To address the factors that shape the configuration of MT networks in β cells, we tested for a potential involvement of motors-dependent MT sliding. Not surprisingly, analysis of existing RNA-sequencing data in functional mouse islet β cells highlighted kinesin-1 KIF5B as the highest expressing β-cell motor protein (Figure 1A; Sanavia et al., 2021). Since this kinesin has been reported to have MT sliding activity in many types of interphase cells, we tested its potential ability to slide MTs in β cells.

Efficient depletion of KIF5B was achieved by utilizing two independent lentiviral-based shRNA against mouse Kif5b in mouse insulinoma cell line MIN6 (Figure 1B, Figure 1—figure supplement 1A). To visualize MT sliding, shRNA-treated MIN6 cells expressing mEmerald-tubulin were imaged by live-cell spinning disk confocal microscopy. We photobleached MTs in two large cell regions, leaving a thin unbleached band (‘fluorescent belt’) and analyzed relocation of MTs from the ‘fluorescent belt’ into the bleached areas over time. To minimize the effects of plausible MT polymerization and to reduce photobleaching, MTs were imaged for short time periods (5 min). Strikingly, in control cells (treated with scrambled control shRNA), MTs were efficiently translocated from the ‘fluorescent belt’ into the photobleached area, indicating that MT sliding events are prominent in this cell type (Figure 1C and D). In contrast, MIN6 cells expressing either Kif5b shRNA variants displayed a significant loss of MT sliding ability (Figure 1C, E, and F, Figure 1—video 1 ‘MT sliding FRAP’), indicating that the loss of KIFB leads to the loss of MT sliding.

While the assay described above provides an easy visualization of MT sliding, it allows for visualization of only a subset of the MT network. To further corroborate the above findings, we used a less photodamaging system to visualize MT sliding that does not involve photobleaching and allows for evaluation of displacements within the whole MT network. To this end, we applied an MT probe of fiducial marks, K560Rigor^E236A-SunTag (Tanenbaum et al., 2014) in MIN6 cells (Figure 1G–I). This probe contains the human kinesin-1 motor domain (residues 1–560) with a rigor mutation in the motor domain (K560Rigor^E236A) and fused to 24 copies of a GCN4 peptide. The rigor mutation in the motor domain causes it to bind irreversibly to MTs (Rice et al., 1999). When co-expressed with a pHalo-tagged anti-GCN4 single-chain antibody (ScFv-GCN4-HaloTag-GB1-NLS), K560Rigor^E236A can recruit up to 24 of the Halo ligands to a single position on an MT. The pHalo-tagged anti-GCN4 construct also contains a nuclear localization signal (NLS), which lends itself to reduce the background of the unbound dye. This enables visualization of MT sliding events via single molecule tracking of the fiducial marks along the MT lattice, allowing us to analyze MT sliding behavior within the whole network with high temporal and spatial resolution (Figure 1G–K, Figure 1—figure supplement 1B, Figure 1—video 2 ‘MT sliding Suntag’).

Our data indicate that in cells treated with scrambled control shRNA, a subset of K560Rigor^E236A-SunTag fiducial marks underwent rapid directional movements, interpreted as MT sliding events (Figure 1G and K). In contrast, in cells expressing Kif5b-specific shRNAs, the fraction of rapidly moving fiducial marks was significantly reduced, while the fraction of stationary fiducial marks increased (Figure 1H, I, and J), indicating the suppression of MT sliding. Collectively, these results indicate that KIF5B is necessary for MT sliding in MIN6 cells.

Because MT sliding mediated by KIF5B is a prominent phenomenon in β cells, we sought to test whether it has functional consequences for MT networks in these cells. Tubulin immunostaining revealed striking differences in MT organization between MIN6 cells treated with scrambled control shRNA versus Kif5b-specific shRNAs. While control cells had convoluted non-radial MTs with a prominent sub-membrane array, typical for β cells (Figure 2A), KIF5B-depleted cells featured extra-dense MTs in the cell center and sparse receding MTs at the periphery (Figure 2B and C). Significant reduction of tubulin staining intensity at the cell periphery (Figure 2D) confirms the robustness of this phenotype in MIN6 cells.

We proceeded to investigate MT organization within primary β cells of the Kif5b KO mouse line (Kif5b^fl/−:RIP2-Cre, Figure 2—figure supplement 1). These mice exhibit a β-cell-specific deletion of KIF5B, as previously outlined (Cui et al., 2011). Our examination revealed a consistent presence of a significant peripheral array in the C57BL/6J control mice, while the KO counterparts exhibited a partial loss of this peripheral bundle. Specifically, the measured tubulin intensity at the cell periphery was significantly reduced in the KO mice compared to their wild-type (wt) counterparts (Figure 2—figure supplement 1A–C). Distinct extra-dense MTs were not common in islet β cells, possibly due to the 3D shape of a β cell in tissue and/or compensatory mechanisms in organisms. Thus, our results indicate that loss of KIF5B leads to a strong defect in MT location to the cell periphery in both a β-cell culture model and primary β cells within intact islets, which could be a direct consequence of disrupted sliding of MTs from the cell center to the periphery.

An alternative explanation for the loss of peripheral MTs would be the loss of the mechanical modification of the MT lattice by the force provided by moving kinesin motors. In cells with radial MT organization, this mechanism was shown to promote MT rescues and growth toward the cell periphery (Andreu-Carbó et al., 2022). To test the contribution of kinesin motor movement at MTs to MT density at the cell periphery, we have attempted to rescue the KIF5B-depletion phenotype by re-expressing kinesin-1 motor (see Figure 3A-1), which is capable of moving along MT but is unable to bind MTs as cargos. Neither in KIF5B-depleted nor in wt MIN6 cells, peripheral MTs were affected by the motor re-expression (Figure 2—figure supplement 2), indicating that kinesin-dependent MT growth does not noticeably contribute to the formation of the peripheral MT array in this cell type.

Given the known significance of the peripheral MT array, which normally consists of well-organized MTs parallel to the cell membrane (Bracey et al., 2020), we have further analyzed directionality of MTs remaining at the cell periphery after KIF5B depletion. Previously, we published a custom image analysis algorithm (Bracey et al., 2020) allowing for detailed quantitative characterization of MTs directionality in relation to the nearest cell border (Figure 2—figure supplement 3). Here, we applied the same computational analysis to MT imaging data in MIN6 cells with perturbed KIF5B level and/or function. After deconvolution for increased signal-to-noise ratio, single 2D slices of MT images were subjected to thresholding. Two thresholding options were analyzed for the unbiased approach. A threshold specifically optimized (perfect) for the peripheral MT array in each cell was considered along with a standard threshold across the analyzed cell population (Figure 2—figure supplement 4). Thereafter, the directionality of MTs with respect to the cell border was determined. Every pixel of the image was analyzed with inconclusive pixels disregarded. MT directionality was quantified as a function of the distance from the cell border and directionality of peripheral MTs within 1 µm of the cell border quantified. Across the analysis, two thresholding conditions provided similar outputs. Our results indicate that in MIN6 cells treated with nontargeting control shRNA (Figure 2F), the distribution of MT angles in the cell periphery is vastly parallel and co-aligned with the cell boundary, as previously reported for primary islet β cells (Bracey et al., 2020). In contrast, the loss of KIF5B via shRNA depletion resulted in a significant loss of parallel MTs at the periphery as compared to control (Figure 2E–H).

Furthermore, the MT directionality analysis in β cells within islets from the Kif5b KO mice revealed a significant decrease in MT alignment to the cell periphery as compared to the wt mice, closely mirroring our findings from the Kif5b knockdown in MIN6 cells (Figure 2—figure supplement 1D–G). Collectively, our data suggest that KIF5B-mediated MT sliding serves to facilitate MT transport to the cell periphery and align them at this cellular location in both a β-cell culture model and primary β cells within intact islets.

Combined, our data demonstrate a distinct effect of KIF5B perturbation on both the distribution of MTs to the cell periphery and their orientation along the cell boundary. These data suggest that KIF5B-driven MT sliding is a decisive mechanism of the sub-membrane MT array generation, likely via redistribution of centrally nucleated MTs and their subsequent alignment at the cell edge. Thus, MT sliding is likely a critical component in functional MT organization in β cells.

While membrane cargo transport by KIF5s requires association of the heavy chain with the kinesin light chains (KLCs) and/or other adaptors, transportation of MTs as cargos occurs due to direct binding of KIF5 to MTs through the ATP-independent MT binding domain in heavy chain tail (C-terminus) (Jolly et al., 2010; Seeger and Rice, 2010).

To specifically establish the role of MT sliding by KIF5B in β cells, we sought to evaluate the effects of suppressing the binding of KIF5B tail to MTs. To this end, we used a previously generated construct (Ravindran et al., 2017), which is a motor-less version of wt kinesin-1 motor KIF5B containing the cargo-binding and ATP-independent MT binding domain and tagged with mCherry (mCh) at the amino terminus (Figure 3A). When overexpressed, this construct acts as a dominant-negative (DN) tool preventing the association of the tail of endogenous KIF5B with MTs. This tool is referred to as KIFDN^wt (KIF5B DN wt) moving forward (Figure 3A).

To confirm that the KIF5B tail domain binds to MTs in MIN6 and acts as a DN, we co-expressed KIFDN^wt and mEmerald-tubulin. When subjected to the FRAP assay, we detected a complete loss of MT sliding events as compared to a control (Figure 3B–D). To prevent tail engagement of the MT lattice through the ATP-independent binding domain, we opted to make point mutations in the tail domain to change the residues 892-DRKRYQ to 892-DAAAYA, thus generating KIFDN^mut (Figure 3A). Photobleaching assay in cells co-expressing the KIFDN^mut with mEmerald-tubulin indicated that the MT sliding activity was not blocked in the presence of the mutated construct (Figure 3B and E, Figure 3—video 1 KFDN FRAP), confirming that KIF5B tail domain binding to MTs is needed for MT sliding in β cells.

The DN constructs are also tagged with the FK506-rapamycin-binding protein (FKBP), as indicated in Figure 3A. This allows to heterodimerize them with a motor domain fused with the FKBP-rapamycin binding (FRB) domain using A/C heterodimerizer (rapalog) and reconstitute a functional motor (Inobe and Nukina, 2016). We restored kinesin-1 activity by connecting the motor-less KIF5B, KIFDN^wt, to kinesin-1 motor domain as a way to rescue the effects of DN approach of KIF5B tail overexpression. To this end, we co-expressed MIN6 cells with the tail domain, mEmerald-tubulin, and the KIF5C motor domain fused to FRB domain (Figure 3A). Once the tail and motor domains were dimerized with rapalog, we saw that the once blocked MT sliding events of the KIFDN^wt tail alone were now reversed (Figure 3B and F, Figure 3—video 2 KFDN + motor FRAP). In contrast, under conditions of heterodimerization of KIFDN^mut with the motor, MT sliding was greatly impaired (Figure 3B and G, Figure 3—video 2 KFDN + motor FRAP), indicating that the motor with the mutated ATP-independent binding domain cannot use MTs as cargos. Interestingly, the endogenous motor in this case was unable to efficiently transport MTs, suggesting that the endogenous motor pool engaged in MT sliding was significantly smaller than the overexpressed nonfunctional motor.

Overall, the results of the DN approach confirm that MT sliding in β cells is driven by KIF5B through direct kinesin-1 tail binding to cargo MTs.

Keeping in mind that KIF5B has additional major functions in addition to MT sliding, we sought to test the consequence of MT sliding more directly by turning to overexpression of the DN constructs. Thus, we took advantage of our heterodimerization approach to analyze MT patterns in cells with active kinesin-1, which is able or unable to slide MTs (see Figure 3F versus Figure 3G). We analyzed MIN6 cells that express either the KIFDN^wt or KIFDN^mut tail domains alone (Figure 4—figure supplement 1) or co-expressing and heterodimerized with the motor domain (Figure 4B and C). Cells were fixed and immunostained for tubulin to identify the MT network. As expected, overexpression of the KIFDN^wt tail construct alone acted as DN toward MT distribution to the cell periphery, resulting in decreased peripheral tubulin intensity (Figure 4—figure supplement 1A and C), while in cells expressing KIFDN^mut, MT patterns were comparable to control (Figure 4—figure supplement 1B and C).

Interestingly, expression of heterodimerized kinesin motors led to impaired MT network configurations compared to NT control (Figure 4). Specifically, blocking of MT sliding by overexpression of KIFDN^mut heterodimerized with the motor led to the decrease in peripheral tubulin intensity (Figure 4C and D) and impaired MT aligning along the cell border (Figure 4E and G). These data indicate that KIF5B-driven relocation of centrally nucleated MTs to β-cell periphery requires kinesin tail domain binding to ‘cargo’ MTs. Strikingly, overexpression of functional heterodimerized motor, which was capable of MT sliding and populating of the cell periphery with MTs as detected by tubulin intensity readings (KIFDN^wt heterodimerized with the motor, Figure 4B and D), also led to a deficient MT aligning at the periphery (Figure 4E and H). This can be interpreted as a result of unregulated sliding in these experimental conditions, since excessive kinesin-1-dependent sliding can lead to MT bending (Straube et al., 2006). This suggests that proper organization of MTs within the sub-membrane array requires fine-tuning of MT sliding activity.

Collectively, this indicates that regulated KIF5B activity is essential for redistributing MTs to the cell border and sustaining an aligned peripheral MT array.

Our data discussed above suggest that overexpression of functional kinesin-1 disrupts MT alignment at the cell periphery, inducing their bending and buckling (Figure 4E and G). To test if this defect is a result of excessive MT sliding, we employed a small molecule, kinesore, which is known to dramatically promote MT sliding by kinesin-1 (Randall et al., 2017). Kinesore targets kinesin cargo adaptor function by impairing KLC from binding kinesin heavy chain. As a result, kinesin heavy chain will excessively engage MTs through the C-terminal, ATP-independent MT-binding domain, leading to exaggerated MT sliding and the loss of membrane cargo transport by kinesin-1 (Randall et al., 2017). To this end, we pretreated MIN6 cells with 50 µm kinesore and stained for MTs (Figure 5).

We have observed an exaggerated MT looping resulting in a slight decrease in peripheral MT intensity (Figure 5C). To validate whether the effect of kinesore on MT morphology was KIF5B-dependent, we applied kinesore treatment to cells treated with either scrambled control or Kif5b-specific shRNA (Figure 5—figure supplement 1). MT organization in KIF5B-depleted cells was not affected by kinesore, indicating that the observed effect was likely due to kinesore-induced MT sliding exaggeration (Figure 5—figure supplement 2). Further analysis of the peripheral bundle indicated that MT alignment was strongly impaired upon kinesore-driven MT remodeling as compared with vehicle (DMSO) treatment (Figure 5D–G). The loss of co-aligned MTs and loss of tubulin density at the periphery indicate that MT sliding must be gated to prevent overcorrected MT networks.

It has previously been reported that kinesin-1 switches activity level in the presence of glucose stimuli (Donelan et al., 2002). We predicted that, as KIF5B activity modulates the MT sliding events, they would also change depending on the glucose concentration. To test this, we preincubated MIN6 cells with media containing a low concentration of 2.8 mM glucose (Figure 6A). We applied the photobleaching assay under these conditions and detected little to no MT sliding events. When switching glucose to a high concentration of 20 mM, MT sliding and remodeling events were significantly increased (Figure 6B, Figure 6—video 1 ‘FRAP low and high glucose’). Quantification of the sliding events demonstrated that MIN6 displaced MTs via MT sliding significantly more efficiently upon glucose stimulation (Figure 6C). We then turned to single-molecule tracking of MT lattice fiducial marks (K560Rigor^E236A-SunTag) to further investigate this observation. Consistent with the photobleaching assay, the fiducial marks were predominantly stationary in cells preincubated in 2.8 mM glucose (Figure 6D) but frequently underwent directed relocation events indicative of MT sliding in cells after being stimulated with 20 mM glucose (Figure 6E, Figure 6—video 2 ‘SunTag low and high glucose’).

These data demonstrate that glucose-stimulated remodeling of the MT network involves regulated MT sliding. Given the importance of MT sliding for peripheral MT abundance (Figure 7A) and alignment (Figure 7B), this effect may be essential to remodel parts of the peripheral MT array during GSIS to allow for secretion (Figure 7C) and/or to restore the initial array after glucose-dependent destabilization (Figure 7D).

MIN6 cells (a gift from Dr. Miyazaki, Osaka University Medical School, Japan) between passages 40 and 60 were utilized (Ishihara et al., 1993; Miyazaki et al., 1990). Cells were authenticated on arrival (passage 40) for murine origin by STR profiling. Thereafter, β-cell identity was routinely confirmed by insulin content per cell (by immunostaining) and responsiveness to glucose (by standard GSIS ELISA, Alpco Diagnostics [Salem, NH], cat #80-INSMSU-E10). Cells were also routinely tested for mycoplasma by MycoFluor (ABD Biosciences, Rockville, MD, USA). Cells were maintained in 25 mM glucose Dulbecco’s Modified Eagle Medium (DMEM) (Life Technologies, Frederick, MD, USA) supplemented with 10% fetal bovine serum (FBS), 0.001% β-mercaptoethanol, 0.1 mg/mL penicillin, and 0.1 mg/mL streptomycin in 5% CO₂ at 37°C.

Mouse usage followed protocols approved by the Vanderbilt University Institutional Animal Care and Use Committee for GG and IK (protocol #M18000195). Mice were euthanized by isoflurane inhalation. C57BL/6J (Strain #:008637; RRID:IMSR_JAX:008637) mice were from The Jackson Laboratory (Bar Harbor, ME, USA). Conditional knockout Kif5b^fl/−:RIP2-Cre were bred and genotyped as previously described (Cui et al., 2011).

Islets were isolated from 8- to 16-week-old mice. Briefly, ∼2 mL of 0.8 mg/mL collagenase P (MilliporeSigma, St. Louis, MI, USA) in Hanks’ balanced salt solution (Corning, Corning, NY, USA) was injected into the pancreas through the common bile duct. The pancreas was digested at 37°C for 20 min. Islets were handpicked into RPMI 1640 media with 11 mmol/L glucose (Gibco, Dublin, Ireland) plus 10% heat-inactivated (HI) FBS (Atlanta Biologicals, Flowery Branch, GA, USA) and cultured at 37°C with 5% CO₂. For MIN6 cells, DMEM with 25 mmol/L glucose, 0.071 mmol/L β-mercaptoethanol (MilliporeSigma), 10% HI FBS, 100 μU/mL penicillin, and 100 μg/mL streptomycin (Gibco) was used.

Primary antibodies for immunofluorescence were: mouse anti-β-tubulin (Sigma-Aldrich, 1:1000), rabbit anti-β-tubulin (Sigma-Aldrich, 1:1000), and rabbit anti-KIF5B (Abcam), Alexa488-, Alexa568-, and Alexa647-conjugated highly cross-absorbed secondary antibodies (Invitrogen). Coverslips were mounted in Vectashield Mounting Medium (Vector Laboratories). Cells were treated with indicated drugs for 3 hr unless otherwise indicated. Drugs used were: kinesore (Tocris Bioscience).

The Kif5b-targeting shRNA [shRNA Kif5b] #1, [TL510740B, 5’-ACTCTACGGAACACTATTCAGTGGCTGGA] and [shRNA Kif5b] #2, [TL51074CB 5’ – AGACCGTAAACGCTATCAGCAAGAAGTAG] are in the plasmid backbone pGFP-C-shLenti and were from Origene (Rockville, MD, USA). The nontargeting shRNA control was pGFP-C-shLenti, also from Origene.

The Scr shRNA-mEmerald-tubulin, Kif5b shRNA#1-mEmerald-tubulin, Kif5b shRNA#2-mEmerald-tubulin were all generated from their respective TGFP containing constructs. Using the NotI and PmeI sites, the TGFP was swapped for mEmerald-tubulin-C-18.

The FKBP-mCherry-KIF5B(568-964) construct (gift from Kristen Verhey, University of Michigan) has previously been described (Ravindran et al., 2017).

By using site-directed mutagenesis, we made eight point mutations in the tail domain to change residues RKRYQ to AAAYA in the ATP-independent MT-binding domain. The point mutations were sufficient to rescue MT sliding in the cell. As previously published, this disrupts the tail domain to bind to the acidic e-hook of the MT tail. Point mutations were introduced using a site-directed mutagenesis kit, In-Fusion Snap Assembly (Takara).

Lentivirus production and infection followed standard methods (Huang et al., 2018). MIN6 cells were treated with a given shRNA expressing an mEmerald-tubulin/cytosolic eGFP marker for 96 hr prior to imaging to achieve KD efficiency. For nonviral vectors, MIN6 cells were transfected using Amaxa Nucleofection (Lonza), and experiments were conducted 24–48 hr thereafter.

Figure 1: (D–F) Maximum intensity projections of spinning disc confocal stacks through the ventral 1.2 µm of the cell for each time point. (G–I) Single focal plane spinning disc confocal time frames.

Figure 1—figure supplement 1: (A–C) Single focal planes from laser scanning confocal stacks.

Figure 2: (A–C) Maximum intensity projections of laser scanning confocal stacks over the ventral 1 μm of the cell. The tubulin channel (top) is shown as inverted grayscale images. The KIF5B channel (bottom) was pseudo-colored cyan. (F–H) Thresholded single focal plane images under the nucleus of the cell selected for directionality analysis.

Figure 2—figure supplement 1: (A, B) Maximum intensity projections of laser scanning confocal stacks over the ventral 1 µm of the islet. Inverted grayscale images. (D, E) Thresholded single focal plane images under the nucleus of the cell selected for directionality analysis.

Figure 2—figure supplement 2: (A–D) Maximum intensity projections of laser scanning confocal stacks through the whole cell. The tubulin channel (top) is shown as inverted grayscale images. The BFP-motor channel (bottom) was pseudo-colored cyan.

Figure 2—figure supplement 3: Single-plane laser confocal images shown as inverted grayscale images (A1–A2), thresholded images (A4, A5), and MATLAB analysis outcomes (A8). Images in (A6–A7) are non-to-scale schematics.

Figure 2—figure supplement 4: Single-plane laser confocal images shown as inverted grayscale images (A1–A2), thresholded images (A3, A4), and MATLAB analysis outcomes (A5, A6).

Figure 3: Maximum intensity projections of spinning disc confocal stacks through the ventral 1.2 µm of the cell for each time point (C–G).

Figure 4: Cells shown are maximum intensity projections of laser scanning confocal stacks through the ventral 1 μm of the cell. The tubulin channel (top) is shown as inverted grayscale images (A–C). Thresholded single focal plane images under the nucleus of the cell selected for directionality analysis (F–H).

Figure 4—figure supplement 1: (A–B) Cells shown are maximum intensity projections of laser scanning confocal stacks through the ventral 1 μm of the cell. The tubulin channel (top) is shown as inverted grayscale images. mCherry-labeled KIFDN constructs (bottom) are pseudo-colored magenta.

Figure 5: Cells shown are maximum intensity projections of laser scanning confocal stacks of the ventral 1 μm of the cell. The tubulin channel is shown as inverted grayscale images (A–B). Thresholded single focal plane images under the nucleus of the cell selected for directionality analysis (D–E).

Figure 5—figure supplement 2: Cells shown are maximum intensity projections of laser scanning confocal stacks of the bottom 1 μm of the cell. The tubulin channel is shown as inverted grayscale images.

Figure 6: Maximum intensity projections of spinning disc confocal stacks through the ventral 1.2 µm of the cell for each time point (A–B). Single focal plane spinning disc confocal time frames (D–E).

For all images, whole-image contrast for each channel was adjusted equally within each figure.

For all experiments, n per group is as indicated by the figure legend and the scatter dot plots indicate the mean of each group and error bars indicate the standard error of the mean. All graphs and statistical analyses were generated using Excel (Microsoft) and Prism software (GraphPad). Statistical significance for all in vitro and in vivo assays was analyzed using an unpaired t-test, one-way ANOVA with Sidak’s multiple comparisons test, Kolmogorov-Smirnov test as indicated in the figure legends. For each analysis, p<0.05 was considered statistically significant, and *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001.

Materials developed in this study are available upon request from the corresponding author.