Low Rho activity in hepatocytes prevents apical from basolateral cargo separation during trans-Golgi network to surface transport

Abstract

Hepatocytes, the main epithelial cells of the liver, organize their polarized membrane domains differently from ductal epithelia. They also differ in their biosynthetic delivery of single-membrane-spanning and glycophosphatidylinositol-anchored proteins to the apical domain. While ductal epithelia target apical proteins to varying degrees from the trans-Golgi network (TGN) to the apical surface directly, hepatocytes target them first to the basolateral domain, from where they undergo basolateral-to-apical transcytosis. How TGN-to-surface transport differs in both scenarios is unknown. Here, we report that the basolateral detour of a hepatocyte apical protein is due, in part, to low RhoA activity at the TGN, which prevents its segregation from basolateral transport carriers. Activating Rho in hepatocytic cells, which switches their polarity from hepatocytic to ductal, also led to apical-basolateral cargo segregation at the TGN as is typical for ductal cells, affirming a central role for Rho-signaling in different aspects of the hepatocytic polarity phenotype. Nevertheless, Rho-induced cargo segregation was not sufficient to target the apical protein directly; thus, failure to recruit apical targeting machinery also contributes to its indirect itinerary.

1 |. INTRODUCTION

Hepatocytes, the parenchymal epithelial cells of the liver, present with a unique polarity phenotype: Rather than establishing a single luminal and basal domain as is typical for duct- or ascini-forming epithelial cells, hepatocytes contact blood vessels at two basal domains. They furthermore feature not one but multiple luminal surfaces. Interrupting their cell-cell contacting domains, the hepatocyte lumina form a branched canalicular network, called the bile canaliculi, which serve as conduits for the transport of excreted bile acids into the bile duct. Hepatocyte polarity is thus optimized for two central metabolic functions: the intense bi-directional exchange of macromolecules with the blood and the circulation of bile acids between the liver and the duodenum, where they aid in digestion.

Hepatocytes are not only unique in how they organize their surface domains but also in how they target newly-synthesized apical proteins. In studies conducted in the 1980, Hubbard’s group combined in vivo metabolic pulse-chase labeling of newly synthesized liver proteins with the isolation of hepatocyte apical and basal membrane domains to demonstrate that hepatocytes target native single-membrane spanning and glycophosphatidylinositol (GPI)-anchored apical proteins from the trans-Golgi network (TGN) first to the basolateral domain from where they then reach their resident surface via basolateral-to-apical transcytosis.^1,2 Although all epithelia studied so far possess a transcytotic pathway, none appears to utilize it for biosynthetic apical protein delivery to the same extent as hepatocytes.³ Instead, most epithelial cells target apical proteins to varying extents from the TGN directly to the apical domain.⁴ In some instances, the same apical protein, for example, dipeptidyl peptidase IV (DPPIV), was shown to be directly targeted in Madin-Darby canine kidney (MDCK) cells^5,6 but indirectly, that is, via transcytosis, in hepatocyte-derived WIF-B9 cells^7,8 or HepG2 cells.⁹

In the direct targeting mode, apical and basolateral proteins are segregated at the TGN and packaged into separate post-TGN transport carriers with distinct apical and basolateral destinations. While the sorting and targeting mechanisms underlying this delivery mode are becoming increasingly clear,^10–13 we still do not understand even on a basic level how TGN-to-surface targeting differs in hepatocytes. Two principle scenarios can be envisioned: (a) Hepatocytes either fail to segregate apical and basolateral cargo into distinct carriers, generating a common basolaterally-directed transport carrier instead, or (b) they do package apical and basolateral proteins into separate post-TGN transport carriers but target both types of carriers to the basolateral surface.

To resolve which of these two scenarios applies, we analyzed the synchronous passage of an apical-basolateral cargo pair, DPPIV and vesicular stomatitis virus G (VSVG) protein, through the secretory pathway of hepatocytes and hepatocyte-derived WIF-B9 cells by dual-color live and fixed cell imaging. Our results reveal that hepatocytes target the apical and basolateral model cargoes from the TGN to the cell surface in common carriers. We further determined that this is due to lack of a Rho-dependent sorting mechanism, which operates by fragmenting TGN-derived tubules that contain apical and basolateral domains into separate apical and basolateral carriers. Finally, we established that segregation into distinct TGN-derived carriers by itself is not sufficient to re-direct the DPPIV cargo into the direct apical targeting pathway.

2 |. RESULTS

2.1 |. DPPIV and VSVG undergo synchronous endoplasmic reticulum to basolateral surface transport in polarized WIF-B9 cells

We initially studied apical-basolateral protein sorting during endoplasmic reticulum (ER) to surface transport in polarized hepatocytic WIF-B9 cells,¹⁴ since this cell line has been shown to faithfully mimic hepatocyte protein trafficking.¹⁵ We established a pulse-chase regimen for an apical-basolateral cargo pair as shown in Figure 1A,B: GFP-tagged DPPIV an apical hepatocyte protein whose indirect targeting is well established, and mTagBFP-tagged VSVG, a basolateral model cargo, were co-expressed and accumulated in the ER via DPPIV aggregation through an engineered conditional aggregation domain (CAD)¹⁶ and VSVG by exploiting the aggregation of the ts045 mutant at its nonpermissive temperature of 40° C.¹⁷ Upon release from aggregation, the cargoes were staged in the TGN by incubation at 20°C, a temperature that prevents their TGN exit.¹⁸ TGN-to-surface transport was subsequently induced by a temperature shift to 37°C. Importantly for the analysis of apical-basolateral sorting, DPPIV and VSVG arrived at and exited the TGN with similar kinetics. We conclude this from their near-identical co-localization dynamics with the TGN marker TGN38 (Figure 1D,E). After being released from the TGN for 1 hour, both cargoes appeared at the basolateral domain (Figure 1B) and were absent from the luminal surface. After 2 hours at 37°C, a cohort of DPPIV had undergone transcytosis to the lumen. This was apparent from the shift in GFP fluorescence (Figure 1B, compare luminal GFP at 1 and 2 hours at 37°C) and corroborated by basolateral-to-luminal transport of a DPPIV antibody. The antibody, which is directed against the protein’s ectodomain (light brown labeling in Figure 1B) bound basolaterally localized DPPIV (Figure 1B, 1 hour) and was subsequently chased with its antigen to the luminal domain (Figure 1B, 2 hours). Importantly, we previously established¹⁹ and verified here (Figure S1A) that in contrast to WIF-B9 cells, the same recombinant DPPIV protein is targeted to the apical domain directly in MDCK cells.

FIGURE 1.

DPPIV and VSVG exit the ER and traverse the secretory pathway in a synchronous fashion. (A) Schematic of our main experimental design to study simultaneous biosynthetic targeting of apical and basolateral cargo in hepatocytic cells: (1) Cells co-transduced with adenoviruses expressing apical DPPIV-CAD-GFP and basolateral VSVG-ts045-mTagBFP retain both cargo proteins in the ER at 40°C. (2) After temperature switch to 20°C in the presence of the AP21998 ligand, which relieves CAD-mediated aggregation, and cycloheximide both cargoes are released from the ER and accumulated in the TGN. (3) Cells are incubated at 37°C and the exit and targeting of apical and basolateral cargo from the TGN to their respective domains are analyzed. (B) Biosynthetic itinerary of DPPIV-CAD and VSVG-ts045 in WIF-B9 cells. Note, that both proteins appear at the lateral domain after a 1-hour chase at 37°C, while only DPPIV but not VSVG appears in the lumen after a 2-hour chase. DPPIV basolateral-to-apical transcytosis was established by surface binding and chase of an antibody against its ectodomain (light brown). ZO-1 labeling demarcates the luminal domain. (C) Selected frames from a time-lapse experiment of cells expressing DPPIV-Dendra2 at 37°C after the exit from the 20°C Golgi block. The photoconverted pool of DPPIV-Dendra2 (red) at the Golgi apparatus is targeted indirectly to the apical domain/lumen. (D and E) Kinetics of the apical or basolateral cargo arrival to the TGN after temperature switch from 40°C to 20°C (D) or exit from the 20°C Golgi block (E). The colocalization of both cargoes and the TGN stained with a TGN38 antibody was quantified at the indicated times (bar graphs). n = 15 cells/experiment were analyzed from three independent experiments. Data shown are means + SD. n.s., not significant analyzed by t-test. (B and C) Black or white arrowheads indicate the lumen and the localization of the apical cargo at the lateral (1 hour) or the apical domain (2 hours), respectively

FIGURE 2.

DPPIV and VSVG exit TGN and arrive at the surface in common TGN-derived tubules. Fixed cell images (A,B) and stills from time-lapse sequences (C-E) of WIF-B9 cells (A-D) or primary hepatocytes (E) co-expressing DPPIV-CAD-GFP and mTagBFP-or SNAP-tagged VSVG-ts045 15 minutes after the release of the 20°C TGN exit block. Yellow arrowheads indicate TGN-derived tubules, in which VSVG and DPPIV are colocalized. (A) Confocal image centered on the Golgi region of a cell labeled for TGN38; the magnified area on the right, of a TGN tubulation (white arrowhead) shows colocalization of DPPIV and VSVG. (B) Frames from a 3D-SIM image sequence (see Movie S1) showing a DPPIV and VSVG positive tubule (orange arrowheads) spanning from the Golgi to the cell surface (indicated by the dashed line). (C,E) Frames from confocal dual-color time-lapse sequences of TGN exit of DPPIV-CAD-GFP and VSVG-ts045-mTagBFP in tubular structures that move away from the Golgi area (see Movies S2 and S4). (E, graph) Classification of tubular and vesicular TGN-derived carriers from time-lapse experiments as containing either DPPIV only, VSVG only or DPPIV and VSVG cargo proteins. n = 30 cells from at least three independent experiments were analyzed. Data shown are means - SD. (D) Frames from a TIRF dual-color time-lapse sequence of DPPIV-CAD-GFP and VSVG-ts045-SNAP-TMR* dual positive carriers arriving at the basal surface (see Movie S3). Left images in C-E correspond to x-y-t projections (x-y-t_p) of the selected time points from the time-lapse sequences (eg, 0-6 seconds in C)

We also established a transcytotic targeting pattern in WIF-B9 cells for a Dendra2-coupled DPPIV protein that lacked the CAD moiety (Figure 1C). The pulse in this case was set by photo-converting the TGN-accumulated population after a 20°C block (see red channel in the “Post-conversion” panel of Figure 1C). During a chase at 37°C photoconverted DPPIV was first detectable at the basolateral domain (white arrowhead at 1-hour chase) before it accumulated at the luminal domain (white arrowhead at 2-hour chase). When a basolateral population was photoconverted, it appeared at the luminal domain after a 1-hour chase, indicating that basolateral-to-apical transcytosis had occurred (Figure S1B).

2.2 |. Plasma membrane transport of DPPIV and VSVG in WIF-B9 cells and hepatocytes occurs in common TGN-derived tubules

Having validated our tools, we analyzed DPPIV and VSVG exit from the TGN of WIF-B9 cells 15 minutes after releasing the 20°C exit block. Fixed cell images were taken by confocal microscopy (Figure 2A) and higher resolution three-dimensional (3D) structured illumination microscopy (SIM) (Figure 2B and Movie S1) revealed that DPPIV and VSVG emerged from the Golgi mass in the same tubules, which were also positive for the TGN marker TGN38 (Figure 2A). Characteristic for these TGN-derived tubules was that they are extremely long, often spanning from the Golgi to the cell surface (Figure 2B). Both proteins accumulated in boli along these tubules (Figure 2A,B, yellow arrowheads) as has previously been observed for TGN-derived cargo in other cell types (eg, ^20–22). To determine whether the two cargoes stayed together or segregated during transport to the cell surface, we conducted dual-color live cell imaging, both in confocal, and total internal reflection fluorescence (TIRF) mode. Quantification of TGN exit events revealed that most transport carriers, which severed from and moved away from the Golgi, contained both cargoes (Figure 2C, Movie S2 and Figure 4A; Control). TIRF mode imaging limits the focal distance to 100 nm from the coverslip and hence only captures transport carriers when they are close to the basal domain. We performed live dual-color TIRF imaging 15 minutes after release of the 20°C block and observed that both cargoes were still in common carriers when they arrived at the basal domain (Figure 2D, Movie S3 and Figure 4B; Control). From the collective imaging data, we conclude that in WIF-B9 cells, DPPIV and VSVG are targeted from the TGN to the basolateral surface in common transport carriers.

FIGURE 4.

RhoA activation increases VSVG-DPPIV segregation into distinct TGN-derived carriers. Selected frames from confocal (A) or TIRF (B) dual-color time-lapse sequences after the release of DPPIV-CAD-GFP and VSVG-ts045-mTagBFP (A; see Movies S2 and S6) or VSVG-ts045-SNAP-TMR* (B; see Movies S3 and S7) from the 20°C temperature block in control and Rho activated WIF-B9 cells. Left images correspond to x-y-t projections (x-y-t_p) of selected time points from the time-lapse sequences (eg, 26,32,36 seconds in A Control). Magnified areas (right) show TGN-derived carriers that moved away from the Golgi area (A) or arrive to the basal surface (B) as dual (Control) or single (+ Rho Activator) cargo carriers. Green, red or yellow arrowheads indicate either apical, basolateral or dual cargo carriers, respectively. Classification of TGN-derived carriers at the peri-TGN area (A) or at the basal surface (B) containing either a single or both cargo for n = 30 cells from at least three independent live cell imaging experiments. Data shown are means - SD. *, P ≤ .05; **, P ≤ .01, when compared to the control cells and analyzed by t-test

We verified this key finding in primary rat hepatocytes, which released DPPIV and VSVG in carriers morphologically similar to those in WIF-B9 cells (Figure 2E and Movie S4) and, by quantification, had a comparable majority fraction of carriers containing both cargoes (compare graph in Figure 2E to that for WIF-B9 controls in Figure 4A).

2.3 |. Rho activation in WIF-B9 cells decreases the length of TGN-derived tubules and increases VSVG-DPPIV segregation into distinct carriers

Intrigued by the extreme length of the WIF-B9 and hepatocyte TGN-derived transport carriers, we tested whether the tubule phenotype resulted from the simultaneous transport of more than one recombinant cargo through the secretory pathway or was due to excessive cargo accumulation in the TGN during the 20°C block. We observed however that DPPIV and VSVG formed extensive TGN-derived tubules in WIF-B9 cells even when expressed individually (Figure S2A,B and Movie S5) and when chased through the secretory pathway without accumulation in the TGN (Figure S2C,D and Movie S5). As observed under these experimental conditions the WIF-B9 TGN-derived tubules were reminiscent of a scission defect caused by inhibition of Rho or its effectors described in several cell types.^21,23–25 Rho was furthermore of interest because we previously found that Rho activation switched WIF-B9 cell polarity to a ductal phenotype.²⁶

We therefore employed the same Rho activator (Rho activator II) used on WIF-B9 cells in this earlier study to determine if Rho activation affected their TGN tubulations and TGN-derived carriers. This activator, which deaminates Gln63 in RhoA, thereby blocking GTP-hydrolysis,²⁷ led to prominent RhoA recruitment to the WIF-B9 cell Golgi area (Figure 3A) in addition to activating the Rho effector myosin II in stress fibers (detected by myosin light chain II phosphorylation, [P-MLC2]), which we associated with the “ductal” polarity switch (Figure 3C²⁶). RhoA recruitment to the Golgi was accompanied by higher RhoA activity in the TGN, as measured with a Förster resonance energy transfer (FRET)-based RhoA activity sensor²⁸ (Figure 3B). At steady state, that is, in the absence of recombinant cargo and without 20° C accumulation, Rho activation significantly reduced the number of GM130- and TGN38-positive tubulations emanating from the WIF-B9 cell Golgi as determined in fixed cells (Figure 3D). The Rho activator also made the Golgi apparatus and TGN ribbon more focused around the nucleus (Figure 3D). The tubulated Golgi phenotype in untreated cells was independent of WIF-B9’s polarization state as it also occurred in cells that lacked a luminal domain and hence were not yet fully differentiated (Figure 3D, “Non polarized”).

FIGURE 3.

The hepatocyte Golgi tubulation result from low Rho activity. (A-D) Control and Rho activated WIF-B9 cells were analyzed for RhoA and P-MLC2 localization (A and C respectively), RhoA activity (B) and morphology of the Golgi apparatus (D). Rho activated cells show significantly increased endogenous RhoA (A) and RhoA biosensor activity (B) at the TGN, which is associated with reduced Golgi tubulation (D). Note that Rho activation increases stress fibers (gray arrowheads in panel A) and shifts activated myosin II from the contractile hepatocytic luminal domain (blue arrowhead) to basal stress fibers (C). (A, B and D) n = 20-30 cells/experiment were analyzed from three independent experiments. Data shown are means + SD. *, P ≤ .05; **, P ≤ .01; n.s., not significant analyzed by t-test. (A-D) Black or white arrowheads indicate the hepatocytic luminal domain or Golgi tubulations, respectively

We next analyzed DPPIV and VSVG exit from the TGN and basal surface arrival in the presence of the Rho activator. Live cell imaging revealed that the cargoes exited in smaller transport carriers (Figure 4A; + Rho Activator and Movie S6) when compared to controls (Figure 4A; Controls, Figure 2C and Movie S2). Quantitation revealed that Rho activation not only reduced the fraction of DPPIV and VSVG exiting the TGN in dual cargo carriers but also significantly decreased tubule length and shortened the time it took for tubular transport carriers to undergo scission from the Golgi mass (Table 1). This indicates that Rho activation in WIF-B9 cells promotes fission from the Golgi of TGN-derived carriers and might contribute to their fragmentation. Consistent with the interpretation that cargo exited in smaller carriers upon Rho activation, the total number of TGN-derived carriers counted in 1-minute time-lapse sequences from 30 cells in WIF-B9 control cells was 266 compared to 385 in their Rho activated counterparts

TABLE 1.

Features of TGN-derived tubular transport carriers in control vs Rho activated WIF-B9 cells

	Tubular carriers (%)	Length (μm)	Time (s)
Control	37.0 ± 3.9	5.0 ± 0.6	33.5 ± 6.1
+ Rho Activator	13.3 ± 3.4**	3.0 ± 0.2**	13.4 ± 5.7*

Note: “Tubular carriers”: defined as carriers of ≥2 μm in length, as percentage of total TGN exit events/cell. “Length”: length of tubules at time of scission from the Golgi mass. “Time”: time from emergence to scission of the TGN-derived tubules in 1-minute time-lapse sequences. Data were compiled from VSVG and DPPIV co-expressing cells. 77.5% of tubules in control and 54% of tubules in Rho activated cells contained both cargoes. Values indicate mean ± SD from n = 30 cells from at least three independent live cell imaging experiments.

^*P ≤ .05,

^**P ≤ .01 as analyzed by t-test comparing control to Rho activator treated cells.

Remarkably, we found that the generation of smaller transport carriers was accompanied by an increased segregation of VSVG and DPPIV into distinct carriers (Figure 4). This increased cargo separation was consistently observed both by confocal time-lapse imaging of carriers exiting the TGN (Figure 4A and Movie S6) and by TIRF time-lapse imaging of carriers in the vicinity of the basal surface (Figure 4B and Movie S7). We conclude that low Rho activity in WIF-B9 cells is the main cause for both, Golgi tubulation and the lack of apical-basolateral cargo separation.

2.4 |. Segregation from VSVG-containing post-TGN carriers is not sufficient for TGN-to-apical surface targeting of DPPIV

Segregation of DPPIV- and VSVG into separate transport carriers is a pre-requisite for their targeting to different membrane domains; we wanted to know if it is also sufficient in Rho activator treated WIF-B9 cells to direct DPPIV from the TGN to the apical instead of the basolateral surface. We addressed this question with two approaches.

First, we utilized the pulse-chase regimen in which DPPIV-CAD-GFP is sequentially held in the ER and TGN, and therefore none of the protein is present at the cell surface prior to its release from the TGN exit block (Figure 5A). We determined that 30 minutes after lifting the TGN block a measurable fraction of DPPIV cargo had arrived at the lateral domain. This was the case in both control and Rho activator treated WIF-B9 cells (Figure 5C). When we quantified the lateral GFP-fluorescence intensities normalized to their TGN levels, no difference in basolateral surface delivery was discernable between Rho activated and control cells (Figure 5C, graph). We also verified that the basolaterally delivered DPPIV population in Rho activated cells underwent basolateral to apical transcytosis by employing our antibody-binding and transport assay described in Figure 1B. Note that, as we reported previously,²⁶ the Rho activator changes the polarity phenotype of WIF-B9 cells from hepatocytic, that is, with tight junctions encircling lumina between cells, to ductal, that is, with tight junctions organized in a “chickenwire” pattern and the luminal domain at the apex. As seen in the x-z views of these ductal-polarized Rho activated WIF-B9 cells (Figures 5A,D and S3), when DPPIV ectodomain antibody was added 1 hour after initiation of TGN exit, the DPPIV cargo was detected at the basolateral domain (1-hour time point, labeled for β-catenin in Figure 5D). When the surface-bound antibody was then chased for an hour at 37°C (2-hour time point in Figures 5A,D and S3), staining was also detected at the apical surface (confirmed by PKC ζ and ezrin labeling in Figure S3), consistent with DPPIV undergoing basolateral-to-apical transcytosis. In our second approach, we interrogated in Rho activator treated WIF-B9 cells the itinerary of a photoconverted TGN population of DPPIV-Dendra2 (Figure 5B). As in control cells (compare to Figure 1C) the photoconverted protein appeared at the lateral domain 1-hour post-conversion, and at the apex, that is, the luminal domain, 2-hour post-conversion. Collectively, these observations indicate that apical delivery of DPPIV still occurs via basolateral-to-apical transcytosis even when the protein becomes segregated from basolateral cargo in post-TGN carriers.

FIGURE 5.

Segregation of DPPIV from VSVG-containing post-TGN carriers is not sufficient for direct apical targeting. (A-D) Indirect targeting from the TGN of DPPIV-CAD-GFP (A,C,D) or DPPIV-Dendra2 (B) in WIF-B9 cells treated with a Rho activator. (A) ER-to-surface chase of DPPIV and VSVG as described in Figure 1A,B. DPPIV transcytosis from the basolateral to the apical domain was established by surface antibody binding and chase (light brown) shown in both x-y and lateral x-z views. Note that the Rho activated WIF-B9 cells establish simple polarity with their circumferential tight junction belt (labeled by ZO-1, white) oriented parallel to the substratum. (B) Selected frames from a time-lapse experiment of cells expressing DPPIV-Dendra2 show the sequential lateral and then apical enrichment of a DPPIV-Dendra2 pool (red) that had been photoconverted at the Golgi apparatus. (C) The amount of DPPIV-CAD-GFP (spectrum map) at the lateral domain as fraction of the remaining pool at the TGN (purple arrowheads) was quantified at the indicated times of chase from the TGN (right). n = 20 cells/experiment were analyzed from three independent experiments. Data shown are means + SD. n.s., not significant analyzed by t-test. (D) DPPIV ectodomain antibody was bound to DPPIV arriving at the basolateral surface after release of the TGN exit block (1 hour) and then chased for an additional 1 hour at 37°C (2 hours). The lateral domain was detected with antibodies against β-catenin protein. (A-D) Black or white arrowheads indicate the lateral lumen (C,D; Controls) or the localization of the apical cargo at the lateral (30 minutes, 1 hour) or apical (2 hours) domain (A-D), respectively

3 |. DISCUSSION

Despite being discovered some 30 years ago, the molecular causes for the basolateral detour in the biosynthetic delivery of apical proteins in hepatocytes have remained largely unknown. Our study, using the apical protein DPPIV as example, provides two fundamental insights into this question: (a) We show that DPPIV is targeted from the TGN to the plasma membrane with the basolateral protein VSVG in common transport carriers; therefore both proteins are a priori delivered to the same surface domain. Failure to separate from basolateral cargo and to generate distinct TGN-derived apical transport carriers could in principle account for DPPIVs basolateral delivery. (b) We determined however that forcing the breakup of the common carriers into separate DPPIV and VSVG-positive vesicles in itself is not sufficient to re-direct DPPIV from the TGN to the apical domain. Instead, vesicles resulting from the common tubules undergo basolateral targeting, regardless of the cargo present. In our TGN exit time-lapse experiments, the separate DPPIV and VSVG carriers appear to have the same trajectories, suggesting that they move along common microtubule tracks, which would carry them to the same surface domains. It should be pointed out that hepatocytes do possess the machinery for TGN-to-apical surface transport; it is utilized for the targeting of apical ABC transporters, a class of polytopic membrane proteins that include bile acid transporters.²⁹ Entry into this pathway might require sorting into TGN-tubules that are distinct from those containing both VSVG and DPPIV.

To date, two experimental manipulations in hepatocytic cells lines have been reported to shunt DPPIV from the indirect into the direct pathway: a p27^(Kip) mutation in the cell line HepG2³⁰ and the recombinant expression of myelin and lymphocyte protein 1 (MAL1), a multi-spanning membrane protein that clusters proteins into lipid rafts, in WIF-B9 cells.⁸ The mechanism for the first claim has not been investigated; as for MAL1 expression, it induced lipid raft association of DPPIV and several GPI-anchored Golgi cargo^8,31; lipid rafts mediate sorting of certain proteins into the direct apical pathway of MDCK cells.³² However, MAL1 is not present in the liver and its exogenous expression in WIF-B9 cells also inhibited the efficient surface transport of apical proteins, leading to their intracellular accumulation at steady state.⁸ Thus, the machinery that distinguishes sorting of ABC transporters from that of other apical proteins in hepatocytes still remains to be elucidated.

Our finding that Golgi-associated Rho activity in hepatocytes determines whether DPPIV and VSVG are targeted to the cell surface in common or in separate TGN-derived transport carriers is remarkable in two respects:

First, while RhoA effectors have previously been implicated in the scission of either apical or basolateral transport carriers from the Golgi and their subsequent fragmentation, we demonstrate here that Rho-dependent scission serves to separate apical from basolateral cargo during TGN-to-surface targeting. Known Rho effectors at TGN-derived tubules include Rho Kinase and its target myosin II,^21,33 LIMK1^24,34 and mDia.²³ Recently, Golgi-associated protein kinase D (PKD) was identified as another RhoA target.³⁵ In MDCK cells, myosin II and PKD have been linked to the exit or scission of VSVG-containing carriers^25,33 while LIMK1 activates the scission of apical carriers.²⁴ We consider it likely that upon treatment with Rho activator several of these effectors operate on the hepatocytic DPPIV-VSVG containing tubules to yield the distinct transport carriers. We cannot rule out that our Rho activator II also targeted RhoD, an atypical Rho GTPase and besides RhoA the only other Rho protein localized to the Golgi.³⁶ However, the reported effects of RhoD manipulation are less consistent with our phenotypic observations than those established for RhoA. Whether activating scission factors to execute the physical separation of pre-established apical and basolateral membrane domains is the sole Rho function here or whether Rho additionally contributes to DPPIV-VSVG sorting remains to be established. Determining whether Rho activation promotes further sorting would require a 3D quantitative electron microscopy analysis which is beyond the scope of the present study.

The other remarkable aspect is that the common DPPIV-VSVG carriers described here represent the third hallmark of the hepatocyte polarity phenotype that we found to be regulated by RhoA. In previous work, we reported that the hepatocytic and ductal lumen organization, presented by WIF-B9 and MDCK cells, respectively, can be switched to the respective other phenotypes by manipulating RhoA levels or activity.^26,37 Similarly, their distinct mode of cell division—ductal cells bisect their luminal surface in cytokinesis while hepatocytes do not—was governed by RhoA.³⁸ In all three cases, low Rho activity was associated with the hepatocyte-specific features while Rho activation promoted the ductal phenotype. Our new findings thus affirm a central role for Rho-signaling in the hepatocytic polarity phenotype. At the same time, while Rho activation switched WIF-B9 lumen polarity to ductal, it did so without switching DPPIV targeting from indirect to direct. That the switch to ductal polarity was not strictly associated with its direct apical targeting is not surprising since at least a fraction of DPPIV is indirectly targeted also in ductal-polarized intestinal Caco-2 cells.³⁹ A different question is whether the unique lumen polarity of hepatocytes is contingent on their predominant indirect apical targeting mode. Addressing it requires knowledge of the mechanisms that govern cargo inclusion into or exclusion from the direct hepatocyte targeting pathway. The advances we report here provide a framework for future investigations into this question.

4 |. MATERIAL AND METHODS

4.1 |. Cell culture

WIF-B9 cells (provided by A. Hubbard, John Hopkins University School of Medicine, Baltimore, Maryland, and D. Cassio, Universite Paris-Sud, Orsay, France) were grown in modified F-12 Coon’s modification medium (F6636; Sigma) supplemented with 5% (vol/vol) FBS (100-106; Gemini), 1% GlutaMAX (35050-061; Gibco), 0.5 μg/mL amphotericin, and 5 mM HEPES buffer. WIF-B9 cells were cultured in the presence of 10⁻⁵ M hypoxanthine, 4 × 10⁻⁸ M aminopterin, and 1.6 × 10⁻⁶ M thymidine. MDCK cells (provided by K. Mostov, University of California, San Francisco, California) cells were grown in Dulbecco’s Modified Eagle Medium [DMEM] (10-013-CV; Corning) supplemented with 10% FBS (S11550; Atlanta Biologicals), 2 mM l-alanyl-l-glutamine (25-015-CI; Corning), and 5 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES) buffer. Isolation of rat hepatocytes was performed following the two step perfusion method described in Neufeld 1997.⁴⁰ Hepatocytes were cultured in DMEM without phenol red (17-205-CV; Corning), supplemented with 10% fetal bovine serum (FBS), 20 μg/mL insulin, 30 ng/mL glucagon, 15 μg/mL hydrocortisone and 1% penicillin-streptomycin. Cells were maintained at 37°C in a 5% (MDCK and hepatocytes) or 7% (WIF-B9) CO₂ humidified atmosphere with daily change in culture medium. For culture maintenance, cells were seeded at ~7 × 10³ cells/cm² and cultivated up to 2 (MDCK) or 4 days (WIF-B9) before replating. For experiments, polarized WIF-B9 cells (10 to 12 days in culture) or MDCK cells were plated at 1.5-2.0 × 10⁵ cells/cm² on German glass cover slips (CLS-1760-012; Chemglass), MatTek glass-bottom dishes (P35G-1.5-14-C; MatTek Corporation) or 12 mm Transwells with 0.4 μm pore polyester membrane inserts (3460; Corning). For experiments described in Figures 2–4 and S2, cells were plated at 0.5 × 10⁵ cells/cm² on German glass cover slips or MatTek dishes prewashed with an aqueous solution of collagen type I (20 μg/mL). For Rho activation, cells were plated in the presence of Rho activator II at 1 μg/mL and treated for 24 hours. Rho activator was reapplied 4 hours before cell fixation or live imaging analysis.

4.2 |. cDNAs, adenoviruses and virus transduction

The adenoviruses expressing DPPIV-CAD-GFP, VSVG-ts045-GFP, VSVG-ts045-mTagBFP and VSVG-ts045-SNAP were reported previously.¹⁹ For the DPPIV-Dendra2 construct, rat liver DPPIV (kindly provided by O. Weisz, University of Pittsburgh Medical School, Pittsburgh⁶) was cloned via BamHI-HindIII sites into the pEGFP-C3 vector; the GFP-moiety was subsequently swapped for Dendra2. The ECFP/Citrine-YFP RhoA biosensor cDNA was generated and kindly provided to us by Louis Hodgson (Albert Einstein College of Medicine, New York). It is a synonymous codon-optimized version⁴¹ of his single-chain biosensor.²⁸ These cDNAs were utilized to generate adenoviruses with the AdEasy system⁴²; viruses were amplified and propagated according to reference 43. Cells expressing VSVG-ts045-GFP and co-expressing DPPIV-CAD-GFP and VSVG-ts045-mTagBFP or VSVG-ts045-SNAP were obtained by adenovirus infection in Opti-MEM (51985034; Gibco) for 1 hour with 1 to 10 plaque-forming units/cell and 10 to 12 hours expression at 40°C in 24 hours freshly seeded cells. Infection with the recombinant adenoviruses expressing the RhoA-CFP-YFP biosensor or the DPPIV-CAD-GFP cargo at 37°C were carried out 10 to 12 hours or 24 hours (DPPIV-Dendra2) prior to the experiment.

4.3 |. Antibodies and reagents

The following antibodies were used in this study: GM130 (mouse; 610822; BD Transduction Laboratories), DPPIV (CD26, clone OX-61; mouse; MCA924R: Bio-Rad), ezrin (rabbit; ab41672, Abcam), P-MLC2 (Phospho-Myosin Light Chain 2; rabbit; 3674; Cell Signaling), PKC ζ (sc-216; Santa Cruz Biotechnology, Inc.), TGN38 (provided by D. Shields, Albert Einstein College of Medicine, New York), RhoA (rabbit; ab86297; Abcam), VSVG cytoplasmic domain (clone P4D5; hybridoma provided by E. Rodriguez-Boulan, Weill Cornell Medical College, New York), and ZO-1 (rat; MABT11; EMD Millipore). Rho activator II (CN03) was from Cytoskeleton (Denver, Colorado). Cycloheximide was from MP Biomedicals (Santa Ana, California) and AP21988 was from ARIAD Pharmaceuticals (Cambridge, Massachusetts). Phalloidin-CF405 and -Atto 647 N were purchased from Biotium (Fremont, California) and Sigma (St. Louis, Missouri), respectively. SNAP-Cell TMR-Star (TMR*) substrate was from New England BioLabs (Ipswich, Massachusetts).

4.4 |. Immunofluorescence and antibody uptake

For immunofluorescence experiments, cells were fixed in 4% paraformaldehyde in phosphate-buffered saline (PBS) buffer for 20 minutes at room temperature or 4°C (FRET experiments), permeabilized with 0.1% saponin or 0.2% TX-100 in PBS for 10 to 15 minutes and blocked in 1% bovine serum albumin (BSA) before primary antibodies incubation. F(ab′)₂ fragment secondary antibodies coupled to CF405M, Alexa Fluor 488, 647 and Rhodamine Red-X were used. For surface immunofluorescence on MDCK cells, the DPPIV antibody was applied to both top and bottom sides of the transwell membrane for 45 minutes at 4°C to 10°C in bicarbonate-free DMEM containing 20 mM HEPES before fixation. For antibody uptake/transcytosis in WIF-B9 cells (Figures 1B, 5A,D and S3), the DPPIV antibody was applied after 1 hour of DPPIV-CAD-GFP exit from the TGN at to the bottom side of the transwell membrane, then after gently washes, cells were incubated at 37°C for 1 hour. At each time, the cells were fixed, and the localization of the DPPIV antibody was detected with fluorescent anti-mouse secondary antibodies.

4.5 |. Confocal, SIM super-resolution and TIRF microscopy

Confocal microscopy was performed in a TCS SP5 confocal microscope (Leica Microsystems Heidelberg, Manheim, Germany) equipped with 405 nm UV and 488 nm, 543 and 633 nm visible laser lines. Alignment between UV and visible wavelengths were routinely tested. Cells were imaged using an HCX PL APO lambda blue 63× 1.40NA or HCX PL APO CS 40× 1.25NA oil UV objectives. Fixed cells were imaged on glass cover slips mounted in Mowiol or non-hardening, glycerol-based, aqueous mounting medium. Confocal (pinhole, 1 Airy Unit; pixel size, 63.1-137.6 nm, scan speed 1000 Hz) x-y or x-z single sections and x-y-z stacks (z-step size, 1 μm) were taken. Live cell imaging was performed on MatTek chambers at 37°C in a CO₂ enriched atmosphere in growth medium. x-y-t stacks (pinhole, 1 Airy Unit; pixel size, 80.2-160.5 nm; sequential scan between lines, scan speed 8000 Hz) were automatically acquired at 2 seconds per frame rate. For cells expressing DPPIV-Dendra2 x-y-t and x-z-t stacks, were manually acquired at the indicated times. Selected sections from the x-y-z or x-y-t stacks were projected (x-y_p, or x-y-t_p) as indicated in the corresponding panels. Images were collected with LAS AF v.2.6.0.7266 (Leica Microsystems) and analyzed with ImageJ v.1.52q (National Institutes of Health) software. 3D-SIM was performed in a Nikon ECLIPSE Ti-E microscope (Nikon Instruments Inc., Melville, New York) equipped with a deep-cooled, back-illuminated EMCCD Andor iXon3 DU897 camera using 1×1 binning (calibrated pixel size, 32 nm), and an SR Apo TIRF 100× 1.49NA oil objective. Reconstructed images were obtained from x-y-z stacks (z-step size, 0.1 μm). Samples were immunostained with an anti-VSVG antibody to enhance the VSVG-ts045-mTagBFP signal. Images were collected with Nikon NIS-elements software. TIRF microscopy was performed in a modified Olympus IX71 microscope described in Dovas et al.⁴⁴ The objective used was a 150× 1.45NA TIRF. Cells were imaged on MatTek chambers at 37°C in a humidified atmosphere in a growth medium containing 10 mM HEPES. GFP and TMR* were excited with the 488 and 561 nm laser lines, respectively, and time-lapse sequences acquisition was at the rate of 3.3 frames/s. Images were collected with MetaMorph software and processed with Image J.

4.6 |. FRET analysis and photoconversion

FRET of RhoA ECFP/citrine-YFP was measured in fixed cells. FRET efficiency was calculated with the acceptor-bleaching FRET module (FRET-AB) of the LAS AF software. For the bleaching, a region of interest at the TGN was exposed to the 488 nm argon laser (main power; 80% with 514 nm laser line; 100%, for 10 iterations). The FRET efficiency was evaluated only in regions of interest with at least 75% of reduced fluorescence intensity in the acceptor species. For photoconversion experiments, a Golgi area in live cells expressing DPPIV-Dendra2 was exposed to the 405 nm diode laser (power; 90%, for 2 to 4 seconds). Photoconversion occurred, as demonstrated by the decrease in green fluorescence (507 emission peak) and the simultaneous increase in red fluorescence (573 nm emission peak).

4.7 |. Quantification and statistical analysis

The colocalization analysis was performed with Leica LAS AF software adjusting the threshold and background parameters for channels 1 and 2 to 70%. The values were normalized to the highest colocalization rate. The percentage of TGN exit events of tubular or vesicular carriers (~300 nm minimum diameter) that move away from the Golgi area (≥2 μm) was obtained by the analysis of individual and merged channels of time-lapse sequences. The mean intensity and perimeter measurements were obtained with the ImageJ tools. For statistical computation and estimation of significance, we used GraphPad Prism version 7.0 software. Graphs represent means + or − SD, where (n) represents the number of cells analyzed. Statistical significance was determined using unpaired two-tailed t-test for three independent experiments.