Arl13b promotes gastric tumorigenesis by regulating Smo trafficking and activation of the Hedgehog signaling pathway

Abstract

Inhibitors of the Hedgehog (Hh) pathway transducer Smoothened (Smo) have been approved for cancer treatment, but Smo mutations often lead to tumor resistance and it remains unclear how Smo is regulated. In this study, we identified the small GTPase Arl13b as a novel partner and regulator of Smo. Arl13b regulated Smo stability, trafficking and localization which are each crucial for Hh signaling. In gastric cancer cells, Arl13b stimulated proliferation, migration and invasion in vitro and in vivo. In clinical specimens of gastric cancer, Arl13b expression correlated strongly with tumor size and depth of invasion; patients with high levels of Arl13b had a poor prognosis. Our results show how Arl13b participates in Hh pathway activation in gastric cancer.

Introduction

The hedgehog (Hh) signaling pathway was initially identified in Drosophila melanogaster(1), but its core components and regulatory mechanism are evolutionarily conserved from invertebrates to vertebrates, with few exceptions(2). In vertebrates, Hh signaling is initiated by the binding of Hh ligands (Shh, Ihh, and Dhh) to a 12-transmembrane receptor Patched (Ptch) present in receiver cells. The binding of an Hh ligand to Ptch relieves the inhibition of Smoothened (Smo) by Ptch, resulting in the translocation of Smo to the primary cilium (PC). Upon entering in the cilia, de-repressed Smo triggers the activation of zinc-finger transcription factors Gli1, Gli2, and Gli3, which then translocate to the nucleus and modulate the expression of their downstream target genes(3). Hh signaling plays an important role in body patterning and organ development during embryogenesis by regulating the proliferation, migration, and differentiation of target cells in a concentration dependent manner(4). However, the Hh signaling pathway is mostly quiescent in adult tissues(5). Aberrantly activated Hh signaling could lead to oncogenesis in various tissues or organs, including basal cell carcinoma, medulloblastoma, pancreatic cancer, colon cancer, gastric cancer and glioblastoma(6–10). Smo plays an important signal transduction role in these physiological and oncopathological events by transducing Hh signaling.

Smo is a membrane frizzled family G-protein-coupled receptor (GPCR), and its trafficking to and enrichment in the PC are key steps in the activation of the Hh signaling pathway. Smo contains an extracellular N-terminal domain, a central seven-transmembrane domain, and an intracellular C-terminal domain (C-tail)(11). The N-terminal domain and seven-transmembrane helical domain primarily function as physiologic ligand-binding sites that are critical for Smo activity(12), while the C-tail mediates the activation of Hh signaling(13–15). Post-translational modifications, such as phosphorylation or ubiquitination, and protein-protein interactions occurring on the C-tail have profound effects on Smo activity(12,15–17). In response to Hh ligands, the C-tail of Smo becomes hyperphosphorylated, which triggers a conformational change associated with the enrichment of Smo in the PC and activates downstream signals in vertebrate, whereas its ubiquitination promotes the endocytosis and degradation of Smo(18). Regulatory or effector proteins, such as β-arrestin, Costal 2, and Gprasp2, facilitate Smo activation by directly binding to its C-tail(12,15,19). Although several molecules have been identified to regulate Smo via its C-tail in response to Hh signaling, it remains unclear whether there are other yet unknown Smo regulators that are important for the Hh signaling pathway.

In an attempt to address this issue, we identified Arl13b as a novel regulator of Smo in the Hh pathway. Arl13b is an ADP-ribosylation factor (Arf)-like small GTPase that is involved in the formation and function of cilia as well as vesicle trafficking, cellular differentiation, cell movement, and cytoskeletal processes(20). Arl13b is specifically enriched in the cilia of many organisms. In the absence of Arl13b, the cilia are short and display a specific structural defect in the ciliary axoneme(21). Mutations in the human Arl13b gene result in Joubert syndrome (JS), an autosomal recessive disorder that is known to have perturbed ciliary function(22,23). Arl13b null mutant mice display the phenotypes of JS patients and show coupled defects in cilia structure and the Hh signaling pathway(24). Although one study showed that Arl13b could affect the localization of some Hh signaling components, such as the distribution of Smo within the cilium(25), it remains completely unknown how exactly Arl13b functions in response to Hh signaling, including whether it works directly with Smo in this signaling pathway and whether it plays a role in tumor formation associated with this signaling pathway.

Our study presented here was designed to address these questions. As detailed below, we found that Arl13b can promote the tumor growth of gastric cancer by directly regulating Smo trafficking and subsequent activation of the Hh signaling pathway. Our results also suggest that Arl13b could serve as a novel molecular target for developing anti-gastric cancer therapy.

Materials and Methods

Cell Line

Cell lines AGS and 293T were purchased from the American Type Culture Collection (ATCC, Manassas, VA) between 2010 and 2015. NIH3T3 was purchased from ATCC in March 2016. MKN45 and MKN28 cells were purchased from the China Centre for Typical Culture Collection at Wuhan University between 2010 and 2013. AGS, MKN28, MKN45 and 293T cells were authenticated using short tandem repeat (STR) profiling and were negative for mycoplasma contamination detecting via PCR-based assay in December 2016. These cells used in the study were cultured not more than 3 months after resuscitation and cultured as recommended by the suppliers in a humidified incubator with 5% CO₂ at 37°C.

Western Blotting and Real-time PCR

Protein extracts were obtained using extraction buffer and were analyzed via Western blotting (WB). The band intensity was analyzed with Image J software. Real-time PCR was performed with an ABI Sep Plus One sequence detection system (Applied Biosystems, Foster City, CA).

Protein-protein Interaction

To detect the interaction between Smo and Arl13b, Arl13bΔ19 GST-fusion proteins were produced in BL21, purified, and immobilized on glutathione Sepharose 4B beads (Amersham Pharmacia). The beads were then incubated with lysates from 293T cells transfected with Flag-Smo. Bead-associated proteins were subjected to SDS-PAGE and WB analysis. Furthermore, MBP-Smo(550–787) that pre-purified on amylose resin was incubated with lysates from 293T cells transfected with Flag-Arl13b, the precipitated proteins were used to analysis via WB assays.

To detect the interaction of Smo and Arl13b in mammalian cells, cells or cells transfected with indicated vectors were solubilized in the lysis buffer containing 0.5% Lubrol-PX, 50 mM KCl, 2 mM CaCl₂, 20% glycerol, 50 mM Tris-HCl, and proteases and phosphatases inhibitors (pH7.4). Pre-cleared cell lysates were incubated with 0.5 μg of indicated antibodies at 4°C overnight. The reaction was incubated with 50 μl of 1:1 slurry beads conjugated with protein G or A (Roche) for 3 hr at 4°C. Beads were washed 4 times with lysis buffer before the addition of SDS sample buffer and subjected to WB analysis as described previously (26).

To determine the direct interaction between Smo and Arl13b, MBP, MBP-Smo(550–787) and His6-Arl13bΔ19 were produced in E.coli and purified. The target proteins that purity was above 90% were subjected to further pull-down assay. Then, amylose resin contain MBP or MBP-Smo(550–787) was incubated with His6-Arl13bΔ19 at 4°C for 3 hr, and were wash 3 times with ice-cold PBS. Finally, the precipitated proteins were used to SDS-PAGE analysis and Coomassie brilliant blue staining.

Immunofluorescence and Immunohistochemistry

To detect Co-locallization of Arl13b and Smo, AGS cells were treated with N-Shh for 12 hr. To detect localization of Smo in the PC, NIH3T3 cells with or without Arl13b knockdown were treated with FBS-free DMEM for 24 hr and stimulated with N-Shh for 2 hr. Cells cultured on coverslips were fixed in 4% PFA in PBS and incubated at 4°C for 30 min with 0.1% TritonX-100. For immunostaining, fixed slides and cells were incubated at 4°C overnight with primary antibodies in PBS containing 0.5% goat serum, 1% BSA and 0.1% TritonX-100. After the samples were washed 3 times with PBS, they were incubated at room temperature for 1 hr with Alexa488-, 594- or 647-conjugated goat anti-rabbit/mouse secondary antibodies, and mounted with Vectashield mounting medium (Vector). Representative images were acquired using an LSM700 confocal microscope (Zeiss, Germany).

Immunohistochemical (IHC) staining was performed as described previously(27). Tissue sections were incubated in 10 mM sodium citrate buffer (pH 6.0) at sub-boiling temperatures for 10 min, rinsed in PBS, and incubated with 10% normal goat serum to block non-specific staining. Sections were incubated with a primary antibody (1:200) at 4°C in a humidified chamber overnight and immunoreactivity was visualized using a Polink-2 HRP DAB Detection Kit following the manufacturer’s procedure. Images were captured using an FSX100 microscope equipped with a digital camera system (Olympus). Samples were examined by 3 individual researchers to independently obtain pathological information using the German semi-quantitative scoring method. Each specimen was scored for the intensity of nucleic, cytoplasmic, and membrane staining (no staining = 0; weak staining = 1, moderate staining =2, strong staining = 3) and for the extent of stained cells (0% = 0, 1~24% = 1, 25~49% = 2, 50~74% = 3, 75~100% = 4). The final immunoreactive score was the product of the intensity score multiplied by the extent score. Consecutive sections were stained by H&E to distinguish cancer tissues from adjacent normal epithelium.

Cell Proliferation, Migration and Invasion Assay

To analyze cell proliferation, an approximately equal number of MKN45, AGS or MKN28 cells (~3×10³/well) were equivalently plated in 6-well plates (with triplicate wells for each cell type), in DMEM supplemented with 10% fetal bovine serum (FBS). The cells were cultured for up to 12–15 days and then stained with crystal violet. Dishes were graphed, and positive colonies containing more than 50 cells were counted under a microscope.

Cell migration was measured using a scratch (wounding healing) assay. Cells were plated in 6-well plates to create a confluent monolayer. Then, the monolayer was scraped in a straight line to create a scratch or wound with a p200 pipet tip. The cells were washed once with growth medium and incubated in DMEM containing 2% FBS. The remaining cells were treated with N-Shh conditional medium and imaged using a phase-contrast microscope. The wound area was quantified using NIH Image-Pro Plus software. The data are expressed as the means of four independent experiments ± standard deviation.

Cell invasion assays were performed in Transwell plates (8 μm pore size, 6.5 mm diameter; Corning Life Sciences, Lowell, MA) precoated with Matrigel Basement Membrane Matrix (1 mg/ml; BD Biosciences, Franklin Lakes, NJ) according to the manufacturer’s protocol. Briefly, 3 × 10⁴ cells in 200 μl of FBS-free medium were seeded into upper chambers. Bottom wells in the system were filled with 800 μl of N-Shh conditional medium supplemented with 1–2% FBS. After the assays had been run for 24 hr at 37°C, non-migrated or non-invaded cells were removed from the upper surface of the filter. Cells on the lower surface of the membrane were fixed with ice-cold methanol and stained with crystal violet. Cell numbers were counted under an optical microscope. Each experiment was repeated at least three times.

Flow Cytometric Analysis

293T cells were transfected with the shRNA-control or shRNA-Arl13b vector. Three days after transfection, cells were detached with enzyme free cell dissociation buffer (Gibco) and stained with an anti-Smo-N (extracellular domain) antibody and followed by staining with Alexa Fluor 488-conjugated goat anti-rabbit IgG (H+L) in PBS. The cells were then subjected to analysis of Smo cell surface level using a Beckman cytometer (MOFLO XDP).

Luciferase Assay

AGS or MKN28 cells were seeded into a 24-well plate. After the cells were cultured overnight, they were co-transfected with the pGL4.7–8×GBS reporter plasmid and the pGL4.2-TK plasmid with miRNAi-Arl13b or miRNAi control, and Arl13b-myc or myc vector, respectively. Luciferase assays were performed 48 hr after transfection using a Dual Luciferase Reporter Assay System (Promega, WI).

Lentivirus Infection and Xenografts

The Lenti-X-shRNA Tet-On construct (pGV307-RFP) for shRNA-Arl13b knockdown was generated, packed, and purified by GeneChem (Shanghai, China). The shRNA target sequence was 5′-CAGATAGAACCATGT-3′. Lentivirus infection was done according to the protocol provided by the manufacturer. Briefly, 5 × 10⁴ MKN45, SGC7901 or MKN28 cells were incubated with 1 × 10⁸ IU of virus and 8 μg/ml polybrene (Sigma-Aldrich, St. Louis, MO) for 12 hr. Cells were induced in 2 μg/ml doxycycline for 48 hr and followed by 5μg/ml puromycin 14 days after infection to select stably infected cells.

For in vivo experiments, 2 × 10⁷ stably infected MKN45, SGC7901 (Lenti-control (Vector) and Lenti-shRNA-Arl13b (Arl13b-KD)) or MKN28 (Lenti-control (Vector) and Lenti-Arl13b (Arl13b)) cells were resuspended in sterile PBS (200 μl) and injected subcutaneously into both flanks of 5-week-old female BALB/c-nu mice (SLAC Laboratory Animal CO. Ltd, Hunan, China). One week after injection, the mice were administered 2 μg/ml doxycycline and 5% sucrose in sterile drinking water. The doxycycline-containing water was replenished every 3 days. Tumor sizes in both flanks of the mice were measured using Vernier calipers thrice weekly, and the tumor volume was calculated using the formula volume, V = (L × W²)/2. After 4 weeks, the xenografts were harvested for IHC and Western blot analysis. Eight female nude mice (4–5 weeks old) were included in each group. All animal experiments were approved by the Ethical Committee of the First Affiliated Hospital of Nanchang University (Permit Number: 2011-021). All surgeries were performed under sodium pentobarbital anesthesia, with minimized suffering.

Human Tissue Specimens

Human specimens were retrieved via surgical intervention without prior radiotherapy or chemotherapy. All samples were collected at the First Affiliated Hospital of Nanchang University between January 2009 and September 2013. The inclusion criteria are described in the Supplementary Information. The study protocol was approved by the Institutional Review Board of the Frist Affiliated Hospital of Nanchang University. Detailed clinical and pathological information for the patients is summarized in Table 1.

Table 1.

Association of Arl13b and Smo expression levels with different clinicopathologic characteristics in gastric cancers.

Clinicopathologic		Arl13b expression					Smo expression
		low		High		P value	low		High		P value
Measurement data	n	Mean	SD	Mean	SD		Mean	SD	Mean	SD
Age(y)	154	58.25	12.14	57.98	15.35	0.898	58.32	10.510	58.08	13.421	0.924
Tumor size (cm)	154	3.75	2.05	4.49	1.99	0.026*	3.25	1.869	4.30	2.053	0.006**
Enumeration data	n	Count	n%	Count	n%	P	Count	n%	Count	n%	P
Gender
Male		58	37.70%	42	27.30%		25	16.23%	75	48.70%
Female		29	18.80%	25	16.20%		9	5.84%	45	29.22%
Total	154	87	56.50%	67	43.50%	0.608	34	22.08%	120	77.92%	0.234
Degree of Differentiation
Well		2	1.30%	1	0.60%		1	0.65%	2	1.30%
Moderately		36	23.40%	18	11.70%		11	7.14%	43	27.92%
Poor		49	31.80%	48	31.20%		22	14.29%	75	48.70%
Total	154	87	56.50%	67	43.50%	0.125	34	22.08%	120	77.92%	0.719
T factor
T1		18	11.70%	6	3.90%		12	7.79%	12	7.79%
T2		13	8.40%	2	1.30%		7	4.55%	8	5.19%
T3		53	34.40%	56	36.40%		14	9.09%	95	61.69%
T4		3	1.90%	3	1.90%		1	0.65%	5	3.25%
Total	154	87	56.50%	67	43.50%	0.005**	34	22.08%	120	77.92%	0.000**
Lymph node metastasis(N factor)
N0		29	18.80%	16	10.40%		14	9.09%	31	20.13%
N1		32	20.80%	19	12.30%		14	9.09%	37	24.03%
N2		15	9.70%	23	14.90%		5	3.25%	33	21.43%
N3		11	7.10%	9	5.80%		1	0.65%	19	12.34%
Total	154	87	56.50%	67	43.50%	0.091	34	22.08%	120	77.92%	0.04*
Neural invasion
Yes		34	22.10%	34	22.10%		8	5.19%	60	38.96%
No		53	34.40%	33	21.40%		26	16.88%	60	38.96%
Total	154	87	56.50%	67	43.50%	0.148	34	22.08%	120	77.92%	0.006**
Vascular invasion
Yes		27	17.50%	30	19.50%		7	4.55%	50	32.47%
No		60	39.00%	37	24.00%		27	17.53%	70	45.45%
Total	154	87	56.50%	67	43.50%	0.08	34	22.08%	120	77.92%	0.025*
Distant invasion
Yes		5	3.20%	5	3.20%		1	0.65%	9	5.84%
No		82	53.20%	62	40.30%		33	21.43%	111	72.08%
Total	154	87	56.50%	67	43.50%	0.748	34	22.08%	120	77.92%	0.461
Clinical stage
stage I		24	15.60%	7	4.50%		14	9.09%	17	11.04%
stage II		11	7.10%	12	7.80%		7	4.55%	16	10.39%
stage III		35	22.70%	30	19.50%		10	6.49%	55	35.71%
stage IV		17	11.00%	18	11.70%		3	1.95%	32	20.78%
Total	154	87	56.50%	67	43.50%	0.063	34	22.08%	120	77.92%	0.001**

Statistical Analysis

Unless otherwise indicated, the data were expressed as the mean ± SD from experiments performed at least three times. Differences between two groups were assessed with Student’s t-test or one-way ANOVA. Differences were considered significant if p<0.05. All analyses were carried out using SPSS v.13.0 software (SPSS Inc., Chicago, IL).

Results

Identification of Arl13b as a Smo-interacting Partner

Although Smo is well established to play a central role in transducing Hh signaling(12), the mechanism(s) underlying its signal transduction function still remains poorly understood. To address this issue, we searched for Smo-interacting partners by performing a yeast two-hybrid screen using the functionally important C-tail of Smo(15) as a bait. From this screen, we found three positive clones that encode two different N-terminal fragments of Arl13b, indicating that Arl13b could be a potential Smo-interacting partner. To confirm this hypothesis, we performed an in vitro GST pull-down assay using Flag-Smo expressed in 293T cells and purified GST-Arl13b lacking its N-terminal 19 amino acids (GST-Arl13bΔ19). We used GST-Arl13bΔ19 because its full-length form is often difficult to be purified from bacteria due to this N-terminal hydrophobic region(28). As shown in Fig. 1A, recombinant Smo was pulled down with GST-Arl13bΔ19. This Smo-Arl13b interaction was direct, as it was confirmed by using another pull-down assay with purified MBP-Smo (residues 550–787) (MBP-Smo (550–787)) and His6-Arl13bΔ19 (Fig. 1B).

Figure 1. The interaction between Arl13b and Smo.

(A) Arl13b interacts with Smo. (B) Arl13b binds Smo directly. Purified MBP-Smo(550–787) immobilized on amylose resin was incubated with pre-purified His6-Arl13bΔ19. (C–D) Interaction of Arl13b with Smo in mammalian cells. (E) Endogenous Arl13b binds to Smo in gastric cancer cells. (F) N-Shh stimulation increases the interaction of endogenous Arl13b with Smo in gastric cancer cells. (G) Co-localization of Arl13b and Smo. Cells were co-stained with antibodies against Arl13b (Alexa Fluor 594, red) and Smo (Alexa Fluor 488, green). Arrow indicates co-localization.

To further validate the Arl13b-Smo interaction in mammalian cells, we performed co-immunoprecipitation (IP) experiments and found that Flag-Smo and GFP-Arl13b were co-IP with each other in 293T cells (Fig. 1C and 1D). Consistent with these results, endogenous Smo protein could be immunoprecipitated with the Arl13b antibody (Fig. 1E), indicating that Smo and Arl13b can form a protein complex in cells. This interaction between endogenous Smo and Arl13b was markedly induced by Hh ligands (Fig. 1F). Finally, immunofluorescence staining showed that Arl13b and Smo co-localized in the cytoplasm of the cells (Fig. 1G), indicating that binding of these proteins occurs in the cytoplasm.

Next, we tried to map their binding domains using different fragments of these two proteins. Smo contains three major segments (Fig. 2A). In agreement with the results from the yeast two-hybrid screening, MBP-Smo(550–787), but not MBP alone, was able to pull down full-length Arl13b (Fig. 2B), suggesting that the C-tail of Smo is sufficient for Arl13b-binding. This result was verified in mammalian cells transfected with the plasmids encoding the C-tail of Smo and full-length Arl13b (Fig. 2C). Arl13b is comprised of an N-terminal GTP-binding domain, a central coil-coiled domain, and a C-terminal proline rich domain(28) (Fig. 2D). Using several truncated Arl13b proteins (Fig. 2D), we performed a set of co-IP-WB experiments and found that while the coil-coiled domain and proline rich region of Arl13b were dispensable for its interaction with Smo, the GTP-binding domain was essential for Arl13b to form a complex with Smo, indicating that Arl13b interacts with Smo through its GTP-binding domain (Fig. 2D and 2E). An in vitro GST pull-down assay validated the binding of this GTP-binding domain(20–150) to the C-terminus of Smo (Fig. 2F).

Figure 2. Mapping the domains responsible for the Smo-Arl13b interaction.

(A) Schematic illustration of the domains of Smo and its association with Arl13b. (B–C) C-terminal tail of Smo interacts with Arl13b. (D) Diagrammatic representation of Arl13b and various deletions used to determine the Smo binding domain. (E) Identification of the Arl13b domains responsible for Smo interaction. Lysates from 293T cells transfected with Flag-Smo and various GFP-tagged Arl13b derivatives were subjected to IP with anti-Flag antibody. Asterisks indicate the domains interacted with Smo. (F) N-terminal region of Arl13b interacts with C-terminal tail of Smo. (G–H) The Smo-Arl13b interaction was influenced by the GTPase activity of Arl13b. 293T cells lysates cotransfected with wild type (WT), G28V mutant or R79Q mutant of GFP-Arl13b and Flag-Smo(550–787) or empty vector were immunoprecipitated with anti-Flag antibody. Data was shown as mean ± SD (n = 3). **p<0.01.

Since the N-terminal Arl13b is responsible for its GTPase activity, we generated the active G28V mutant(23) and inactive R79Q mutant(29) of Arl13b to evaluate whether the GTP-binding site is critical for the interaction between Arl13b and Smo. Indeed, R79Q showed a reduced interaction with Smo, whereas G28V increased the Arl13b-Smo binding (Fig. 2G and 2H). Correlated with this result, mycophenolic acid (MPA), an inhibitor of inosine monophosphate dehydrogenase (IMPDH), which can reduce intracellular GTP levels(30), was found to inhibit the interaction between Arl13b and Smo (Supplemental Fig. S1). Taken together, these results identify Arl13b as a novel Smo-binding protein and demonstrate that the C-tail of Smo directly interacts with the GTP-binding domain of Arl13b.

Arl13b Promotes Hh Signaling Through Smo Stabilization

Next, we determined the functional outcome(s) of the Arl13b-Smo interaction. Since the Arl13b-Smo interaction is enhanced upon N-Shh stimulation (Fig. 1F), we first tested whether Arl13b could affect the level of Smo. Interestingly, titrating the amount of ectopic Flag-Arl13b led to an increase of endogenous Smo protein levels in a dose-dependent manner (Fig. 3A), but alteration of Arl13b did not affect Smo transcription (Supplemental Fig. S2A and S2B), suggesting that Arl13b may regulate Smo stability by directly binding to this protein. To test this conjecture, we performed an Arl13b knockdown experiment followed by analysis of the half-life of Smo. Knockdown of Arl13b led to a marked decrease of the Smo’s half-life from ~6 hr to less than 2 hr (Fig. 3B, 3C and Supplemental Fig. S2C). Additionally, proteasome inhibitor MG-132, but not lysosomal inhibitor Monensin, increased the level of Smo protein (Supplemental Fig. S2D). Consistently, knockdown of Arl13b led to an increase in the ubiquitylation of Flag-Smo, which was more apparent in the presence of MG-132 (Fig. 3D). These results indicate that Arl13b protects Smo from ubiquitination and ubiquitination-dependent proteolysis. Intriguingly, The N-terminal Smo-binding domain of Arl13b (residues 1–150) reduced the Smo protein levels in a dose-dependent manner (Fig. 3E) and inhibited the proliferation of gastric cancer cells (Fig. 3F and 3G), suggesting that binding to Smo itself is not sufficient to protect Smo from ubiquitination-mediated degradation; instead, this binding might prevent the protective effect of full-length Arl13b on Smo in a dominant-negative fashion.

Figure 3. Arl13b enhances Smo stability, trafficking and activity.

(A) Overexpression of Arl13b increases the Smo protein level. 293T cells were transfected with indicated dozes of Flag-Ar13b. (B–C) Arl13b knockdown accelerates the degradation of Smo protein. 293T cells expressing miR-control or miR-Arl13b-478 were treated with cycloheximide (CHX) for indicated times. KD, knockdown. Data was shown as mean ± SD (n= 3). **p<0.01. (D) Ubiquitylation- and proteasome-dependent degradation of Smo. AGS cells cotransfected with Flag-Smo and miR-control or miR-Arl13b-478 were treated with or without 50 μM MG-132. (E) Arl13b(1–150) functions as a dominant negative modulator on the stabilization of Smo. 293T cells were transfected with indicated doses of GFP-Arl13b(1–150). (F–G) Arl13b(1–150) significantly reduces the colony formation of gastric cancer cells. AGS cells transfected with GFP-Arl13b, GFP-Arl13b(1–150) or GFP were sorted and used for colony formation assays. Data was shown as mean ± SD (n= 3). **p<0.01. (H) Knockdown of Arl13b reduces Smo surface expression. (I–J) Arl13b regulates localization of Smo in the PC. Cells were co-stained with antibodies against Ac-tubulin (Ac-Tub) (Alexa Fluor 594, red) and Smo (Alexa Fluor 647, green). Arrow indicates PC. Data was shown as mean ± SD (n=40). **p<0.01.

We then determined whether Arl13b might affect Smo trafficking. Immunofluorescence staining revealed that Smo was enriched in the cytoplasmic membrane of Arl13b-expressed cells, but reduced on the surface of Arl13b-depleted cells (Supplemental Fig. S2E), which was confirmed by flow cytometric analysis (Fig. 3H). Also, knockdown of Arl13b reduced the localization of Smo to the PC (Fig. 3I, 3J and Supplemental Fig. S2F). We further evaluated whether Arl13b regulates the effects of Smo inhibitors, the expression of the Hh signaling target genes Gli1 and Bcl2 were detected, and we found that Smo inhibitors were ineffective in the presence of overexpression of Arl13b (Supplemental Fig. S2G and S2H). These results indicate that Arl13b can mediate the trafficking of Smo and its activity.

It was reported that mice lacking Arl13b exhibit abnormal Hh signaling(21), and Arl13b is required for the entry of Smo into the cilium via unknown mechanisms(25). We found Arl13b regulates the expression of Gli1 in NIH3T3 cells which is Hh responsive (Supplemental Fig. S3A and S3B). We further determined whether Arl13b affects the Hh signaling response in gastric cancer cells. Indeed, knockdown of Arl13b reduced, but overexpression of Arl13b enhanced luciferase expression driven by a Gli2-binding sites-containing promoter (Supplemental Fig. S3C and S3D). In line with these results, the expression of several Hh signaling components, its downstream targets and the epithelial mesenchymal transition (EMT)-related proteins was reduced in Arl13b-depleted cells, but up-regulated in Arl13b-overexpressed cells with or without Hh treatment (Supplemental Fig. S3E and S3F). These results indicate that Arl13b not only regulates the activity of Smo, but also influences the Hh signaling response, and suggest that Arl13b might play a role in the proliferation, migration and invasion of cancer cells.

Arl13b Facilitates Shh-induced Malignancy of Gastric Cancer Cells

Although Hh signaling has been shown to be crucial for gastric cancer development and metastasis(31), the role of Arl13b in these oncogenic processes remains unknown. To test this, we employed two cell lines AGS and MKN45, which contain higher levels of Arl13b (Supplemental Fig. S4A). Knockdown of Arl13b markedly reduced the expression of Gli2, Gli1, and Smo as well as that of EMT-related proteins (Fig. 4A), suggesting that Arl13b might be required for the proliferation, migration and invasion of these cancer cells. Indeed, knockdown of Arl13b dramatically blocked the N-Shh-induced wound healing of MKN45 cells, and also inhibited N-Shh-mediated invasion of the cells (Fig. 4B–4E). These results were also reproduced using AGS cells (Supplemental Fig. S4B–S4E). Additionally, Arl13b knockdown significantly retarded the N-Shh-induced colony formation and growth of MKN45 cells (Fig. 4F, Supplemental Fig. S4F and S4G). These results indicate that Arl13b is critical for the proliferation, migration and invasion of gastric cancer cells.

Figure 4. Arl13b increases Shh-induced gastric cancer migration and invasion.

(A) Arl13b regulates the levels of Hh pathway components and EMT-related proteins in gastric cancer cells. Lysates of MKN45 contorl (Vector) or Arl13b stable knockdown (Arl13b-KD) cells were subjected to WB. (B–C) Arl13b knockdown reduces N-Shh-induced migration of MKN45 cells. Images were acquired at indicated time. Data was shown as mean ± SD (n=3). **p<0.01. (D–E) Knockdown of Arl13b decreases N-Shh-stimulated invasion of MKN45 cells. Data was shown as mean ± SD (n=4). **p<0.01. (F) Arl13b knockdown inhibits the N-Shh-stimulated colony formation of MKN45 cells. The data was shown as mean ± SD (n=3). **p<0.01. (G–H) Arl13b overexpression enhances the N-Shh-induced migration of MKN28 cells. Data was shown as mean ± SD (n=3). **p<0.01. (I–J) Arl13b overexpression significantly increases the N-Shh-stimulated invasion of MKN28 cells. Data was shown as mean ± SD (n=4). **p<0.01. (K) Arl13b overexpression increases N-Shh-stimulated colony formation in MKN28 cells. Data was shown as mean ± SD (n=3). **p<0.01.

Conversely, overexpression of Arl13b using the lentivirus system in MKN28 cells (Supplemental Fig. S5A) drastically increased the percentage of wound repair with N-Shh treatment of the cells (Fig. 4G and 4H). Also, Arl13b overexpression significantly enhanced N-Shh-induced cell invasion, colony formation and growth (Fig. 4I–4K, Supplemental Fig. S5B and S5C). Altogether, these results demonstrate that Arl13b plays a crucial role in the Hh signaling-mediated proliferation, migration and invasion of gastric cancer cells.

Arl13b Accelerates Gastric Carcinogenesis In Vivo

To determine whether the cellular phenotypes described above could be reproduced in a more biologically significant context, we generated xenograft tumor models using gastric cancer cell lines MKN45, SGC7901 and MKN28. These cell lines expressing scrambled shRNA, Arl13b-shRNA(Arl13b-KD), or ectopic Arl13b were subcutaneously injected into the flanks of nude mice. Knockdown of Arl13b led to dramatic decreases of the average tumor volume and the average tumor weight compared with the control group(Fig. 5A–5C). In line with these results, the expression of Gli2, Gli1, Smo, snail, N-cadherin, Vimentin and p-FAK was dramatically reduced in the tumor tissues of the Arl13b-KD group (Fig. 5D, left panels), similar to the results obtained in vitro (Fig. 4A). The IHC results indicated that Ki-67 strictly associated with cell proliferation, and MMP9 involved in cancer invasion and metastasis were significantly decreased in Arl13b knockdown xenograft tumors (Fig. 5E). Furthermore, we observed similar results in tumor xenograft of SGC7901 cells (Supplemental Fig. S6). Consistent with these results, the overexpression of Arl13b resulted in a dramatic increase in both tumor volume (by ~3-fold) and tumor weight (by ~2-fold) compared with that of the control group (Fig. 5F–5H). Consistently, overexpression of Arl13b significantly increased the level of Gli2, Gli1, Smo, snail, N-cadherin, Vimentin and p-FAK in Arl13b xenograft tumors (Fig. 5D, right panels). The levels of Smo, Ki67 and MMP9 were also markedly enhanced in ectopic Arl13b-expressed xenograft tumors (Fig. 5E). These results demonstrate that Arl13b plays a critical role in driving the growth of xenograft gastric tumors by mediating the Hh signaling pathway.

Figure 5. Arl13b promotes gastric carcinogenesis in vivo.

(A–C) Arl13b knockdown inhibits tumor growth. Eight nude mice were injected subcutaneously with 2×10⁷ cells/mouse for each of the indicated stable cell lines of MKN45-Arl13b-KD. Results were presented as tumor volume (A), isolated tumors (B), and tumor weights (C). Data was shown as mean ± SD (n=8). *p<0.05, **p<0.01. (D) Arl13b regulates levels of Hh pathway components and EMT related proteins in vivo. (E) Arl13b regulates metastasis and proliferation of tumor cells. Tumors were isolated, fixed, and employed to IHC assays. (F–H) Overexpression of Arl13b promotes tumor growth. Eight nude mice were injected subcutaneously with 2×10⁷ cells/mouse for each the indicated stable cell lines of MKN28-Arl13b. Results were represented as tumor volume (F), isolated tumors (G), and tumor weights (H). Data was shown as mean ± SD (n=8). *p<0.05, **p<0.01.

Arl13b Expression Levels Correlate with the Clinical Pathogenesis of Gastric Cancer

To translate the above described oncogenic role of Arl13b in gastric cancer development and progression to clinical significance, we recruited 154 eligible gastric cancer patients and analyzed the expression of Arl13b and Smo, as well as the correlation between their expression and the stage or progression of gastric cancers in pairs of tumor specimens. Firstly, we evaluated the specificity of the antibodies used in IHC assays, and found that anti-Arl13b and anti-Smo antibodies could effectively and specifically recognize the indicated proteins (Supplemental Fig. S7). Then, we found Arl13b and Smo protein levels were significantly increased in gastric tumors compared with the paired adjacent tissues via IHC analysis (Fig. 6A–6C), consistent with the results from the xenograft experiments described above (Fig. 5). In 8 pairs of representative tumor and adjacent tissues, the levels of Arl13b and Smo were revealed to be much higher than that in adjacent gastric tissues (Fig. 6D). These results indicate that both Arl13b and Smo are highly expressed in human gastric cancers.

Figure 6. High levels of Arl13b and Smo are closely related to poor survival in gastric cancer patients.

(A–C) Arl13b and Smo are overexpressed in gastric cancer tissues compared with adjacent tissues as examined by IHC. (A) The representative images with magnified local images reflecting detailed information were shown on the right. Plotting the Arl13b and Smo scores in each gastric carcinoma and adjacent tissues (B) and box plots of the scores of Arl13b and Smo expression (C) are shown. Statistical significance was analyzed using the Mann-Whitney U test, n=154. (D) Arl13b and Smo are highly expressed in gastric cancer tissues. Proteins isolated from gastric cancer and adjacent nontumorous tissues from 8 patients were subjected to WB. A, adjacent; T, tumor. (E–H) The high levels of Arl13b and Smo are correlated with poor survival of patients with gastric cancer. Univariate analysis was carried out to analyze disease-free survival and overall survival of patients with gastric cancer. (I) Hypothetical model of Arl13b promotes tumorigenesis via mediating Smo trafficking and activating of Hh signaling. In the absence of Shh (upper panel), Ptch enriches in and around the PC and inhibits Smo translocation to PC. The repressed form of Gli (GliR) cleaved by proteasome accumulates and enters the nucleus to block expression of target genes. In the presence of Shh ligand (down panel), Arl13b strongly interacts with Smo facilitating trafficking of Smo to the PC. PC localized Smo induces accumulation of Gli activated form (GliFL) in the cytoplasm and nucleus, results in activation of target genes, and then promotes the proliferation of cancer cells.

Next, we analyzed the correlation of the expression of Arl13b and Smo with the survival rates of patients with gastric cancers using the Kaplan-Meier method. As shown in Fig. 6E–6H, univariate analysis of tumor samples for disease-free survival and overall survival revealed that the group of patients exhibiting lower expression of Arl13b or Smo displayed a better prognosis (Fig. 6E–6H). We then analyzed the correlation between the expression of Arl13b or Smo and the clinical and pathological features of patients with gastric cancers. We used scores of 0 to 12 to estimate the expression of Arl13b or Smo, defining a score of less than 6 as low expression, and a score equal to or greater than 6 as high expression. Similar to the case for Smo (Table 1), the higher expression of Arl13b in gastric cancer samples was strongly correlated with both tumor size (p=0.026) and the depth of invasion (p=0.005). But, in contrast to the case for Smo, little correlation was observed between the expression of Arl13b and tumor TNM stage (Table 1).

In line with the above IHC results, analysis of the GEO database also showed much higher expression of Arl13b at the RNA level in gastric cancer tissues than that in paired normal tissues (Supplemental Fig. S8A and S8B). Based on gene expression patterns, gastric cancers are classified into three subtypes: metabolic, proliferative and mesenchymal, which have features of 5-fluorouracil sensitivity, genomic instability and cancer stem cells, respectively(32). Arl13b was found to be highly expressed in the proliferative and mesenchymal subtypes of gastric cancers (Supplemental Fig. S8C). Hence, taken together, these results demonstrate that Arl13b is highly associated with the progression of human gastric cancer and could be a biomarker for the prediction of prognosis.

Discussion

The seven-transmembrane receptor Smo is a key player in the Hh signaling pathway. However, it remains poorly understood how Smo is regulated in response to Hh signaling. The present study identified Arl13b as a novel regulator and partner of Smo via directly binding to this protein. As further discussed below, Arl13b can regulate the trafficking and subsequent activation of Smo in response to Hh signaling, and play a crucial role in promoting the progression of gastric cancer (Fig. 6I).

Arl13b Physically Interacts with Smo

Although previous studies have shown that both Smo and Arl13b are required for Hh signaling, and Arl13b regulates the distribution of Smo within the PC(25), it remains completely unknown how Arl13b regulates the localization of Smo and whether these protiens directly interact with each other. Using the C-terminus of human Smo as bait in yeast two-hybrid screening, we identified Arl13b as a potential novel Smo-interacting protein (Fig. 1). Our further biochemical and cellular analyses not only validated the physical interaction between Smo and Arl13b (Fig. 1B), but also mapped their binding domains to the C-tail of Smo and the N-terminal GTP-binding motif of Arl13b (Fig. 2). Remarkably, this interaction was responsive to Hh signaling.

Our study further demonstrated that the GTP-binding motif is critical for the interaction between Arl13b and Smo using two point mutations in Arl13b. The small GTPases of the Ras superfamily contain a P loop (GXXXXGKS/T) that is required for guanine nucleotide binding(33). A previous study showed that substitution of the first G with V frequently results in decreased GTP hydrolysis and generates a dominant-active form(23). Similarly, we introduced a mutation at the corresponding site of Arl13b and generated a G28V mutation. We also made an R79Q substitution in the N-terminus of Arl13b, which was previously identified in JS patients to interfere with GTP binding in a dominant-negative fashion(29). Using these mutants of Arl13b, we showed that the Arl13b-Smo interaction is dependent on its GTPase activity, as the G28V mutation enhanced the Arl13b-Smo interaction, while the R79Q mutation suppressed this interaction (Fig. 2G and 2H). Also, an IMPDH inhibitor MPA that can decrease cellular GTP levels reduced this interaction. These results indicate that the GTPase domain is pivotal for the interaction of Arl13b with Smo, and suggest that this interaction may serves as a target site for screening inhibitors that could be developed into a potential anti-cancer therapy as further discussed below.

Arl13b Regulates Smo Trafficking and Hh Signaling Activity

Our functional studies of the Arl13b-Smo interaction revealed that Arl13b can markedly enhance the stability of Smo (Fig. 3A). By binding to Smo, Arl13b blocked its ubiquitination and thus prevented its degradation (Fig. 3D). Arl13b appeared to inhibit Smo degradation through a proteasome-dependent mechanism, but not via the lysosome-dependent mechanism, in mammalian cells. Since activated Smo is often stabilized on the plasma membrane in Drosophila or on the PC in vertebrates(34), our results suggest that Arl13b might affect the trafficking of Smo in mammalian cells.

Indeed, this was the case, as Arl13b promoted the plasma membrane localization of Smo and inhibited Smo ubiquitination that mediates the endocytic trafficking of Smo (Fig. 3H and Supplemental Fig. S2E). In vertebrates, trafficking of Smo to the PC is essential for Smo signal transduction(13). In the presence of Hh ligand, Smo is transported to and enriched in cilia, triggering the activation of Hh signaling. There are three models proposed for Hh ligand induced Smo transport to the PC, (1) direct pathway: direct trafficking of Smo from the Golgi to the base of the cilium; (2) lateral transport pathway: transport to the cell surface followed by lateral transport to the cilium; (3) recycling pathway: surface localization followed by internalization into a recycling pathway(35). Previously, Smo was shown to move from the plasma membrane to the ciliary membrane through the lateral transport pathway(35). A later study showed that Arl13b regulates Smo trafficking to the PC via the endocytic recycling pathway(20). Together with our results as presented here (Fig. 3I and 3J), these studies suggest that Arl13b may mediate lateral transport of Smo from the plasma membrane to the PC by directly binding to Smo and inhibiting its ubiquitination and degradation.

Our study establishes Arl13b as a positive regulator of Hh signaling, which is consistent with the previous finding that mice lacking Arl13b display aberrant Hh signaling(21). Interestingly, our study also showed that the N-terminal domain of Arl13b (residues 1–150) acts as a dominant-negative modulator as it competed with full length Arl13b for Smo-binding, leading to the degradation of Smo (Fig. 3E). Hence, our results suggest that targeting the Arl13b-Smo interaction with peptides or small compounds may be a good approach to inactivate the Hh signaling.

Aberrant Hh signaling has been implicated in various cancers, including gastric cancer, basal cell carcinoma and pancreatic cancer. Activating mutations of Smo (e.g., V321M, L412F, F460L, W535L and R562Q) have strong effects on its activity and can drive the formation of basal cell cancer and medulloblastoma(36–38). Several Smo inhibitors, such as cyclopamine, IPI-926, GDC-0449 and LDE-225, are currently in clinical trials for various cancer treatments(39–42). GDC-0449 has been approved for the treatment of locally advanced and metastatic basal cell cancer and medulloblastoma(43). Cyclopamine potently inhibits the proliferation of gastric cancer cells, which express high levels of Smo(9). However, a de novo mutation in Smo (W535L, SmoM2) results in tumor resistance against GDC-0449 or other Smo inhibitors(37). Upregulation of Arl13b significantly inhibited the activity of cyclopamine (Fig. S2H). Identification of peptides or small compounds that interfere with the Arl13b-Smo interaction might offer a way to overcome the resistance caused by Smo mutation, and to develop a new molecule-targeted therapy against Smo-related cancers by synergizing the effect of Smo inhibitors.

Arl13b Serves as a Biomarker of Cancer

Although the PC, a slender microtubule-based subcellular organelle, projects on the surface of most mammalian cells and plays key roles in vertebrate development and tissue homeostasis, this organelle is also highly involved in cancers and inherited human diseases, such as cystic kidney disease(44). Genetic studies in mice have shown that the PC is essential for Hh signaling transduction(45). Arl13b is a marker of the PC. Recent studies have demonstrated that Arl13b regulates cell migration and cell cycle progression(46,47). However, it remains unclear how exactly Arl13b functions in cilia and whether Arl13b plays a role in tumorigenesis. Our studies as presented here not only illustrate the biochemical function of Arl13b by directly interacting with Smo, stabilizing this Hh signaling protein, and mediating Hh signaling as discussed above, but also unveil the oncogenic role of this GTPase protein in gastric cancer.

Gastric cancer is one of the most frequent and fatal malignancies worldwide(48). It was reported that abnormal activation of Hh signaling plays a role in the progression of gastric cancer(31). In this study, by investigating the effects of Ar13b on the progression of gastric cancer cells in vitro and in vivo, we found that knockdown of Arl13b leads to a significant reduction in the proliferation, migration and invasion of gastric cancer cells, but overexpression of Arl13b causes a marked increase in the proliferation, migration and invasion of gastric cancer cells in vitro and in vivo (Fig. 4 and 5). Remarkably, by analyzing clinical gastric cancer specimens, we found that Arl13b levels are highly correlated with tumor size and the late stage of invasion in gastric cancers. More remarkably, patients with a higher level of Arl13b display a poor prognosis (Fig. 6E–6H and Table 1).

Our further analysis showed that the RNA level of Arl13b is increased in the proliferative and mesenchymal subtypes of gastric cancers, compared with the metabolic subtype. Correspondingly, genes critical for cell cycle and DNA replication were up-regulated in the proliferative subtype, and the mesenchymal subtype has the features of cancer stem cells with aberrantly activated Hh signaling(32). Together, our results suggest that Arl13b could serve as a biomarker for the late stage of gastric cancers, especially Hh signaling associated ones. Although additional studies are needed to understand how the level and/or activity of Arl13b is regulated in these gastric cancers, our present study offers useful information for future precision oncology with a new biomarker for this type of cancer.