Sema3B, Sema3C, Sema3D, Sema3E and Sema3F mRNA (Fig. 3a). Moreover, we found that other cancer cell lines, such as LLC, cloneM‐3 (mouse melanoma) and MM102‐TC (mouse mammary gland carcinoma) also express several class 3 semaphorins. Interestingly, the tumor cells we observed expressed more or less either Npn1 or Npn2, receptors for class 3 semaphorins. Further, western blotting of tumor cell revealed the presence of Sema3A and Sema3C proteins from both Colon26 and B16 cells (Fig. 3b).

Figure 3.

Tumor cell lines express class 3 semaphorins. (a) Reverse transcription–polymerase chain reaction analysis of mRNA extracted from colon 26 (C26) cells, B16 cells, lewis lung carcinoma (LLC), cloneM‐3 (mouse melanoma) and MM102‐TC (mouse mammary gland carcinoma). Distilled water was used as a negative control. Non‐reverse transcribed RNA did not generate any polymerase chain reaction product. β‐Actin was used as an internal loading control. (b) Western blotting of C26 and B16 cells. Glyceraldehyde‐3‐phosphate was used as an internal control.

As class 3 semaphorins are soluble ligands for Npn1, we tried to inactivate the function of Sema3 in the cultures. For this purpose we generated a recombinant fusion protein (Npn1–Fc) of the extracellular domain of murine Npn1 and the Fc fragment of human IgG.^{( 7} , ^{8 )} In the presence of Npn1‐Fc (50 µg/mL), DRG cells cultured in C26 tumor‐conditioned medium (Fig. 4e,f) appeared to show more extensive axonal outgrowth than those cultured in C26 tumor‐conditioned medium containing the CD4‐Fc control protein (Fig. 4c,d), but less than those cultured in control medium (Fig. 4a,b). Morphometric analyses confirmed this impression, as the DRG halo diameters in C26 tumor‐conditioned media with Npn1‐Fc were significantly larger than those in C26 tumor‐conditioned media alone, but still significantly smaller than that with control medium (Fig. 4g). A lower concentration of Npn1‐Fc (10 µg/mL) produced results that were not significantly different from those in the C26 tumor‐conditioned medium, and an excess of the Npn1‐Fc (250 µg/mL) did not add to its rescue effects for axonal outgrowth (Fig. 4g). As observed in the effect of Npn1‐Fc for C26 tumor‐conditioned media, in the presence of Npn1‐Fc DRG cells cultured in B16 tumor‐conditioned medium appeared to exhibit more extensive axonal outgrowth than those cultured in B16 tumor‐conditioned medium containing the CD4‐Fc control protein (Suppl. Fig. 1).

Figure 4.

Inactivation of class 3 semaphorins rescues dorsal root ganglion (DRG) axonal outgrowth. Confocal images of (a,b) DRG primary cultures grown in control medium, (c,d) colon 26 tumor‐conditioned medium with CD4‐Fc control protein, or (e,f) colon 26 tumor‐conditioned medium with 50 µg/mL neuropilin‐1‐Fc protein. Cultured cells were stained with anti‐β‐III tubulin antibody. (b,d,f) High‐power views of the areas indicated by boxes in (a,c,e). Scale bar = 500 µm. (g) Quantitative evaluation of the DRG halo diameter. Control, control medium; tumor, colon 26 tumor‐conditioned medium. Data show mean ± SEM from five random fields. Shown are representative data from one of three independent experiments. *P < 0.01.

Discussion

The nervous and vascular systems share several anatomical parallels and both systems utilize a complex branching network of neuronal cells or blood vessels to control all regions of the body. From the anatomical and juxtapositional similarities of the nervous and vascular systems, it has been suggested that axons might guide blood vessels, and vice versa. Indeed, VEGF (or VEGF‐A) from neuronal cells guides blood vessels,^{( 6 )} and signals from vessels, such as artemin and neurotrophin 3, attract axons to track alongside the vessel.^{( 19} , ^{20 )} In this manner, the neuronal and vascular systems are coordinately well organized in normal tissues; however, in tumor tissues, with a few exceptions, little innervation is observed. In adulthood, regenerated tissue responding to tissue damage is reinnervated along with angiogenesis. Although angiogenesis is commonly observed in the tumor environment and regenerated normal tissue area, why innervation does not occur in the tumor environment has not been elucidated at the molecular level. In the present study, we found one candidate that inhibits innervation in tumors, class 3 semaphorins, molecules well known to induce repulsion of axons.^{( 15} , ¹⁶ , ¹⁷ , ^{18 )}

A series of genetic and biochemical screens identified proteins acting in an instructive manner to actively attract or repel axons, and four types of molecules relating to axon guidance have been isolated, including members of the semaphorin, ephrin, netrin and slit families.^{( 21 )} In the present study, it was clear that soluble factors inhibited axon outgrowth because we used the culture supernatant of tumor cells to examine outgrowth in cultured DRG cells. Therefore, among the four kinds of axon guidance molecules described above, we excluded the ephrin family because they are membrane proteins.^{( 22 )} Moreover, two of the other families of molecules, netrins and slits, function as both attractive and repulsive cues to axons depending on their concentrations in the foci, and originally attract commissural axons to the midline of the brain.^{( 21 )} Conversely, several papers have reported semaphorin expression associated with tumor angiogenesis in the tumor environment.^{( 23} , ^{24 )} Based on this evidence, we examined semaphorin expression in tumor cells.

Semaphorins are a large family of signaling proteins, both secreted and membrane bound. They are divided into eight classes. Among these, class 3 semaphorins (Sema3A–G) are the only secreted forms in vertebrates. Moreover, all class 3 semaphorins were shown to be chemorepulsive for many class of axons, and so far Sema3A, Sema3C, Sema3D and Sema3E were found to bind Npn1 with high affinity.^{( 23 )} Among the class 3 semaphorins, Sema3A has been most intensively studied in relation to axon guidance.^{( 15} , ¹⁶ , ¹⁷ , ^{18 )} Sema3A shows repulsive activity toward a variety of neuronal types, including motor, sensory, olfactory and hippocampal neurons.^{( 25} , ²⁶ , ²⁷ , ²⁸ , ²⁹ , ^{30 )} Npn1 and Npn2 were the first receptors identified for Sema3A.^{( 31} , ^{32 )} The transmembrane protein Npn1 forms a homodimer receptor complex for Sema3A. Mice lacking Npn1 have a similar phenotype to those lacking Sema3A, namely, marked defasciculation of nerve bundles and aberrant projections of sensory nerves.^{( 33} , ^{34 )} Although Npn1 itself does not contain a kinase domain or binding site of the adaptor protein for signal transduction in the cytoplasmic domain, together with plexins it forms a functional receptor complex to induce signals into cells.^{( 13} , ^{35 )} In this way, Sema3A was studied extensively and is suggested to induce strong chemorepulsion for almost all axons. Therefore, we first tried to knockdown the Sema3A gene in C26 and B16 tumor cells, with Sema3A‐depleted C26 or B16 tumor‐conditioned media being used in DRG culture. However, merely depletion of Sema3A from tumor‐conditioned media did not alter tumor‐conditioned media‐mediated inhibition of axon outgrowth from DRG (data not shown). Consistent with the result of the Sema3A knockdown, our present data showed that almost all class 3 semaphorins were produced from tumor cells, and that soluble Npn1 protein rescued axon outgrowth resulting from tumor‐conditioned media. Therefore, we confirm that class 3 semaphorins derived from tumor cells are involved in the inhibition of axon outgrowth. However, soluble Npn1 did not completely rescue the extension of axons to the control level. It is possible that other secreted molecules, including some members of class 3 semaphorins, bind to Npn2 alone. Moreover, in the present study, we focused on soluble proteins for axon extension; however, axon outgrowth is controlled not only by secreted factors but also by cell‐to‐cell contact‐dependent mechanisms. Therefore, molecules expressed on cells within the tumor, such as endothelial cells and tumor fibroblasts, as well as tumor cells, might affect axon guidance in a cell‐to‐cell contact manner.

It is well known that blood vessels in tumors are disorganized compared to those observed in normal tissue, and permeability is relatively suppressed in the tumor environment. Recently, a new concept has emerged whereby the normalization of blood vessels in tumors allows for penetration of anticancer drugs deep into the site of the tumors.^{( 36 )} In normal tissues blood vessels are closely linked with neuronal components, suggesting that axon outgrowth accompanying angiogenesis in tumors may contribute to normalization of blood vessels in tumors. To develop new strategies for normalizing tumor blood vessels, the molecular mechanisms contributing to inhibition of axon guidance in tumors should be elucidated.

Supporting information

Supporting info item