Lipid raft interacting galectin 8 regulates primary ciliogenesis

Primary cilium is a specialized sensory organelle that transmits environmental information into cells. Its length is tightly controlled by various mechanisms such as the frequency or the cargo size of the intraflagellar transport trains which deliver the building materials such as tubulin subunits essential for the growing cilia. Here, we show the sialoglycan interacting galectin 8 regulates the process of primary ciliogenesis. As the epithelia become polarized, there are more galectin 8 being apically secreted and these extracellular galectin 8 molecules apparently bind to a lipid raft enriched domain at the base of the primary cilia through interacting with lipid raft components, such as GD3 ganglioside and scaffold protein caveolin 1. Furthermore, the binding of galectin 8 at this critical region triggers rapid growth of primary cilia by perturbing the barrier function of the transition zone (TZ). Our study also demonstrates the functionality of this barrier depends on intact organization of lipid rafts at the cilia as genetically knockout of Cav1 and pharmacologically inhibition of lipid raft both phenocopy the effect of apical addition of recombinant galectin 8; that is, rapid elongation of primary cilia and redistribution of cilia proteins from TZ to the growing axoneme. Indeed, as cilia elongated, endogenous galectin 8, caveolin 1, and TZ component, TMEM231, also transited from the TZ to the growing axoneme. We also noted that the interaction between caveolin 1 and TMEM231 could be perturbed by exogenous galectin 8. Taken together, we proposed that galectin 8 promoted primary cilia elongation through impeding the barrier function of the TZ by interfering with the interaction between caveolin 1 and TMEM231.


| INTRODUCTION
Primary cilium is a solitary protruding organelle found at the apical surface of most cells.Ultrastructurally, it can be distinguished as a pole-like structure composed of a so-named 9+0 organized microtubule cylinder axoneme that emanates from a centriole-derived structure called the basal body. 1 The primary cilium serves as a specialized sensory hub where intensive signaling can be transduced from outside environmental stimuli to influence various cellular behaviors such as cell cycle progression.Although complex morphology exists in cilia of specific cell types, primary cilia in the majority of cells possess a plain rod-like appearance; therefore, an apparent morphological attribute in most cilia is its length, which is tightly modulated in response to environmental factors such as fluid-induced shear force. 2 Cilia formation is governed by a conserved intraflagellar transport (IFT) process which is mediated by the balanced interaction between anterograde motor kinesin 2 and retrograde motor dynein with the IFT-B and IFT-A protein complexes. 3,46][7][8][9] Most of these models suggest the existence of a molecular gate governing the entry rate of IFTdependent cargo from the cytosol into the cilia.Despite the detailed biochemical nature of this hypothetic gate remains elusive, the subcellular location of this gate is believed to be at the transition zone (TZ) region of primary cilia. 10,11he TZ is defined as the region at the base of the primary cilium where transition fibers/distal appendages anchor the basal body to the ciliary base.Despite a great number of evidence supporting the role of TZ as a diffusion barrier that can efficiently prevent the free mixing of ciliary components with other membranous and soluble molecules from subcellular regions elsewhere of cilia, the biochemical composition of TZ is not well characterized.It has been reported that the TZ membrane is the most detergent resistant of all surface membranes in Chlamydomonas reinhardtii 12 suggesting that a high-ordered lipid phase resides at the TZ.This characteristic is reminiscent of the lipid rafts existing at other cellular membranes.However, whether lipid rafts also play a regulatory role in the TZ region to govern the length of primary cilia remains an interesting issue to be explored.
In the current study, we provided reinforcing evidence that lipid rafts like domain exists at the base of the primary cilium and plays a regulatory role in controlling the length of primary cilia.Furthermore, galectin 8 (Gal8), a ganglioside-specific beta-galactoside binding lectin type protein, can be secreted from polarizing epithelia and promote ciliogenesis, through its interaction with caveolin 1 and ganglioside, two of the instrumental components of lipid rafts.

| Cell culture
Madin-Darby canine kidney (MDCK) type II cells were cultured in DMEM supplemented with 10% fetal bovine serum and penicillin/streptomycin/ampicillin at 37°C in a humidified incubator containing 5% CO 2 .To generate Gal8 knocked down MDCK clones, cells were transfected with the siRNA-expressing vector using Lipofectamine 2000 (#11668019, Invitrogen) according to the manufacturer's instructions before selection at 0.25 mg/mL hygromycin B (#400052, Calbiochem).For the ciliogenesis-related assay, MDCK cells were plated as an instant monolayer and allowed to be polarized in 12 mm Transwell (#3401, Corning) for 7 days or the indicated period specified in the relevant figure legends.

| Generation of the MDCK clones stably expressing Arl13B-mCherry or Neon-centrin 2
The plasmids (gift of Dr. Yu-Chun Lin) expressing mouse Arl13B C-terminally tagged with mCherry or centrin 2 N-terminally tagged with Neon were transfected into MDCK cells respectively using Lipofectamine 2000 (#11668019, Invitrogen) according to the manufacturer's instructions.The transfected cells were selected with 0.8 mg/mL G418 for two weeks before the surviving candidates were pooled and subjected to two rounds of fluorescence-activated cell sorting by the FACSAria™ III (BD Biosciences) to enrich the mCherry or Neon-positive cells to more than 95% in the cell population.

Western blot
The GST-fused full-length Gal8 were produced in E. coli BL21 strain and conventionally purified on Glutathione Sepharose® 4B beads (GE17-0756-01, GE Healthcare).MDCK cells were lysed in CSK buffer (50 mM NaCl, 300 mM sucrose, 10 mM PIPES (pH 6.8), 3 mM MgCl 2 , and 0.5% (v/v) Triton X-100) and incubated overnight at 4°C with GST fused Gal8 captured on Sepharose® 4B beads.The beads were then washed with CSK buffer 3 times with intermittent sedimentation by centrifugation.Pulled-down proteins were collected by boiling in an SDS sample buffer.Then, the samples were resolved by the SDS-PAGE and analyzed by Western blotting.For Western blotting, proteins were blotted onto nitrocellulose membranes (NBA085C001EA, PerkinElmer) which were blocked in 5% milk/Phosphate-buffered saline (PBS) containing 0.05% (v/v) Tween-20 (PBST).The membranes were incubated with primary antibody overnight at 4°C followed by 3 washes with PBST and vigorous shaking.The membranes were incubated with horseradish peroxidase-conjugated secondary antibodies (NA931V or NA934V, GE Healthcare) at room temperature for 1 h, and visualization was performed using enhanced chemiluminescence reagents (PK-NEL122, PerkinElmer) and exposure to X-ray film (GE28-9068-39, GE Healthcare).

| Immunoprecipitation
To purify Gal8 secreted in the culture medium, a clone of MDCK cells stably expressing FLAG-tagged Gal8 was cultured on 24 mm Transwell (#3412, Corning) until polarization, and then apical and basolateral culture medium was collected on the predetermined dates (2, 6, 8, and 10 days).Collected cultured media were incubated with mouse anti-FLAG antibody-conjugated agarose beads (A2220, Sigma-Aldrich) and rotated at 4°C overnight.The immunoprecipitated materials were separated by the SDS-PAGE followed by Western blotting analysis using appropriate antibodies.

| Lipid-binding assay
Sphingo strips (S-6000, Echelon Biosciences) were incubated with His-or GST-Gal8 for 1 h at room temperature.The membranes were blocked in 5% milk/PBS containing 0.05% (v/v) Tween-20 (PBST).The membranes were incubated with primary antibody overnight at 4°C followed by 3 washes with PBST and vigorous shaking.The membranes were incubated with horseradish peroxidase-conjugated secondary antibodies (NA931V or NA934V, GE Healthcare) at room temperature for 1 h, and visualization was performed using enhanced chemiluminescence reagents (PK-NEL122, PerkinElmer) and exposure to X-ray film (GE28-9068-39, GE Healthcare).

| Lipid floatation assay
Confluent MDCK cells grown on 75 mm Transwell (#7910, Corning) were lysed for 20 min on ice in 2 mL of TNE buffer (20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 5 mM EDTA) containing 1% Triton X-100 and passed five times through a 26-gauge needle.The 2 mL lysate was mixed with 2 mL 80% sucrose in TNE and placed at the bottom of a centrifuge tube (331372, 14 × 89 mm, Beckman Coulter).A discontinuous sucrose gradient (5%-30% in TNE) was layered on top of the lysates, and the samples were centrifuged at 39 000 rpm for 18 h in an ultracentrifuge (SW41; Beckman Coulter).The fractions were separated by SDS-PAGE followed by Western blotting analysis using appropriate antibodies.

| Binding of gangliosides to inhibit the effect of Gal8 recombinant protein on cilia elongation
The fresh liquid stock of ganglioside solution (2 mM) was prepared by reconstituting the lyophilized ganglioside GM1 (AV-860065P, Avanti Polar Lipids) or GD3 (AV-860060P, Avanti Polar Lipids) with the appropriate amount of tissue culture grade DMSO (D2650, Sigma-Aldrich).Recombinant His-Gal8 was incubated with the indicated concentration of GM1 or GD3 as specified in the experimental study, respectively, at 4°C for 30 min.Then, these ganglioside pre-loaded His-Gal8 samples were added to the apical compartment of MDCK cells grown on Transwell and incubated at 4°C for 30 min.After extensive washing, the cells were processed for immunofluorescence study.

| Immunofluorescence staining and microscopy
MDCK cells cultured on 12 mm Transwell (#3401, Corning) were washed twice with PBS containing Ca 2+ and Mg 2+ (phosphate-buffered saline containing 1 mM CaCl 2 and 0.5 mM MgCl 2 ) and fixed with 3.7% (w/v) paraformaldehyde.Permeabilization was carried out by incubation with CSK buffer at room temperature for 15 min.Cells were then washed twice with PBS and blocked with PBS containing 1% (w/v) bovine serum albumin (BSA), 10% (v/v) goat serum, and 50 mM NH 4 Cl at room temperature for 1 h.After washing briefly with PBS containing 0.2% (w/v) BSA, the cells were incubated with primary antibody overnight at 4°C.Cells were then washed thrice with PBS containing 0.2% (w/v) BSA and incubated with a secondary antibody at room temperature for 2-4 h.After washing thrice with PBS containing 0.2% (w/v) BSA, cells were mounted with ProLong™ Diamond Antifade Mountant (P36961, Thermo Fisher Scientific).Images were acquired with a Zeiss LSM880 confocal microscope (Carl Zeiss).
The mouse tissue section staining was followed: paraffin sections of mouse kidney tissue were deparaffinized, rehydrated, and autoclaved in Trilogy™ (920P-06, Cell Marque) for antigen retrieval before being permeabilized with CSK buffer for 15 min.Then, the sections were washed twice with PBS, 5 min each, and blocked in blocking solution (PBS containing 1% BSA, 10% goat serum, and 50 mM NH 4 Cl) for 1 h followed by washing twice with PBS containing 0.2% BSA.Sections of the kidneys were incubated with primary antibody at 4°C overnight, were washed three times, each for 5 min, in PBS containing 0.2% BSA, and were incubated with a secondary antibody which is goat anti-mouse or rabbit with fluorescein-conjugated for 90 min at room temperature in the dark.Sections of the kidney were washed three times, each for 5 min, in PBS containing 0.2% BSA and mounted with ProLong™ Diamond Antifade Mountant (P36961, Thermo Fisher Scientific).
2.12 | Surface biotinylation and mass spectrometry MDCK cells were cultured on 75 mm Transwell (#7910, Corning) for 7 days for polarization.Transepithelial electric resistance was measured each time before performing the experiments to confirm the tightness of the monolayer.Filters were washed three times with cold PBS containing Ca 2+ and Mg 2+ .Sulfo-NHS-biotin (#21217, Thermo Fisher Scientific) in PBS containing Ca 2+ and Mg 2+ was applied to apical (4 mL) chambers.Filters were incubated on ice for 30 min on a rocker platform.Unreacted biotin was quenched by washing cells in five changes of Tris-saline (10 mM Tris-HCl, pH 7.4, 120 mM NaCl).Then each filter was incubated with 4 mL of GST-Gal8 (0.1 μg/μL) on ice for 30 min on a rocker platform.After washing with three changes of cold PBS, the cells were then extracted in CSK buffer for 15 min at 4°C. Cell lysates were centrifuged (12 000 g, 10 min) and the collected supernatant was then incubated with Glutathione Sepharose® beads (GE17-0756-01, GE Healthcare) at room temperature for 3 h.Beads would be washed three times with PBS and eluted twice with elution buffer (50 mM Tris-HCl, 10 mM reduced glutathione, pH 8.0).The eluted lysates were incubated with immobilized avidin (#20219, Pierce) overnight at 4°C.The collected beads were washed with PBS 3 times before adding 2× SDS sample buffer (v/v) and resolved by SDS-PAGE.The gel was then silver stained, followed by in-gel tryptic digestion and liquid chromatographytandem mass spectrometry as described previously. 14,15o identify the protein, the raw spectrometry data were analyzed using Proteome Discoverer software (version 1.4, Thermo Fisher Scientific).The resulting mass spectra were searched against the Swiss-Prot Canis sequence database using the Mascot search engine (Matrix Science, version 2.2.04).The annotated subcellular localization of our identified proteins was according to the Ingenuity Pathway Analysis (IPA, Qiagen) software.Proteomic data for the Gal8 interactome have been deposited in the ProteomeXchange Consortium (http:// prote omece ntral.prote omexc hange.org) through the PRIDE partner repository (https:// www.ebi.ac.uk/ pride/ ). 16The dataset identifier is PXD038021.

| Lipid raft labeling with cholera toxin subunit B
MDCK cells cultured on 12 mm Transwell (#3401, Corning) were washed twice with PBS containing Ca 2+ and Mg 2+ and incubated with cholera toxin subunit B (C22843, Invitrogen) on apical space for 30 min on ice.After then, cells were fixed with 3.7% (w/v) paraformaldehyde for 15 min kept on ice, and permeabilized with CSK buffer for 15 min at room temperature.Cells were then washed twice with PBS and blocked with PBS containing 1% (w/v) bovine serum albumin (BSA), 10% (v/v) goat serum, and 50 mM NH4Cl at room temperature for 1 h.After washing briefly with PBS containing 0.2% (w/v) BSA, the cells were incubated with primary antibody overnight at 4°C.Cells were then washed thrice with PBS containing 0.2% (w/v) BSA and incubated with a secondary antibody at room temperature for 2-4 h.After washing thrice with PBS containing 0.2% (w/v) BSA, cells were mounted in ProLong™ Diamond Antifade Mountant (P36961, Thermo Fisher Scientific).Images were acquired with a Zeiss LSM580 or LSM880 confocal microscope (Carl Zeiss).

|
The lipid raft perturbation treatment MDCK cells were cultured on 12 mm Transwell (#3401, Corning) co-treated with 5 mM lovastatin (#438185, Sigma-Aldrich) and 2.5 mM MBCD (C4555, Sigma-Aldrich) at 37°C for 1 h.The cells were washed twice with PBS and preceded as described above for immunofluorescence staining.The images were acquired with a Zeiss LSM880 confocal microscope (Carl Zeiss).

| Quantification of primary cilia length and statistical analyses
The 3D reconstructions of images were generated using microscope software Zen 2011 (Carl Zeiss) and quantification of cilia length in the 3D was made with Fiji as an image processing package (National Institutes of Health).The data of primary cilia length are presented with GraphPad Prism 7. Continuous data are presented as means ± standard deviations.Differences in the values between the two groups were determined by the Student's ttest or Mann-Whitney U test.When more than two groups were compared, one-way ANOVA or the Kruskal-Wallis test was performed, and a value of p < .05 was considered statistically significant.Densitometry analysis was carried out using ImageJ.

| Galectin 8 interacted with the primary cilia membrane and promoted cilia elongation in polarized MDCK monolayer
While we reported the effect of Gal8 knockdown on the apical-basal polarity of MDCK cells via apical targeting of Gp135/podocalyxin, 17 we also noticed that there was an accompanying decrease of ciliated cells in the MDCK clones stably knocked down of Gal8 (Figure 1A).Furthermore, endogenous Gal8 was demonstrated to be distributed in a punctate pattern along the primary cilia (Figure 1B).Based on these results and galectin family proteins are released from cells via a non-canonical secretory pathway, 18 we hypothesized that Gal8 interacted with primary cilia as an autocrine/paracrine factor and contributed to the ciliogenesis process during epithelial morphogenesis.To verify this hypothesis, we first examined the exact location of endogenous Gal8 at primary cilia in an MDCK clone stably expressing mouse Arl13B, which is a primary cilia marker, with C-terminally tagged mCherry.The cells were processed for immunostaining using two different experimental protocols: one group of cells was fixed and permeabilized, while the other group of cells would be fixed without permeabilization.As noted earlier, the endogenous Gal8 was colocalized with primary cilia marker acetylated tubulin in permeabilized Arl13Bmcherry MDCK cells (Figure 1C upper).However, only the Gal8, but not the acetylated tubulin signal could be detected at primary cilia in non-permeabilized Arl13Bmcherry MDCK cells (Figure 1C lower).The above result F I G U R E 1 Gal8 interacted with the primary cilia membrane and regulated ciliogenesis.(A) Right upper, WB showed efficient downregulation of endogenous Gal8 expression in two independent MDCK stable clones expressing shRNA targeted at galectin 8 as compared to the mock control.Morphometric analysis quantifying the extent of ciliogenesis (right lower, bar chart was depicted for means and standard deviations (SDs), n > 200, **p < .01,***p < .001)and representative immunofluorescence images (left) in these cells.(B) Polarized MDCK cells were processed for immunofluorescence using mouse anti-acetylated tubulin and rabbit anti-galectin 8 antibodies.(C) Confocal images showing the patterns of endogenous Gal8 (green) and the primary cilia marker-acetylated tubulin (white) in an MDCK cell line stably expressing mCherry-tagged Arl13B which were either fixed and permeabilized (upper) or only fixed but not be permeabilized (lower) before the procession for double immunostainings.(D) The Coomassie blue-stained gel showed the purity and quantity of each recombinant protein used in the study.Ten microliters of the indicated recombinant proteins as well as the albumin standards were loaded into each well of the SDS-PAGE and separated by electrophoresis before staining the gel with Coomassie blue dye.(E) A double immunofluorescence study in polarized MDCK cells revealed that exogenous His-Gal8 (0.1 μg/μL) added to the apical domain elicited primary cilia lengthening compared with PBS treatment.(F) Polarized MDCK epithelia were added with the indicated concentration of His-Gal8 from the apical surface for 30 min before the length of primary cilia was quantified by immunofluorescence study using acetylated tubulin staining.The median with an interquartile range of experiments was shown as black bars.The statistical significance was calculated by the Kruskal-Wallis with Dunn's test.n > 200, n.s.(not significant) p > .05,***p < .001,****p < .0001.demonstrated the non-permeabilization protocol successfully prevented the access of the staining antibodies to intracellular targets such as acetylated tubulin and further indicated that there was a significant pool of endogenous Gal8 interacting with primary cilia from extracellular space.Next, we generated and purified His-or GST-tagged Gal8 recombinant protein (Figure 1D) and added submicromolar His-tagged Gal8 on the apical domain of polarized MDCK monolayer with the initial goal to see the retention of the exogenous Gal8 at primary cilia.Very surprisingly, we not only observed the signal of recombinant Gal8 retained at primary cilia but also witnessed the length of primary cilia increase dramatically within several minutes after Gal8 treatment (Figure 1E).This cilia-lengthening effect can be discernable when as low as 5 ng/μL (equal to 0.135 μM) of the recombinant His-Gal8 protein was added to polarized MDCK cells (Figure 1F).Taken together, these results implied that endogenous Gal8 could be secreted to the apical domain and promoted primary cilia elongation in polarized epithelial cells.
To exclude the possibility that the surprisingly cilial retention and cilia-lengthening effect of the recombinant Gal8 might result from denatured/misfolded Gal8 protein and its subsequent non-specific binding onto the cilia, His-Gal8 was inactivated by boiling at 100°C for 10 min (His-Gal8Δ) and added to the apical space of MDCK cell cultures.Compared to the His-Gal8 treatment without previous boiling, heat-inactivated His-Gal8 could neither be retained on primary cilia nor induce primary cilia elongation (Figure 2A).To further substantiate the finding we observed from using His-Gal8, we then applied another recombinant protein GST-tagged Gal8 to either the apical or basolateral compartment of polarized MDCK cells.The length of primary cilia increased only when exogenous GST-Gal8 was presented to the apical domain (Figure 2B) implying there is a specific Gal8-interacting partner on the apical surface of polarized MDCK cells.Besides MDCK cells, the recombinant Gal8 protein also displayed a cilia lengthening effect on mouse IMCD3 or NIH3T3 fibroblasts (Figure 2C).To further investigate the effect of Gal8, the recombinant Gal8 protein was washed off the apical chamber of the Transwell after incubation for 30 min.Then, the apical chamber was replete with fresh completed medium.We observed that the Gal8-induced lengthening effect on primary cilia would be quite lasting as primary cilia were still longer than the untreated control at 4 h after recovery (Figure 2D).

MDCK cell culture during the course of cilia lengthening
It has been reported the primary cilia of MDCK cells elongated over the time course of culturing. 19As we demonstrated earlier that extracellular Gal8 might play an instrumental role in ciliogenesis, we set to examine whether apically secreted Gal8 would increase during cell culture.To this end, we collected the culture media from the apical and basolateral compartments of a polarized MDCK cell line that stably expressed FLAG-Gal8 20 and quantified the content of extracellular FLAG-Gal8 after immunoprecipitation of the culture media with mouse anti-FLAG antibodies.We found that increasing Gal8 was transported to the apical pole over cultivation time (Figure 3A).The length of primary cilia also elongated during this observation period (Figure 3B,C).Simultaneously, we investigated the correlation between primary cilia length and subcellular localization of endogenous Gal8.To better discern the exact location of Gal8 during the ciliogenesis, we performed imaging studies in an MDCK cell line stably expressing neon-tagged centrin 2 that marks the position of centrosome and provides spatial cues regarding the orientation of primary cilia which could be a challenge, especially for short cilia.There are basically two expression patterns of Gal8: (1) Gal8 was localized exclusively at the TZ (Figure 3D, left upper) in which the fluorescent signal of Gal8 was noted in between the signals of acetylated tubulin as a marker for axoneme and neon-centrin 2 and (2) Gal8 signal was noted extending into axoneme (Figure 3D, left lower).The expression pattern of Gal8 is more correlated with cilia length than with culture period (Figure 3D, right).After a 2-day culture period when about 70% of cilia were shorter than 1 μm, Gal8 in this population of cells tended to exhibit a TZ pattern on their primary cilia.Conversely, 76.9% of cells after a 10-day culture period had longer cilia length (>1 μm) and Gal8 generally exhibited an axoneme pattern.It is noteworthy that there were cells whose ciliogenesis was exceptional; about 30% of cells after a 2-day culture period had cilia longer than 1 μm while about 23.1% of cilia after a 10-day culture period were still F I G U R E 2 Recombinant Gal8 facilitated primary cilia lengthening through transient interaction specifically occurring on an apical domain.(A) Double immunofluorescence study (left side panels) in polarized MDCK cells revealed that exogenous His-Gal8 (0.1 μg/μL) added on the apical domain elicited primary cilia lengthening compared with PBS treatment or the heat-inactivated His-Gal8.Quantification of primary cilia length for this assay was presented in the right panel.(B) Exogenous GST-Gal8 (0.2 μg/μL) treatment onto the apical domain of polarized MDCK monolayer promoted primary cilia elongation, but no effect was observed when the recombinant protein was applied to the basolateral domain (left side panels).Quantification of primary cilia length for this experiment was presented in the right panel.Data were showing the lengths of primary cilia in individual cells from 3 independent experiments.(C) NIH3T3 fibroblasts and IMCD cells were treated with His-Gal8 (0.1 μg/μL) for 30 min before the length of primary cilia was quantified by immunofluorescence study using acetylated tubulin staining.(D) Polarized MDCK epithelia were added apically with His-Gal8 (0.1 μg/μL) for 30 min followed by recovery in complete medium for 1 (His-Gal8_R1) or 4 (His-Gal8_R4) h before the cells were processed by immunostaining using antibody recognizing acetylated tubulin.Data were quantified with the nonparametric test.n > 200, n.s.p > .05,**p < .01,***p < .001,****p < .0001.short (≦1 μm).It is intriguing that although the cilia length in general increased over the course of culture, it is the cilia length not the age of the culture that better correlates with whether the expression of Gal8 on the cilia is adopting a TZ or axoneme pattern.For example, the Gal8 expression in 78.2% of cells which remained short (≦1 μm) after a 10-day culture was TZ rather than the axoneme pattern noted in their companion cells which had been cultured for the same 10 day's period but possessed longer cilia (>1 μm).These results indicated that Gal8 was secreted to the apical domain in polarizing renal epithelia when primary cilia drastically increased their numbers and lengths.

| Gal8 interacted with membranous ganglioside to promote primary cilia elongation
Previous research demonstrates that Gal8 is distinguished from other galectin family proteins by possessing a preferential affinity for negative-charged oligosaccharides, the sulfated and sialylated glycoconjugates. 21To explore whether this characteristic is responsible for the interaction between the extracellular Gal8 and its receptive partners over the surface of primary cilia, we applied the recombinant Gal8 proteins to polarized MDCK monolayer grown on Transwell at 4°C so that the retention signal after washing off the labeling solution might only reflect the surface pool of extracellular Gal8 but not the endocytosed pool.Compatible with our previous results (Figure 1B), there were clear signals of the recombinant proteins at the primary cilia (Figure 4A).While the addition of GST-Gal8 (WT) recombinant protein onto the apical domain of polarized MDCK monolayer resulted in significant anti-GST antibody labeling to the primary cilia, a triple alanine mutant (R44A/Q46A/R58A) of the recombinant Gal8 (AAA), which was reported to diminish the affinity of Gal8 towards sialic acid-containing glycan, 21 showed much diminished signals retained at primary cilia.To delineate how Gal8 interacted with the primary cilia membrane, we used a dot blot assay to identify which lipid species had a relatively strong binding affinity with Gal8.As noted, recombinant GST-Gal8 and His-Gal8 both displayed a specific affinity to negatively charged lipids, especially the disiaganglioside GD3 (Figure 4B).Gangliosides are glycosphingolipid with several sialic acids linked on the β-d-galactose or N-Acetyl-β-d-galactosamine (GalNAc) of glycan chain which enriched on neurons 22,23 and are a component of lipid raft. 24This result showed that Gal8 interacted with ganglioside GD3 in vitro.Moreover, we also observed that GD3 was mostly located at the base of primary cilia in polarized MDCK cells, and the endogenous GD3 was redistributed into the shaft of the elongated cilia after the addition of recombinant Gal8 in a chaotic pattern (Figure 4C).In fact, all the gangliosides we examined, except GD2, were noted at primary cilia as revealed by their high degree of co-localization with the ciliary marker acetylated tubulin (Figure 4D).We therefore suspect that the gangliosides on primary cilia serve as the docking molecules for the extracellular Gal8 and this interaction is essential for the Gal8-dependent primary cilia elongation.To verify this hypothesis, we saturated first the cell apical domain of polarized MDCK cells with various ganglioside antibodies at 4°C before the application of recombinant Gal8 protein.After GD1a, GD1b, or GT1b antibody pre-treatment, the Gal8-induced primary cilia elongation was significantly deterred; however, the Gal8 elicited ciliary lengthening was not affected by GM1 antibody blocking (Figure 5A,B).To verify the specificity of the ganglioside antibodies used to block the access of Gal8 to particular ganglioside and inhibit primary cilia elongation in the above experiment, gangliosides GM1 and GD3 from commercial sources were pre-mixed with His-Gal8 respectively.The pre-loaded Gal8 was added to the apical compartment of polarized MDCK cells and analyzed for the length of primary cilia (Figure 5C,D).With GM1 pre-mixing, the His-Gal8 reserved its capability of retaining the cilia as well as promoting primary cilia elongation.By contrast, Gal8 pre-loaded with GD3 failed to retain primary cilia and was compromised in its cilia lengthening effect.All together, these results indicated that apical Gal8 interacted with particular gangliosides such as GD3 on the primary cilia membrane to induce cilia elongation.

| Gal8 perturbed a lipid raft-based barrier to promote primary cilia elongation
In addition to membranous gangliosides as the potential candidate for binding apical Gal8, we searched for the protein substrates that Gal8 interacted with on the primary cilia.As the cilia occupy only 0.3% volume of the cell body, traditional biochemical purification scheme, such as pulled-down analysis using whole cell lysate would be unsatisfactory to reveal the specific binding partner at the cilia for Gal8, which is known to be localized mainly in the cytosol and likely interacts with many potential interacting partners located at other subcellular compartments outside of primary cilia.To identify the ciliary molecule that mediates the cilia-lengthening effect of Gal8, we approached by adopting sequential apical domain selective biotinylation, surface GST-Gal8 fusion protein incubation, followed by a two-step affinity purification scheme (Figure S1).We then processed these apical surface Gal8-interacting candidates by mass spectrometry.We totally identified 254 reproducible candidates in two independent experiments.There were 96 and 117 plasma membrane candidates noted from the first and second experiments respectively.We focused on 66 candidates categorized as plasma membrane F I G U R E 3 Gal8 was secreted to the apical domain during epithelial polarization.(A) Culture media were collected from apical (AP) and basolateral (BL) compartments of FLAG-tagged Gal8 stably expressing MDCK cells grown on Transwell for the indicated periods.These conditioned media were immunoprecipitated with mouse anti-FLAG antibody before being processed for SDS-PAGE together with 10% of total cell lysates (TL) and analyzed by Western blot analysis using mouse anti-FLAG antibody.(B) MDCK cells were grown on Transwell for the indicated periods and processed for immunostaining for acetylated tubulin and Hoechst staining.(C) Quantitative results of the experiment performed in B according to the percentage of ciliated cells (left, data were expressed as means and SDs, n > 200, ***p < .001as compared to day 2) and average primary cilia length (right, individual measurements were quantified with nonparametric test.n > 200, **p < .01,***p < .001as compared to day 2).(D) Left panel, confocal images showing the expression patterns of endogenous Gal8 (blue) and the primary cilia marker-acetylated tubulin (red) in an MDCK cell line stably expressing neon-tagged centrin 2 (green).Right panel, categorized results of the expression pattern of Gal8 as either exclusively localized to TZ or extended to the axoneme of the primary cilia by the length of primary cilia and the culture age of the cells.Data were expressed as means and SDs.proteins (~25% of total candidates) (Table S1) which were repetitively identified in two separate experiments (Figure 6A).Because we previously established that Gal8 interacted with gangliosides which were components of the lipid raft, candidates of lipid raft scaffold proteins, such as flotillin 1 (Flot1), stomatin (Stom), and caveolin 1 (Cav1), especially drew our attention.To examine whether Gal8 interacted with lipid raft scaffold proteins, Flot1, Stom, and Cav1 were over-expressed in HEK293T cells, respectively, and mixed with GST or GST-Gal8 beads for pull down assay.Since the result demonstrated that the purified GST-Gal8 recombinant protein can interact with the exogenously expressed HA-Flot1, Stom-FLAG, or Cav1-V5 (Figure 6B), we sought further biochemical evidence of Gal8 interacting with lipid raft using flotation assay which revealed that Gal8 co-migrated with endogenous Flot1 and Cav1 lipid raft scaffold proteins (Figure 6C).Lipid rafts are in the TZ region of primary cilia. 25To demonstrate the potential interaction between endogenous Gal8 and the lipid rafts at primary cilia in vivo, we labeled live MDCK cells with cholera toxin B subunit (CTxB), a lipid raft marker, before fixating the cells and processing for double immunofluorescence study using mouse anti-acetylated tubulin and rabbit anti-Gal8 antibodies.Interestingly, the CTxB staining was noted at the base of stubby primary cilia with minimal Gal8 nearby (Figure 6D, upper row); however, pronounced Gal8 staining was co-localized with the CTxB signal that was distributed at the bases of other longer primary cilia (Figure 6D, lower row).Furthermore, there was a close correlation between the staining intensity of endogenous Gal8 that was colocalized with CTxB signal and the length of the respective cilium which was rooted at the same region (Figure 6E).TMEM231 is a TZ component that has been demonstrated to contribute an essential role in maintaining the TZ ultrastructure and the gatekeeper to regulate cilia transportation. 26The subcellular localization of TMEM231 was noted to be re-distributed into the elongated cilia upon apical addition of recombinant GST-Gal8 as compared to be located at the cilia base of control cells (Figure 6F).These results implied that the base of the primary cilia (presumably at the TZ) could be a lipid raft-enriched region and binding of the apically secreted Gal8 might elicit relocation of TZ component protein such as TMEM231 and thereby perturb the function of TZ as a cilia transportation barrier to facilitate primary cilia elongation.

| Cav1 organized a lipid raft barrier on the TZ to restrict primary cilia growth
The TZ membrane at the base of cilia is the most detergent resistant of all surface membranes in C. reinhardtii, 12 and a high-ordered lipid phase exists at the TZ to limit free lateral diffusion of membrane proteins and lipids between ciliary and plasma membranes.
It has been shown that knockout of KIF13B, a key regulator of ciliary TZ configuration and membrane composition, led to lipid raft scaffold protein Cav1 translocated from TZ to primary cilia axoneme. 27Since we demonstrated that the interaction with lipid raft component ganglioside was critical to the cilia-lengthening effect of Gal8 and lipid raft marker CTxB was located at the bases of primary cilia, we set up to examine the effect of acutely inhibiting lipid raft on cilia length as well as the location of lipid raft scaffold proteins.When MDCK cells were treated with lovastatin and methyl-b-cyclodextrin (MBCD) to inhibit cholesterol biochemical synthesis and deplete cholesterol on lipid raft, primary cilia increased their lengths and Cav1 was re-distributed into the shaft of these elongated cilia (Figure 7A, co-treatment of lovastatin and MBCD).Since we had demonstrated the membrane at the bases of primary cilia, presumably the TZ region, were enriched with lipid rafts and we also identified lipid raft organizing proteins as the apical Gal8 interacting partners, we hypothesize that primary cilia TZ is a lipid raft-rich region and extracellular Gal8 binding to TZ would perturb the molecular organization and function there to allow fast elongation of primary cilia.Compatible with this hypothesis, apical application of either the purified recombinant His-Gal8 (Figure 7B) or GST-Gal8 (Figure 7C) would induce Cav1 relocation into the axoneme of the elongated primary cilia, which phenocopied the pattern observed in cells being depleted of lipid rafts.To further confirm the role of Cav1 in the effect of Gal8 on facilitating cilia growth, we generated Cav1 knock-out MDCK cell lines using CRISPR-Cas9 technology.After DNA sequencing analysis, we identified several candidates and one clone which preserved the wild-type genomic sequence.Subsequent Western blotting analysis showed intact Cav1 expression in the clone which apparently preserved the Cav1 genomic sequence (Figure S2, clone #26); therefore, this clone was used for later study as a mock control.All the other candidates showed much reduction in Cav1 expression with two clones (Figure S2, clones #21 and #41) showed F I G U R E 4 Gal8 interacted with gangliosides on the primary cilia and this interaction contributed to the cilia-elongating effect of Gal8.(A) Purified recombinant GST-Gal8 (wild type, WT) or the sialic acid binding defective mutant GST-Gal8-AAA protein (AAA) (0.2 μg/μL) was applied onto the apical domain of polarized MDCK monolayer at 4°C for 30 min, respectively.After extensive washing, immunofluorescence staining with mouse anti-acetylated tubulin and rabbit anti-GST antibodies was used to assess the extent of these recombinant proteins retained on the surfaces of primary cilia.Both primary cilia length and colocalization of the Gal8 recombinant proteins with the cilial marker acetylated tubulin were quantified.The data were quantified with the nonparametric test, and the median with interquartile range was shown as black bars, n > 200, **p < .01,***p < .001,****p < .0001.(B) The potential interaction between sphingolipid and Gal8 was evaluated by incubation of a commercial sphingolipid strip with GST, GST-Gal8, or His-Gal8, respectively.After extensive washing, the bound recombinant proteins were revealed by immunoblotting analysis with GST or His antibodies.(C) Polarized MDCK cells were treated with PBS, GST, or GST-Gal8, respectively, and processed for double immunofluorescence study with rabbit antiacetylated tubulin and mouse anti-GD3 antibodies.The inserts in the overlapped channels showed the details of the spatial relationship between GD3 and cilia within the boxed areas.(D) Polarized MDCK cells were processed for immunofluorescence using rabbit antiacetylated tubulin and the indicated ganglioside-specific mouse antibodies.
| 15 of 21 SYU et a faint signal at the region where the endogenous Cav1 was supposed to be on the Western blot.To avoid ambiguity, we selected the two clones (Figure S2, clones #17 and #31) which displayed a complete absence of Cav1 expression for further assays.Immunofluorescence analysis demonstrated that Cav1 expression was located at the base of primary cilia in the mock while there was a complete loss of staining signal in the Cav1 knockout clones (Figure 7D, left).Quantitation further confirmed the primary cilia length increased in the Cav1-depleted MDCK cell lines (Figure 7D, right).Taken together, these data imply that Cav1, presumably through its lipid raft scaffold function, helps maintain the structural organization and the diffusion barrier function of TZ.Moreover, exogenous Gal8 treatment phenocopied the effect of Cav1 knockout; that is, TZ disorganization and cilia elongation.Thus, we hypothesize that Cav1 maintains a balanced interaction with some TZ components to organize the structure and barrier function of TZ.The additional Gal8 available to TZ would perturb the balanced interaction between Cav1 and some TZ components to compromise the barrier function of TZ and eventually induce cilia elongation.To test this hypothesis, cell lysates were harvested from an MDCK cell clone stably expressing FLAG-tagged Gal8 (MDCK-FG8) and performed a co-immunoprecipitation assay.We observed that Gal8 could interact with both Cav1 and TMEM231 (Figure 7E).Interestingly, the dynamic expression patterns of Cav1 during the ciliogenesis were similar to those of Gal8; that is, it is the cilia length rather than the age of the culture that defines whether the expression of Cav1 on the cilia is adopting a TZ or axoneme pattern (Figure S3A).The majority of Cav1 in short cilia (≦1 μm) displayed a TZ distribution pattern no matter whether the culture period was 2 or 10-day; however, over 95% of Cav1 adopted the axoneme expression pattern in long primary cilia (>1 μm) after either a 2-or 10-day culture period.This observation together with the result of the co-migration assay (Figure 6C) implied that Gal8 interacted with Cav1 at the primary cilia region.Intriguingly, TMEM231 exhibited a distinguished pattern from those of Gal8 and Cav1 over the course of ciliogenesis (Figure S3B).After the 2-day culture, TMEM231 tended to be located at TZ in both short and long primary cilia.By contrast, TMM231 expression shifted into an axoneme pattern after the 10-day culture.As we demonstrated that Gal8 was secreted to extracellular space during the culture period (Figure 3A), there would be more Gal8 in the apical compartment after a 10-day culture.Taken together, these data signified that the shift of TMEM231 expression pattern in terms of the TMEM231 localization in primary cilia was a response to the culture period and quantity of extracellular Gal8 accumulated during the culture period.To demonstrate the relevance of Gal8 in the interaction between Cav1 and TMEM231, His-Gal8 was added to MDCK cells before cell lysates were collected for co-immunoprecipitation.It is noted that Gal8 decreased the interaction between Cav1 and TMEM231 (Figure 7F).Altogether, these results revealed that the TZ of primary cilia was a lipid raft-rich region which was organized by Cav1 and Gal8 binding to TZ affected its diffusion barrier function to facilitate cilia growth.

| DISCUSSION
In this study, we demonstrated that apically secreted Gal8 binds to the primary cilia membrane and promotes primary cilia elongation in polarized epithelia.We also identified that the TZ of primary cilia is enriched with lipid raft component gangliosides and lipid raft scaffold protein, Cav1, which interact with Gal8.As primary cilia are elongated by extracellular Gal8, these lipid raft components as well as the TZ marker TMEM231 were translocated from the base of primary cilia into the axoneme.These effects of extracellular Gal8 can be phenocopied by pharmacological inhibition of lipid raft and genetic ablation of Cav1.These results imply that the interaction between Gal8 and the TZ perturbs the function of TZ as a diffusion barrier to regulate the molecular inflow and outflow of cilia axoneme.It is noteworthy that although the result of non-permeabilized versus permeabilized experiments confirmed the existence of a pool of extracellular 5 The interaction with GD3 gangliosides on the primary cilia likely contributed to the cilia-elongating effect of Gal8.(A) MDCK cells were seeded onto Transwell.After polarization, the MDCK monolayers were blocked with the indicated ganglioside antibodies added onto the apical domain for 1 h on ice.Afterward, the unbound antibodies were washed off at 4°C before the cells were incubated with purified GST-Gal8 applied into the apical compartment.Quantification of primary cilia length was assessed by immunofluorescence staining using an anti-acetylated tubulin antibody.The data were quantified with nonparametric test, n > 200, n.s.p > .05,*p < .05,**p < .01. (B) Representative confocal images from the study described in (A).(C) After polarization on Transwell, the apical compartments of MDCK monolayers were incubated with PBS, 5 μM purified His-Gal8, or the purified His-Gal8 which had been pre-mixed with the indicated concentration of GM1 or GD3 free lipid.Quantification of primary cilia length was assessed by immunofluorescence staining using an anti-acetylated tubulin antibody.The data were quantified with nonparametric test, n > 200, n.s.gal8 over primary cilia (Figure 1C), it does not exclude the possibility that gal8 also exists intracellularly in the cilia.Therefore, our findings support the role of extracellular galectin-8 in the regulation of cilia length; however, the possibility that intracellular galectin-8 may also contribute to ciliogenesis still exists.
The TZ is a macromolecular complex located at the most proximal region of the cilia axoneme, and immediately distal to the basal body. 10,11This residence of TZ in this region positions it as an ideal organelle to assume the role of a gatekeeper to govern molecular inflow and outflow of the ciliary axoneme.Genetic interaction studies of TZ component genes in Caenorhabditis elegans have identified two major functional modules, Meckel syndrome (MKS) and Nephronophthisis (NPHP).If one or more gene(s) within one module are disrupted, ciliogenesis is mostly normal, although a subset of cilia is slightly shorter. 28,29In stark contrast, deleting genes from both modules severely compromises structures. 30,31These findings imply that members of the MKS and NPHP modules can compensate for each other's function in ciliogenesis.A third group of TZ proteins, for example, MKS-5 (RPGRIP1L/MKS5), act as an essential "scaffold" or "assembly factor" for most, if not all, MKS and NPHP module proteins. 26,32Given the drastic influence of exogenous Gal8 in ciliary morphology, it is likely that Gal8 can exert its effect on more than one of these TZ scaffold proteins via caveolin 1.Our results confirm a previous study that reports a cilium-lengthening effect of knocking out caveolin 1α by gene editing approach. 33The authors conclude that caveolin 1α activates RhoA and its downstream effectors, ROCK and DIA1, to regulate the apical actin meshwork, which affects the arrival of vesicles to the ciliary base, which subsequently affects cilium length.Caveolin-1 has also been reported to affect ciliary functions in sensory neurons of C. elegans by working together with endosome maturation factors rabenosyn-5/VPS45 to regulate ciliary membrane homeostasis. 34 this study, we showed that the frequency of the cells with a primary cilium was diminished by Gal8 knockdown while extracellular Gal8 elicited cilial elongation.This implies that Gal8 behaves as an autocrine/paracrine factor to modulate epithelial morphogenesis.It is generally accepted that galectins, which lack the signal sequence for classical secretory pathways, are sent to extracellular space through unconventional protein transportation. 18oth exosomes and ectosomes are extracellular vesicles that allow cells to release bioactive molecules independent of the classical secretory pathway. 35We favor ectosomes over exosomes as the transportation machinery of apically secreted Gal8 because there are quite a few number of Gal8 interacting partners identified by our proteomic approach (Table S1) which are also confirmed ectosome components such as integrins 36,37 LAMP1, 38 flotilin 1, 39 and ezrin. 40,41lthough those proteins dependent on unconventional routes for their secretion do not share obvious common motifs or features in their sequence or structure, most unconventionally secreted proteins are reported to bind phospholipids.Indeed, anoctamin-6 (ANO6, also known as TMEM16F) is a lipid scramblase that has been demonstrated to facilitate the translocation of annexins, which are known to bind to negatively charged lipids and be secreted through unconventional pathways, across plasma membrane. 42Intriguingly, annexin A1 and anoctamin-6 were both identified in this study as the apical interactome candidates for extracellular Gal8 (Figure 6A and Table S1).Given that anoctamin is expressed at the primary cilia in polarized MDCK monolayer 43 and our study reveals Gal8 interacted with the lipid raft components, it would be interesting to explore whether anoctamin plays a role in the location of Gal8 at primary cilia and the cilia-lengthening effects of Gal8.
Another interesting finding revealed by our apical interactome screening for Gal8 comes from the identification of several ATP-binding cassette (ABC) transporters (Figure 6A and Table S1).Although ABC family transporters have not F I G U R E 6 Gal8 interacted with lipid raft scaffold proteins.(A) The Gal8-interacting candidate membrane proteins identified by mass spectrometry analysis.Venn diagram indicating the two sets of membrane proteins that were identified by two independent mass spectrometry analyses, respectively.Totally, 66 candidates (complete information is shown in Table S1) were identified by both analyses and were further categorized by their functional characterization.(B) Cell lysates were collected from HEK293T cells over-expressing HA-tagged flotilin 1 (HA-Flot1), FLAG-tagged stomatin (Stom-FLAG), and V5-tagged caveolin 1 (Cav1-V5), respectively, and mixed with GST or GST-Gal8 beads for pull-down assay.(C) Polarized MDCK cells stably expressing FLAG-tagged galectin 8 were processed for lipid raft flotation assay.The distribution profile of the FLAG-Gal8 over the sucrose gradient was assessed by Western blotting using antibodies detecting FLAG-Gal8, Flot1, and Cav1.(D) Polarized MDCK cells were processed for apical labeling with Alexa-595 conjugated cholera toxin B subunit (CTxB) before a double immunofluorescence study using rabbit anti-galectin 8 and mouse anti-acetylated tubulin antibodies.(E) Correlation of the experimental results as described in panel (D) between the Gal8 fluorescent intensity colocalized with CTxB signal and the length of the respective cilium which originated from the same region, n > 200.(F) Polarized MDCK cells were treated with GST or GST-Gal8 (0.2 μg/μL), and processed for double immunofluorescence study with mouse anti-acetylated tubulin and rabbit anti-TMEM231 antibodies.
been reported to be involved in the export of galectins, ATP binding cassette A1 (ABCA1) transporter is implicated in the secretion of annexin A1, another membrane lipid binding protein which also lacks leader peptide signal as galectin. 44,45Furthermore, it has been demonstrated in murine leukocytes that the P4-type ATPase ATP11C possesses significant flippase activity. 46Whether lipid scramblases such as anoctamin 6 or any ABC transporter facilitate the unconventional secretory process of galectins is an interesting issue that awaits further investigation.
Ciliopathies are a group of diseases caused by a genetic mutation affecting the genes encoding proteins localized to primary cilia.Some of the ciliopathies are characterized by absence or insufficient development of primary cilia in cells of the involved tissue.Therefore, our study might help decipher the molecular mechanism regulating ciliogenesis and pave the way for the development of future therapeutics for some ciliopathies.
F I G U R E 7 Caveolin 1 is localized on primary cilia and its interaction with galectin 8 regulated cilia elongation.(A) Polarized MDCK cells were treated with lovastatin and methyl-beta-cyclodextrin (MBCD) before being processed for a double immunofluorescence study using mouse anti-acetylated tubulin and rabbit anti-caveolin 1 (Cav1) antibodies to assess the expression pattern of Cav1 at primary cilia (left) and the length of primary cilia (right).The data were quantified with the nonparametric test, n > 200, ***p < .001.(B, C) Cav1 localization patterns at primary cilia in the control and recombinant His-Gal8 (B) or GST-Gal8 (C)-treated MDCK cells.(D) Cav1 expression pattern (left) and the length of primary cilia (right) were shown in two verified caveolin 1 knockout MDCK clones (#17 and #31) and a mock clone (#26).The data were quantified with the nonparametric test, n > 200, ****p < .0001.(E) Cell lysates were extracted from an MDCK clone stably expressing FLAG-tagged Gal8 (MDCK-FG8).Ten percent of the lysates were saved as loading control (input), and the rest of the lysates were processed for immunoprecipitation using non-immune mouse IgG (mIgG) or mouse anti-FLAG antibody covalently linked to agarose (M2 beads).The immunoprecipitated materials were then processed for Western blotting analysis using the antibodies specifically for TMEM231, FLAG, and caveolin 1 respectively.(F) MDCK cells were grown on Transwell for polarization before treatment of His-Gal8 (0.1 μg/μL in PBS) at the apical compartment for 30 min.A parallel cultured sample was treated with PBS as the control.Cell lysates were extracted with 10% of lysates being saved as loading control (input) and the rest of the lysates being processed for immunoprecipitation using mouse anti-caveolin 1 antibody (mCav1) or non-immune mouse IgG (mIgG).The immunoprecipitated materials were then processed for Western blotting analysis using the antibodies specifically for TMEM231, and caveolin 1, respectively.