FMNL2 and -3 regulate Golgi architecture and anterograde transport downstream of Cdc42

The Rho-family small GTPase Cdc42 localizes at plasma membrane and Golgi complex and aside from protrusion and migration operates in vesicle trafficking, endo- and exocytosis as well as establishment and/or maintenance of cell polarity. The formin family members FMNL2 and -3 are actin assembly factors established to regulate cell edge protrusion during migration and invasion. Here we report these formins to additionally accumulate and function at the Golgi apparatus. As opposed to lamellipodia, Golgi targeting of these proteins required both their N-terminal myristoylation and the interaction with Cdc42. Moreover, Golgi association of FMNL2 or -3 induced a phalloidin-detectable actin meshwork around the Golgi. Importantly, functional interference with FMNL2/3 formins by RNAi or CRISPR/Cas9-mediated gene deletion invariably induced Golgi fragmentation in different cell lines. Furthermore, absence of these proteins led to enlargement of endosomes as well as defective maturation and/or sorting into late endosomes and lysosomes. In line with Cdc42 - recently established to regulate anterograde transport through the Golgi by cargo sorting and carrier formation - FMNL2/3 depletion also affected anterograde trafficking of VSV-G from the Golgi to the plasma membrane. Our data thus link FMNL2/3 formins to actin assembly-dependent functions of Cdc42 in anterograde transport through the Golgi apparatus.


SUPPLEMENTARY FIGURES AND LEGENDS
Figure S1: FMNL2 and -3 can localize to both Golgi and lamellipodia tips simultaneously, albeit at variable extents.
Epifluorescence images of B16-F1 cells co-expressing either FMNL2-EGFP or FMNL3-EGFP and myctagged Cdc42-L61. Cells were also labelled with fluorescent phalloidin (right panel). Pictures on the left demonstrate both formins to associate with the Golgi complex upon co-expression of active Cdc42-L61. However, enhanced contrast adjustment of the same images (leading to over-exposure at the Golgi) reveals both formins to simultaneously localize to lamellipodia tips (middle panel), as previously described 21,23 , not apparent in these experimental conditions when normalizing to the strong accumulation at the Golgi apparatus (left panels). B16-F1 melanoma cells transiently transfected with EGFP-tagged FMNL family member (green in merge on the right) and mCherry-tagged, constitutively active Cdc42 (green in merge) were fixed and counterstained for the actin cytoskeleton using blue phalloidin. Red insets are magnified in images shown in merges on the right (without phalloidin). Note robust co-localization of each FMNL family member with constitutively active Cdc42 (L61) at Golgi membranes and vesicular structures. Superresolution images using 3D-Structured Illumination Microscopy (3D-SIM) of transfected B16-F1 cells, as indicated, revealing a clear perinuclear co-localization of all FMNL members with Golgiresident mCherry-Cdc42-L61. Zoomed images on the right display merged color pictures (FMNL family member, green; mCherry-Cdc42-L61, red) of the perinuclear region in higher magnification (see also Supplementary Movie 1 for 3D-animation of FMNL2-EGFP data). Epifluorescence images of B16-F1 cells transiently transfected with FMNL2-or FMNL3-EGFP in combination with constitutively active Cdc42-L61, the fast cycling mutant Cdc42-L28 or mCherrytagged wildtype Cdc42, as stated in the pictures. Myc-tagged Cdc42 variants were visualized using anti-myc staining. Phalloidin was employed to stain for the F-actin cytoskeleton. Note that all Cdc42 constructs promoted recruitment of FMNL2 and -3 to the Golgi complex (red arrowheads), although to variable extents, with Cdc42-L61 being most and wildtype Cdc42 being least effective. Structured illumination microscopy superresolution images of COS-7 cells transiently transfected with either FMNL2-or FMNL3-EGFP, as indicated, and stained for the cis-Golgi compartment using a GM130-reactive antibody. Color-merged images on the right show an accumulation of these formins in the vicinity of cis-Golgi membranes, comparable to the localization data shown below. GST-pull down assay (using Rho-GTPases as indicated) of EGFP-tagged FMNL1ΔDAD (constitutively active version) ectopically expressed in B16-F1 cells and detected with anti-EGFP antibody, revealing binding of FMNL1 to active, GTP-bound Cdc42, but not dominant negative Cdc42 (N17) or active and inactive versions of Rac1, nor to active RhoA-L63 and Rif-L27 (the latter MBP-tagged).Red rectangle highlights specific pull down of EGFP-FMNL1ΔDAD with active, GST-tagged Cdc42 (L61). (b) Negative control showing that neither GST nor MBP alone was capable of pulling down EGFP-tagged FMNL1ΔDAD. (c) Negative control showing that neither constitutively active Cdc42 nor Rac1 was able to pull down EGFP alone. (a) Spinning disc confocal fluorescence microscopy images of B16-F1 cells transiently co-transfected with FMNL2-EGFP and myc-tagged Cdc42-L61 were counterstained with anti-GM130 (Golgi matrix protein 130) and anti-TGN46 antibodies to define the cis-Golgi compartment (top) and TGN (bottom panels), respectively. For middle panels, FMNL2-EGFP and myc-tagged Cdc42-L61 were cotransfected with mRFP-β1,4-GalT, encoding a β-1,4-galactosyltransferase specifically accumulating in the trans-medial Golgi. Merged images display FMNL2-EGFP (green) and respective Golgi subcompartment (red) as indicated by labelling, and red rectangles mark insets magnified on the very right (zoom). (b) Representative scatter plots of perinuclear regions exemplify overlap quantities between proteins as specified. (c) Box and whiskers plots illustrating results from co-localization analysis, showing Pearson´s correlation coefficients for each of the stainings, as indicated. Boxes include 50% (25%-75%) and whiskers 80% (10%-90%) of all measurements, outliers are shown as dots. Lines within boxes show medians, also given as numbers in red, n represents number of cells analyzed. Figure S8: FMNL3 specifically associates with trans-medial Golgi cisternae.
(a) Spinning disc confocal fluorescence microscopy images of B16-F1 cells transiently co-transfected with FMNL3-EGFP and myc-tagged Cdc42-L61, as indicated, and counterstained with anti-GM130 and anti-TGN46 antibodies to mark cis-Golgi compartment (top) and TGN (bottom panels), respectively. For middle panels, FMNL3-EGFP and myc-tagged Cdc42-L61 were co-transfected with mRFP-β1,4-GalT to visualize the trans-medial Golgi. Merged images display FMNL3-EGFP (green) and respective Golgi subcompartment (red) as indicated by labelling, and red rectangles mark insets magnified on the very right (zoom). (b) Representative scatter plots of perinuclear regions exemplify overlap quantities between proteins as specified. (c) Box and whiskers plots illustrating results from colocalization analysis, showing Pearson´s correlation coefficients for each of the stainings. Boxes include 50% (25%-75%) and whiskers 80% (10%-90%) of all measurements, outliers are shown as dots. Lines within boxes are medians, also given as numbers in red; n, number of cells analyzed.

Figure S9: Golgi-located FMNL3 polymerizes F-actin at this organelle.
B16-F1 cells were co-transfected with FMNL3-EGFP and myc-Cdc42-L61, followed by fixation and phalloidin staining. (a) Representative examples of B16-F1 cells illustrating correlative cases of fomrin accumulation at Golgi versus cell periphery with actin accumulation at the Golgi. Colored rectangles illustrate distinct categories defined, with red and blue rectangles marking cells in which the formin is strongly located at the Golgi complex and harboring (red rectangle) or not (blue rectangle) phalloidinstainable actin filament meshworks around Golgi. The third category shows cells in which the formin is not targeted to the Golgi, but remains exclusively localized at the lamellipodium tip (green rectangle). (b) Summary of incidences of actin accumulation at the Golgi in transfected cells, categorized as shown in (a). Data are displayed as arithmetic means of percentages of transfected cells (± sem) harboring or not harboring discernible Golgi-actin, as indicated. Note that actin accumulation was never found in cells lacking FMNL3-EGFP; n > 100. FMNL2/3 KO B16-F1 cells were co-transfected with mCherry-tagged, constitutively active Cdc42 and the EGFP-tagged N-termini of FMNL2-or 3, both comprising respective N-terminal myristoylation sites and GTPase binding domains. Upon fixation, cells were counterstained with phalloidin, as indicated. Note that expressed FMNL2/3 N-termini overlap with Cdc42 accumulation at plasma and intracellular membranes, including the Golgi (red arrowheads). Recruitment behavior was independent of endogenous, full length FMNLs, as transfections were carried out in the absence of endogenous proteins.

Figure S11: Golgi targeting of FMNL2 is solely stimulated by Cdc42.
Fixed and phalloidin-stained B16-F1 cells transiently co-transfected with FMNL2-EGFP plus constitutively active Rho-GTPases, as indicated. Expression of respective GTPase was validated by anti-myc staining. Cdc42 expression induced Golgi membrane association, as described above (top panel). However, none of the remaining Rho-GTPases tested here, including Cdc42-like GTPases TC10 and TCL, promoted FMNL2 recruitment to the Golgi apparatus. This was true even though the latter two and Wrch appeared to target to vesicular/Golgi-like structures, as expected 61,62 (red arrowheads).

Figure S12: Golgi targeting of FMNL3 is exclusively stimulated by Cdc42.
B16-F1 cells were transiently co-transfected with FMNL3-EGFP and constitutively active Rho-GTPases, as indicated, and counterstained for the actin cytoskeleton with phalloidin. Expression of respective GTPase was validated by anti-myc staining. Cdc42 expression induced Golgi membrane association, as expected (top panel). As for FMNL2, none of the remaining Rho-GTPases tested here, including Cdc42-like GTPases TC10 and TCL, promoted FMNL3 recruitment to the Golgi apparatus. This was true in spite of TC10, TCL and Wrch all targeting to vesicular/Golgi-like structures, as expected 61,62 (red arrowheads).

Figure S13: N-terminal myristoylation is essential for Golgi membrane association of FMNL2 and -3.
FMNL2/3 KO cells were co-transfected with mCherry-Cdc42-L61 and either N-or C-terminally tagged FMNL2 or -3 or a non-myristoylatable point mutant of FMNL2 (FMNL2-G2A-EGFP), as indicated. Nterminal tagging was previously established to block N-terminal myristoylation 23 , and is shown here to inhibit Golgi association of both FMNL2-and 3, in spite of the presence of co-transfected, constitutively active Cdc42. Blocking myristoylation of FMNL2 by mutation of residue 2 to alanine (FMNL2-G2A-EGFP) also eliminated Golgi association, in spite of EGFP fused to C-terminus of this variant, confirming that lack of myristoylation alone is sufficient for inhibiting Golgi accumulation. Potential dimerization with endogenous FMNL formins was avoided through the use of FMNL2/3deficient cells. Note that inhibition of N-terminal myristoylation did not block lamellipodial targeting of non-myristoylated FMNL variants (red arrowheads), consistent with previous data 21, 23 . Note that Cterminally tagged FMNL2 and -3 can also target to the lamellipodium, but this is frequently masked in digital images displaying strong Golgi accumulation.

Figure S14: Combined knockdown of FMNL2 and -3 causes pronounced Golgi complex dispersal.
(a) Western Blot demonstrating efficiency of RNAi-mediated protein rundown using a FMNL2/3 reactive antibody and GAPDH as loading control. (b) Left panel: Representative images of compact (blue), spread (red) and dispersed (green) Golgi morphologies, exemplifying categories used for quantifications shown in bar graph on the right. GM130 was used as Golgi marker. Right panel: Bar graph summarizing quantitation of Golgi morphology categories in different experimental conditions, as indicated. Data are arithmetic means and standard errors of means; n equals number of cells analyzed; t-test confirmed statistically significant differences for category: "fragmented". Although differences in individual FMNL2 or FMNL3 knockdown populations already appeared statistically significant, the changes were comparably modest. However, differences observed for double knockdown cells were strong and statistically highly significant, indicating overlapping functions for the two formins in promoting Golgi compaction.   Representative examples of NIH 3T3 fibroblasts stained for GM130 and photographed by widefield fluorescence microscopy (left panel). Right panel: Software-aided identification of separable objects (red boundaries) of GM130-staining (green), with numbers of quantified objects provided in each case. d and e, f). (a, b) Images depict the cytoplasm of control cells close to the nucleus with a distinct Golgi region composed of stacks of cisternae in a well ordered, parallel organization (white asterisks with black frames). (c, d) and (e, f) In FMNL2/3 KO cells, ministacks of short, loosely arranged cisternae and contorted membranes, surrounded by vesicles, are dispersed in the cytoplasm (black asterisks with white frames). More regularly organized stacks and disorganized mini-stacks can be seen located side by side. Nu -nucleus; arches in panels (a) and (b) indicate the distinct Golgi regions apparent in the control cells.

Figure S18: Genetic disruption of FMNL2 and -3 in NIH 3T3 fibroblasts causes Golgi dispersal and fragmentation.
Representative Structured Illumination Microscopy images of control NIH 3T3 and FMNL2/3 double KO cells stained for the Golgi apparatus using a GM130 marker. Two representative examples are shown for each cell line. Note that in case of both KO clones, the Golgi structure appears much more fragmented compared to control fibroblasts (see also Supplementary movies for 3D-reconstructions of the data).  (a) Representative examples of Rab7 (LE marker)/LAMP1 (lysosomal marker) and phalloidin triplestainings of control NIH3T3 and FMNL2/3 KO clones, as indicated. Merges of Rab7 (green) and LAMP1 (red) are shown on the right. Note significant co-localization of both markers and a more diffuse, less confined staining of Rab7 and LAMP1 in case of FMNL2/3 KO cells. (b) Categorization of Rab7 stainings (let panel) and corresponding quantification depicted on the right; n corresponds to examined cells; dara are arithmetic means ± sem. Differences in Rab7 categorization as explored for parameter "diffuse" were determined to be statistically significant using t-test. Quantitative analysis of average cell surface surface intensities of VSV-G-EGFP (arbitrary units) at 60 and 120 min after shift to 32°C (permissive temperature) for control and FMNL2/3 KO cells, as indicated. Data are arithmetic means and ± sem from four independent experiments, n equals cell numbers analyzed; statistics was done using non-parametric Mann-Whitney rank sum test. Note plasma membrane association of VSV-G-EGFP to be reduced in both FMNL2/3 double KO clones in a statistically significant fashion 2h after temperature shift.

Supplementary Movie 4
FMNL2-EGFP associates with intracellular vesicles. Time-lapse fluorescence and phase contrast microscopy of B16-F1 cell migrating on laminin and transiently transfected with FMNL2-EGFP, accumulating at lamellipodia and filopodia tips, as expected. Additionally, however, FMNL2-EGFP targeted to various, highly dynamic vesicles displaying typical, saltatory and bouncing movements within the cell. Time is given in minutes and seconds. Scale bar corresponds to 10 µm.

FMNL3-EGFP accumulates at intracellular vesicular structures.
A B16-F1 cell migrating on laminin and transiently expressing FMNL3-EGFP was recorded by time-lapse fluorescence microscopy and phase contrast optics, as indicated. In analogy to FMNL2-EGFP, FMNL3-EGFP displays prominent accumulation at highly dynamic vesicles. Time is given in minutes and seconds. Scale bar, 10 µm.

Supplementary Movie 6
Morphology of the cis-Golgi in NIH control fibroblast, example 1.
Animated 3D-SIM volume view of a reconstructed stack of an NIH3T3 cell stained for the cis-Golgi marker GM130. Note that visible Golgi cisternae are mostly connected, building continuous Golgi ribbons.

Supplementary Movie 7
Morphology of the cis-Golgi in NIH control fibroblast, example 2. For legend see Supplementary Movie 6.
Supplementary Movie 8 FMNL2/3 removal causes a fragmented Golgi morphology, clone #9/10, example 1. Representative, animated 3D-SIM volume view of a reconstructed stack of a cell representing NIH3T3 clone #9/10, genetically deleted for FMNL2 and -3 and stained for the cis-Golgi marker GM130. Note that the cis-Golgi compartment appears highly fragmented as compared to control NIH 3T3 cells (compare Supplementary Movies 6 and 7). Representative, animated 3D-SIM volume view of a reconstructed stack of a cell representing NIH3T3 clone #46/20, genetically deleted for FMNL2 and -3 and stained for the cis-Golgi marker GM130. Note that the cis-Golgi compartment appears highly fragmented as compared to control NIH 3T3 cells (compare Supplementary Movies 6 and 7).

Supplementary Movie 12
FMNL2-EGFP dynamics during lamellipodium protrusion and membrane ruffling. Time-lapse epifluorescence imaging of NIH 3T3 cell plated on fibronectin and transiently co-expressing FMNL2-EGFP and constitutively active, mCherry-tagged Cdc42-L61. The panel on the right shows FMNL2-