Misplaced formation of microvilli to basolateral domains and intracellular inclusions in enterocytes are pathognomonic features in congenital enteropathy associated with mutation of the apical plasma membrane receptor syntaxin 3 (STX3). Although the demonstrated binding of Myo5b to the Rab8a and Rab11a small GTPases in vitro implicates cytoskeleton-dependent membrane sorting, the mechanisms underlying the microvillar location defect remain unclear. By selective or combinatory disruption of Rab8a and Rab11a membrane traffic in vivo, we demonstrate that transport of distinct cargo to the apical brush border rely on either individual or both Rab regulators, whereas certain basolateral cargos are redundantly transported by both factors. Enterocyte-specific Rab8a and Rab11a double-knockout mouse neonates showed immediate postnatal lethality and more severe enteropathy than single knockouts, with extensive formation of microvilli along basolateral surfaces. Notably, following an inducible Rab11a deletion from neonatal enterocytes, basolateral microvilli were induced within 3 days. These data identify a potentially important and distinct mechanism for a characteristic microvillus defect exhibited by enterocytes of patients with neonatal enteropathy.
Polarization of intestinal epithelial cells (IECs) is fundamental for nutrient absorption in all animals (Bryant and Mostov, 2008). Mature IECs are joined by junctional complexes that result in formation of distinct apical and basolateral domains, each containing a distinct composition of proteins and lipids (Apodaca et al., 2012; Overeem et al., 2015; Rodriguez-Boulan and Macara, 2014). In enterocytes, specialized transporters, ion channels and digestive enzymes are delivered to apically positioned brush border microvilli to facilitate digestion and absorption (Field, 2003; Martin-Belmonte et al., 2007; Repishti et al., 2001). Proper transport and insertion of these proteins involves unidirectional anterograde vesicle traffic along cytoskeletal components and vesicle fusion that relies upon evolutionarily conserved motor proteins, RAB small GTPases, and membrane fusion machinery (Donovan and Bretscher, 2012; Gálvez-Santisteban et al., 2012; Huber et al., 1993b; Li et al., 2002; Low et al., 1996; Roland et al., 2011). Utilizing similar transport processes, functionally distinct basolateral proteins, such as the extracellular matrix receptor integrins, are localized to basolateral membrane surfaces where they serve to attach IECs to the underlying basement membrane (Aumailley et al., 2000; Larjava et al., 1993). The mechanism underlying how apical membrane components and structural features are sorted and excluded from basolateral domains remains poorly understood.
Aberrant apical transport has been found to impair microvillus structure and function, and adversely impact nutrient uptake, causing neonatal enteropathy such as microvillus inclusion disease (MVID). MVID causes intractable diarrhea in neonates, associated with life-threatening dehydration and metabolic disorders (Cutz et al., 1989; Davidson et al., 1978; Ruemmele et al., 2006). Intestines of MVID patients show villus atrophy, loss of microvilli, abnormal accumulation of periodic acid–Schiff (PAS)-positive intracellular vesicles, as well as appearance of characteristic microvillus inclusion bodies in affected enterocytes (Ameen and Salas, 2000; Cutz et al., 1989; Phillips et al., 2000). This disease might have different clinical manifestations, with severe cases affecting neonates in the first day of life, and mild ones exhibiting a delayed disease onset, e.g. in the first 2 months after birth (Cutz et al., 1989; Davidson et al., 1978; Ruemmele et al., 2006). MVID patients have been identified to possess mutations in myosin VB (MYO5B) (Müller et al., 2008; Ruemmele et al., 2010; Szperl et al., 2011; Thoeni et al., 2014; van der Velde et al., 2013). Mouse genetic studies have further added to the understanding of Myo5b motor trafficking in regulating apical transport, formation of microvillus inclusions and associated MVID (Cartón-García et al., 2015; Schneeberger et al., 2015; Weis et al., 2016). Recently, a truncation mutation in an apical t-SNARE gene, syntaxin 3 (STX3), was described to account for a variant form of enteropathy with late disease onset (Wiegerinck et al., 2014) and several features different from typical MVIDs (Knowles et al., 2014b).
Rowland et al. established that MYO5B binding to RAB8A, RAB10 and RAB11A (Roland et al., 2007, 2009) allows the motor to access and transport associated vesicles toward the plasma membrane (Roland et al., 2011). In MYO5B knockdown CaCo2 cells, expression of mutant MYO5B that was uncoupled from either RAB8A or RAB11A resulted in different subcellular changes: apical microvillus deficiency or induction of microvillus inclusion, respectively (Knowles et al., 2014a). Likewise, in MYO5B null CaCo2 cells, which were genetically modified to carry a patient-derived MYO5B mutant (1125G>A), expression of a RAB11A binding-deficient MYO5B showed reduced complex formations with STX3 and the exocyst proteins. However, less effect on binding to STX3 or exocyst proteins was observed in the expression of a RAB8A binding-deficient MYO5B (Vogel et al., 2015).
Although currently no RAB8A or CDC42 mutation has been found in MVID patients, mice with Rab8a or Cdc42 deficiency in IECs showed microvillus atrophy and formation of microvillus inclusion bodies resembling lesions in MVID (Melendez et al., 2013; Sakamori et al., 2012; Sato et al., 2007). In contrast to the early lethality of Myo5b knockout mice, Rab8a knockout mice survived up to weaning stage (Sato et al., 2007), whereas the majority of IEC-specific Cdc42-knockout mice survived to adulthood (Sakamori et al., 2012), suggesting that the appearance of typical microvillus inclusions was not directly correlated with clinical severity. Rab8a and Rab8b double-knockout mice died several days earlier than Rab8a single-knockout mice (Sato et al., 2014). Recently, we identified that global Rab11a ablation in mice results in embryonic lethality (Yu et al., 2014b), while constitutive IEC-specific Rab11a deletion induces microvillus deficiency, accompanied by mislocalization of apical Stx3 (Knowles et al., 2015). The large majority of Rab11afl/fl;Villin-Cre knockout mice survived beyond weaning stage, with a substantial number of female knockouts living to adulthood (Yu et al., 2014a). An independent report of IEC-specific Rab11a knockout mice on a different genetic background showed a relatively earlier postnatal lethality (Sobajima et al., 2015).
Aberrant formation of lateral microvilli was reported in patients with STX3 mutation (Wiegerinck et al., 2014). In contrast, cytoplasmic microvillus inclusions were formed in either Rab8a- or Cdc42-deficient IECs without development of lateral microvilli (Melendez et al., 2013; Sakamori et al., 2012; Sato et al., 2007). Of note, loss of Myo5b also perturbed basolateral integrin α2 localization in Myo5b-deficient adult mouse IECs (Schneeberger et al., 2015). However, no significant basolateral trafficking defect was observed in either Rab8a- or Rab11a-deficient IECs (Knowles et al., 2015; Sato et al., 2007; Sobajima et al., 2015), although each one of them was shown to control basolateral trafficking in cultured cells (Ang et al., 2003; Henry and Sheff, 2008; Huber et al., 1993a,b; Lock and Stow, 2005). Combining the results of multiple Myo5b, Rab8a and Rab11a studies gives rise to the overall impression that the severity of Myo5b loss-of-function phenotypes is directly linked to selective or simultaneous incapacitation of the Rab8a/11a cargo trafficking pathway.
In this study, we assessed disease severity and determined the regulatory pathway for apical-basolateral cargo delivery in neonatal mice with individual or compound disruption of Rab8a and Rab11a traffic. We demonstrate that combinatory loss of both trafficking pathways caused severe microvillus atrophy and extensive formation of basolateral microvilli along the entire length of the small intestine. Although proper apical localization of brush border structural proteins might be carried out by a single Rab8a or Rab11a factor, transport of brush border enzymes and ion transporters required presence of both Rabs. Our data suggested that certain basolateral proteins were redundantly trafficked by Rab8a and Rab11a. Importantly, inducible deletion of Rab11a (by Villin-CreER) from IECs in neonates induced exclusive basolateral microvillus-like formation but not microvillus inclusions. These data improve our understanding of a particular cellular pathology manifested in neonatal enteropathy.
Loss of Rab8a and Rab11a results in greater severity of enteropathy compared with single gene knockout
Prior studies suggested that Rab8a or Rab11a function in distinct and overlapping pathways during apical development and cargo delivery in enterocytes (Knowles et al., 2014a). In order to identify functional redundancies and elucidate the specific contributions of each factor to transport of specific cargos during brush border morphogenesis, we performed genetic studies utilizing Rab8a and Rab11a conditional mouse alleles. Rab8afl/+;Rab11afl/+;Villin-Cre (double-heterozygous) mice showed no overt difference from wild-type (no Cre) littermates (Fig. 1A). By crossing Rab8afl/+;Rab11afl/+;Villin-Cre mice to Rab8afl/fl;Rab11afl/fl mice, we obtained double-knockout (DKOΔIEC) Rab8afl/fl;Rab11afl/fl;Villin-Cre pups, along with pups containing single allele knockouts of either gene (Rab8afl/fl;Rab11afl/+;Villin-Cre or Rab8afl/+;Rab11afl/fl;Villin-Cre). DKOΔIEC mice were born alive but died within the first 24 h (Fig. 1A; Table S1). Pups that retained a single Rab11a allele in IECs (orange line, Fig. 1A) survived a few days longer than those with a single Rab8a allele (green line, Fig. 1A), with median survival times of 24 days and 15 days, respectively (Fig. 1A). The poor postnatal survival rates of DKOΔIEC and Rab8afl/+;Rab11afl/fl;Villin-Cre mice were correlated with their lower birth weights (Fig. 1B), smaller body sizes (Fig. 1C), and signs of dehydration. Compared with wild-type tissues, the ileum, cecum and colon of DKOΔIEC mice appeared transparent. Histologically, DKOΔIEC and Rab8afl/+;Rab11afl/fl;Villin-Cre mice showed blunted villi, with the DKOΔIEC demonstrating the greatest severity of villus atrophy, blunting, and fusion (Fig. 1D,E, compare panels left to right). The distal intestinal segments of DKOΔIEC mice appeared to be more affected than the proximal parts. PAS staining showed that the enterocytes of DKOΔIEC and Rab8afl/+;Rab11afl/fl;Villin-Cre mice contained expanded PAS-positive apical staining that extended into the cytoplasm (Fig. 1E).
DKOΔIEC enterocytes develop extensive basolateral microvilli
To further analyze the villus morphology in DKOΔIEC mice, tissue from P1 duodenal tissues were prepared and examined by scanning electron microscopy (SEM). Comparison of duodenal tissue clearly demonstrates the gross difference in villus structure with DKOΔIEC villi being shorter and ‘bloated’, with extensive surface patchiness and irregularity, as compared to villi from wild-type mice (Fig. 1F,G). When examined at higher magnification, the upper regions of DKOΔIEC villi appeared ‘bald’, and the enterocytes were effectively devoid of surface microvilli (Fig. 1H,I). The loss was less pronounced on cells located along the lower portions of villi (Fig. 1J,K); however, the microvilli that were present were no longer hexagonally packed (Fig. 1K), and appeared to be shorter, with occasional blebbing and fusion with adjacent microvilli (Fig. 1K).
The observed structural abnormality of villi by SEM was further confirmed when tissue was examined at the ultrastructural level by transmission electron microscopy (TEM). Enterocytes from DKOΔIEC mice exhibited significant alterations in brush border architecture. Such alterations included cells lacking an organized brush border (Fig. 2A,B), cells possessing short blunt microvilli (Fig. 2; Fig. S1A, bar graph), and cells having abnormally long apical rootlets (duodenum in Fig. 2; ileum in Fig. S1). In addition to malformation of the apical brush border, there was also abnormal structural organization to the lateral surfaces of enterocytes, typically characterized by well-formed apical adherens junctions and well-aligned opposing plasma membranes (Fig. 2A,C; Fig. S1A). Enterocytes in DKOΔIEC possessed extensive stretches of lateral microvilli containing core bundles of actin filaments (Fig. 2; Fig. S1). Lateral microvilli formed below the level of the tight junction (Fig. 2B; Fig. S1A) and were often clustered adjacent to lateral microvilli protruding from the neighboring cells (Fig. 2A,B; Fig. S1A,B). Qualitatively, lateral microvilli appeared to be flaccid and wavy as if not possessing the full complement of actin-binding proteins present in apical microvilli from wild-type tissue. Although examination of tissue from multiple animals failed to detect typical microvillus inclusion bodies as previously reported for Myo5b-, Rab8a- or Cdc42-deficient enterocytes (Cartón-García et al., 2015; Sakamori et al., 2012; Sato et al., 2007; Schneeberger et al., 2015; Weis et al., 2016), we did observe multiple examples of enclosed patches of microvilli located in the apical cytoplasm beneath the brush border (arrows in Fig. 2A; box in Fig. S1C). Apical microvillus deficiency and formation of lateral microvilli were also detected in embryonic day 18.5 DKOΔIEC intestines (n=2) (Fig. S1D).
DKOΔIEC enterocytes showed additional structural abnormalities, including the presence of large cytoplasmic vesicles or vacuoles (>1 µm in diameter) (asterisk in Fig. 2B; Table S2), and multiple smaller non-electron-dense vesicles. In a number of cases, the entire cytoplasm underlying the extended actin core rootlets consisted of a collection of clear vesicles and tubules. Some of the cytoplasmic inclusions contained individual finger-like projections (Fig. 2C), similar to vesicle bodies described in the enterocytes of MVID patients (Cutz et al., 1989).
In addition to the aforementioned abnormalities in the proximal (duodenum) portions of the intestine, we also examined the ilia of these animals. Similar to DKOΔIEC duodenum, ileal enterocytes exhibited severe microvillus atrophy, pronounced lateral microvilli formation, drastic expansion of the terminal web, and excessive formation of membranous vesicles/tubules (Fig. S1A,B, Table S3). The microvilli of some DKOΔIEC enterocytes located in the distal ileum were observed to fuse with each other (arrow in Fig. S1C). Approximately one in three DKOΔIEC ileal enterocytes exhibited the microvillus defects described above, suggesting a more pronounced structural abnormality as compared to duodenal tissue, where one in 10 enterocytes exhibited evident defects.
Inducible deletion of Rab11a in neonatal IECs results in abnormal microvillus formation
Animals and/or cells deficient in Rab8a (Sato et al., 2007), Rab11a (Knowles et al., 2015) or Myo5b (Weis et al., 2016) all exhibit defects in assembly of the characteristic apical brush border architecture. The defects range from inappropriate membrane sorting to formation of microvillus inclusions (Sato et al., 2007; Knowles et al., 2015; Weis et al., 2016). Additionally, there is evidence that the degree of abnormality might be correlated with a developmental time window around the neonatal stage (Weis et al., 2016). To investigate the impact of developmental timing, we established Rab11afl/fl;Villin-CreER mice and induced the deletion of Rab11a in 3-day-old neonates by tamoxifen administration. Tamoxifen-injected Rab11afl/fl;Villin-CreER mice, but not their wild-type littermates, developed extensive lateral microvilli, (Fig. 3A,C), possessed significantly stunted or shortened microvilli with expanded terminal web (Fig. 3A,B), exhibited abnormal lateral cell-cell contacts leading to large gaps/vacuoles between adjacent cells (asterisk in Fig. 3C), and appeared to have individual microvilli that were larger in diameter than those in control samples (P<0.001) (Fig. 3D). Despite the severity of the phenotype and disruption of normal microvillar assembly and positioning, there was no detectable structure that would be characterized as microvillus inclusions. More typically, there were sizeable ‘pockets’ of microvilli located laterally between cells and (arrows in Fig. 3A) as well as pockets penetrating into the apical cytoplasm (areas outlined by dotted lines in Fig. 3C). Additionally, in Rab11afl/fl;Villin-CreER enterocytes, the cytoplasm immediately below the terminal web was filled with clear non-electron-dense vesicles and tubular networks (arrows in Fig. 3B). Consistent with a previous study (Sato et al., 2007), neonatal Rab8afl/fl;Villin-Cre enterocytes (P2) did not exhibit strong microvillus defects but showed apical accumulation of vesicles (Fig. S2A); Rab8afl/fl;Villin-Cre mice of weaning age demonstrated severe microvillus atrophy and cytoplasmic inclusions of lysosomes or autophagosomes (Fig. S2). However, no lateral microvilli were observed in Rab8afl/fl;Villin-Cre enterocytes at these stages.
Rab8a and Rab11a redundantly contribute to apical localization of villin and ezrin
Given the observed effects that both Rab8a and Rab11a had on assembly of the brush border and the presence of numerous cytoplasmic vesicles/tubular extensions, we examined whether apical localization of microvillus structural components, villin and ezrin, were differentially dependent on Rab8a or Rab11a. Villin, a primary component of the actin bundle core rootlets (Drenckhahn et al., 1983), remained largely at apical domains in Rab8afl/fl;Rab11afl/+;Villin-Cre tissue, with an overall localization very similar to tissue from wild-type animals (Fig. 4A, also see bar graphs). Upon deletion of Rab11a and retention of a single copy of Rab8a (Rab8afl/+;Rab11afl/fl;Villin-Cre enterocytes), there resulted a pronounced decrease in confined apical localization of villin and an observed expansion into the subapical cytoplasmic regions (Fig. 4A). Occasional villin was detected at the lateral junction in Rab8afl/fl;Rab11afl/+;Villin-Cre enterocytes (Fig. 4A, arrow); in Rab8afl/+;Rab11afl/fl;Villin-Cre enterocytes, villin exhibited a pronounced punctate staining pattern (Fig. 4A, arrow). DKOΔIEC mouse enterocytes showed decreased apical staining for villin; instead, villin appeared localized to the subapical and non-apical regions of enterocytes (Fig. 4A). Ezrin, another key partner in construction of microvilli (Tsukita and Yonemura, 1999), also demonstrated localization dependence on Rab expression. In DKOΔIEC enterocytes, ezrin localization essentially mirrored villin staining, with a decrease in apical brush border staining and an accompanying increase in non-apical domains (Fig. 4B, bar graph). Single knockouts exhibited less-pronounced changes as compared to villin. Ezrin remained primarily apical; however, there was a detectable (20%) repositioning to lateral domains in Rab8afl/fl;Rab11afl/+;Villin-Cre enterocytes that appeared to be dependent upon where the cell was positioned along the villus (Fig. 4B; Table S4, Fig. S3A). In Rab8afl/+;Rab11afl/fl;Villin-Cre enterocytes, ezrin localization was nearly identical to that in wild-type tissue, with possible broadening of the brush border domain.
Additionally, we used phospho-ezrin–radixin–moesin (p-ERM) immunostaining as a read-out of functional crosslinking activity between actin filaments and the plasma membrane (Gautreau et al., 2000; Matsui et al., 1998). Wild-type and Rab-knockout animals all exhibited apical p-ERM staining, with a ∼20% decrease in the number of positive cells for DKOΔIEC and Rab8afl/+;Rab11afl/fl;Villin-Cre enterocytes (Fig. 4C, also see bar graph). Although not quantitated, the levels of fluorescent intensity appeared to be somewhat lower in the knockout animals, and this would be consistent with the shortening of microvilli observed by TEM (see above). Also consistent with the reported TEM observations, we detected p-ERM, villin, and ezrin staining within spherical pockets located along lateral surfaces (arrows in Fig. 4A-C). Collectively, the results suggest that villin and ezrin are reliant upon Rab8a and Rab11a for apical membrane localization, and that there could be functional overlap in the delivery process.
Localization of Mst4, syndapin 2, and PKCζ in double-knockout enterocytes
To further explore the differences observed for ezrin and p-ERM localization, we analyzed the intracellular distribution of Mst4 (also known as Stk26), the kinase that phosphorylates ezrin in enterocytes (Baas et al., 2004; ten Klooster et al., 2009). Mst4 was primarily localized to apical membranes and subapical regions in wild-type enterocytes (Fig. 5A, also see bar graph). Rab8afl/fl;Rab11afl/+;Villin-Cre enterocytes retained a primary, but reduced, apical localization with an increased presence at perinuclear and basal cytoplasmic regions (Fig. 5A). In contrast, Rab8afl/+;Rab11afl/fl;Villin-Cre and DKOΔIEC enterocytes no longer possessed a primary apical distribution, and instead the overall localization was diffused throughout the entire cell with very low levels of apical enrichment (Fig. 5A), consistent with the notion that Rab11a depletion prevents subapical enrichment of kinases, thereby inhibiting ezrin phosphorylation (Dhekne et al., 2014). Analyses across multiple sections and animal tissues did not detect any Mst4 staining in the lateral pockets, unlike observed for villin, ezrin and p-ERM (compare Fig. 5A with Fig. 4).
Previously, it was shown that the multifunctional actin regulator cordon bleu (COBL) promotes the assembly of actin core bundles of brush border microvilli (Grega-Larson et al., 2015), and that COBL activity is regulated by the upstream regulator syndapin 2. In particular, the COBL–syndapin 2 complex has been proposed to tether the membrane base of microvilli to the actin core bundle, which in turn promotes actin polymerization from microvillar tips (Grega-Larson et al., 2015). Anti-syndapin 2 staining resulted in distinct tight staining of the apical surface in wild-type tissue, and the apical enrichment was greatly diminished in single-knockout mice and completely lost in double knockouts (Fig. 5B). Rab8afl/fl;Rab11afl/+;Villin-Cre enterocytes retained an enrichment at the apical surface along with a significant increase in cytoplasmic staining and possible perinuclear enhancement (Fig. 5B, arrow). In both Rab11a-deficient and DKOΔIEC enterocytes, syndapin 2 was dispersed throughout the cytoplasm, and apical enrichment was a rare occurrence (Fig. 5B).
We also examined the localization of Stx3, which is known to regulate membrane targeting to apical surfaces of epithelial cells (Vogel et al., 2015). Stx3 relocalizes from being primarily apical in wild-type enterocytes to being cytoplasmically dispersed in DKOΔIEC (Fig. S3B); the same mislocalization of Stx3 was reported for Rab11a-deficient enterocytes from adult animals (Knowles et al., 2015). In addition to examining Mst4, syndapin 2 and Stx3, which are known regulatory elements for brush border assembly, we also examined the localization of PKCζ (also known as Prkcz), a key component of the apical PAR complex that determines apical polarity (Gao and Kaestner, 2010). Apical membrane localization of PKCζ was significantly diminished in Rab8afl/fl;Rab11afl/+;Villin-Cre and DKOΔIEC enterocytes, but was minimally changed or unchanged in Rab11a knockouts (Fig. 5C, bar graphs).
Positioning of brush border enzymes and ion exchangers in Rab8a and Rab11a mutant tissue
Severe clinical outcome of congenital enteropathy is often related largely to disruption of nutrient digestion and uptake often resulting from the inability to properly construct apical microvillar cytoarchitecture (Overeem et al., 2016). Given our observations on survivorship (Fig. 1), malformation of the apical brush border (Figs 2 and 3; Fig. S1), and massive accumulation of mis-sorted/targeted membrane compartments (Figs 4 and 5), we determined the impact of Rab mutations on localization of key brush border enzymes or ion exchangers that are essential for adsorption and survival. Wild-type enterocytes displayed alkaline phosphatase (AP) and sucrase isomaltase (SI) immunofluorescence as a narrow and sharp band at the apical brush border (P1 duodenum in Fig. 6A, P21 duodenum in Fig. 6B). Upon single or double mutation of Rab8a/11a, both AP and SI became mislocalized from the apical brush border and located subapically in a broad cytoplasmic space between the apical surface and the nucleus (Fig. 6A,B). This localization appears to closely correlate with the expanded subapical membrane compartments (vesicles and tubular extensions) observed by TEM. Similar mislocalization was observed when we analyzed for sodium-hydrogen exchanger 3 (NHE3, also known as SLC9A3), an important Na+/H+ exchanger (Fig. 6C). Additionally, DKOΔIEC tissue stained for Lamp2 and LC3 identified a marked increase in subapical staining for both, suggesting that the double knockout leads to expansion of lysosomes and autophagosomes in the subapical domain (Fig. S3C,D). These data collectively suggested that the proper transport of brush border enzymes and ion exchangers were critically dependent on both Rab8a and Rab11a.
Analysis of basolateral cargo traffic in Rab8a- and Rab11a-deficient enterocytes
To address the impact of Rab8a and Rab11a mutations on basolateral trafficking, tissues were stained for E-cadherin, laminin, Na+/K+ ATPase, and integrin β1. Across all tissues examined, neither single- nor double-knockout enterocytes appeared to demonstrate mis-trafficking of either E-cadherin or laminin (Fig. 7A); occasionally the two were colocalized at the basal-most surface of DKOΔIEC enterocytes (Fig. 7A, arrows). The overall proper localization of E-cadherin to lateral domains was consistent with TEM observations, showing that all tissue types were capable of making adherens junctions (Figs 2 and 3; Fig. S1). Basolateral localization of Na+/K+ ATPases in Rab8afl/+;Rab11afl/fl;Villin-Cre and Rab8afl/fl;Rab11afl/+;Villin-Cre enterocytes appeared indistinguishable from wild-type tissue with the presence of pronounced staining at cell edges and additional diffuse, possible vesicular, cytoplasmic staining (Fig. 7B). DKOΔIEC enterocytes possessed diffuse cytoplasmic staining with reduced lateral cell edge staining regardless of position along the villus (Fig. 7B; Table S4). Integrin β1 localization appeared unchanged in Rab8afl/fl;Rab11afl/+;Villin-Cre tissues, but diffuse in Rab8afl/+;Rab11afl/fl;Villin-Cre and DKOΔIEC tissues (Fig. 7C), reflecting a reduced protein level detected in DKOΔIEC tissues by western blotting (Fig. 7D) , as well as previously reported regulation of integrin β1 trafficking by Rab25, a Rab11 subfamily member (Krishnan et al., 2013; Nam et al., 2010).
Rab8a and/or Rab11a deficiency lead to different morphogenesis of neonatal intestinal organoids
The intestinal organoid culture platform can be used to follow morphogenesis of an enteroid from either an isolated single crypt or a single stem cell (Sato et al., 2009). Organoid cultures from small intestines of P1 mice of various mutant genotypes were prepared using a protocol modified from an established method (Sato et al., 2009). P1 DKOΔIEC intestines showed more proliferative cells in vivo (Fig. S4A). The morphogenesis of organoids and number of surviving organoids were recorded daily for each genotype group (Fig. S4B,C). Almost all surviving organoids appeared as spheres at day 1; however, their morphologies became evidently different around day 3–4, when wild-type organoids showed prominent epithelial budding, whereas Rab8afl/fl;Rab11afl/+;Villin-Cre organoids developed spheres with small buds. By contrast, ∼50% of Rab8afl/+;Rab11afl/fl;Villin-Cre and DKOΔIEC organoids became completely sphere-like, with <15% developing more than two small buds (Fig. S4D). More than half of the wild-type organoids became mature epithelial structures that contained actively proliferative buds labeled by 5-ethynyl-2′-deoxyuridine (EdU) (Fig. S4B,D). Collectively, these results indicated that loss of Rab11a promoted the neonatal enteroid to adopt a sphere-like morphology, while loss of Rab8a reduced budding activities.
Prior biochemical and cell biology studies established that proper membrane and protein trafficking via Myo5b- and Rab8a/11a-dependent pathways were integral to proper assembly of the apical cytoarchitecture of IECs (Knowles et al., 2014a; Roland et al., 2011). Further, Myo5b binding to Rab8a or Rab11a is managed by distinct structural motifs and protein complexes (Roland et al., 2011). In CaCo2 cell cultures, specifically uncoupling of Myo5b from either Rab8a or Rab11a resulted in different structural defects in microvillus assembly and positioning (Knowles et al., 2014a; Vogel et al., 2015). Rab8a- or Rab11a-deficient mouse IECs demonstrated microvillus defects, yet the disease manifestations of these mice were less severe than those of Myo5b-deficient mice (Cartón-García et al., 2015; Knowles et al., 2015; Sato et al., 2007; Schneeberger et al., 2015; Sobajima et al., 2015; Weis et al., 2016). The CaCo2 model cannot recapitulate the disease progression of neonatal enteropathy; therefore, it was necessary to examine the in vivo role of Rab8a and Rab11a in specific cargo delivery during neonatal microvillus morphogenesis. By performing extensive mouse genetic experiments, we evaluated the impact of individual or combinatory disruptions of Rab8a and Rab11a traffic on microvillus morphogenesis and on distinct cargo transportations in neonatal enterocytes (Fig. 7E).
Our data suggest that selective or combinatory loss of Rab8a or Rab11a traffic could cause neonatal enteropathy with different severities and disease onsets. The most severe clinical presentation and lethality associated with MYO5B mutations are likely attributable to an accumulative loss of Rab8a- and Rab11a-mediated vesicular traffic. Germ-line Myo5b-knockout mice display perinatal lethality, and died within 12 h or several days after birth (Cartón-García et al., 2015; Weis et al., 2016). This lethality resembled the phenotype observed in mice with simultaneous loss of Rab8a and Rab11a in IECs. The reduced body weight, dehydration and early death of these mice represented severe forms of neonatal enteropathy (Cutz et al., 1989; Davidson et al., 1978). In terms of survivability, IEC-specific deletion of either Rab8a or Rab11a resulted in relatively mild disease outcomes (Sato et al., 2007; Yu et al., 2014a), consistent with the fact that a certain portion of normal intracellular traffic was preserved. Extending from previous studies, we demonstrated that approximately half of the animals that had only a single wild-type Rab8a allele in IECs survived up to weaning stage, and a large majority of mice that retained a single wild-type Rab11a allele in IECs lived beyond weaning stage. The reported observations build a developmental model, in which Rab8a and Rab11a transport both common and distinct protein cargos and that specific cargo delivery has consequence on IEC physiological functionality and structural organization.
In contrast to IEC-specific Rab11a knockouts, Rab8afl/+;Rab11afl/fl;Villin-Cre mice that were deleted of one copy of Rab8a on a Rab11a-deficient genetic makeup exhibited a much more severe disease outcome. All Rab8afl/+;Rab11afl/fl;Villin-Cre pups died within 1 month, suggesting that in the absence of Rab11a, the gene dosage of Rab8a became critical for the overall survivability of the animals. Of note, loss of Rab8a or Rab11a in IECs reciprocally altered the intracellular distribution of the remaining Rab proteins (Sobajima et al., 2015), suggesting that Rab8a and Rab11a might compensate for the absence of one or the other. Consistent with this notion, DKOΔIEC exhibited the poorest disease outcome. These results collectively suggest that in normal IECs, distinct apical cargos might differentially rely on Rab8a or Rab11a for proper transport; however, the loss of a single Rab8a or Rab11a pathway can be partially compensated by the remaining Rab protein.
The aforementioned conclusion of overlapping functionality is supported by the presence of a single Rab8a or Rab11a allele being sufficient to sustain the proper apical transport of sufficient levels of actin cytoskeletal proteins (villin and ezrin in the present study) to construct the apical brush border (Crawley et al., 2014), albeit with distinct organizational defects. Ezrin is the only ERM protein in developing small intestinal epithelium and is required for proper apical membrane assembly (Casaletto et al., 2011; Seton-Rogers, 2013; Zhu et al., 2010). In DKOΔIEC enterocytes, ezrin was dispersed in the cytoplasm and/or basolateral membrane. In addition, p-ERM, which controls the brush border assembly, was found localized to cytoplasmic puncta and thus ill-positioned for ‘zippering’ the plasma membrane along actin core bundles (Bretscher et al., 1997; Nambiar et al., 2010; Tsukita et al., 1997). Its upstream kinase, Mst4, was also found diffusely in cytoplasm. In addition, the mislocalization of PKCζ, as well as the significant elevation of total PKCζ and LKB1 (also known as STK11), protein levels in double-knockout IECs indicated a profound cell polarity defect (Benton and St Johnston, 2003; Krahn et al., 2010; Morais-de-Sá et al., 2010; Sotillos et al., 2004; Walther and Pichaud, 2010). However, the molecular mechanism contributing to the elevation of these proteins in the absence of Rab8a or Rab11a warrants future investigation. Similarly, mislocalization of syndapin 2 away from the bases of microvilli may have been a contributing factor to the short stature of the microvilli in knockout tissue IECs (Grega-Larson et al., 2015). Also, while not examined in this study, it would be reasonable to propose that myosin-1a, a known actin core bundle membrane crosslinker (Tyska et al., 2005; Tyska and Nambiar, 2010), might be similarly mis-trafficked in mutant animals. Mislocalization of these crucial brush border molecules was likely the major contributor to the ultimate death of mutant animals that retained only a single Rab8a or Rab11a allele in IECs. In double-knockout animals, the consequences on structural organization were even more pronounced, resulting in villus fusion, microvillus atrophy and extensive formation of lateral microvilli. The villus fusion resembled the phenotype reported in ezrin-deficient (Saotome et al., 2004) and Myo5b-deficient intestines (Weis et al., 2016).
Additionally, tissues from Rab8a and Rab11a double-knockout animals were examined for apically localized membrane enzymes and transporters as well as basolateral resident transporters and adhesion proteins. Apical localization of SI and AP, and ion transporter NEH3 all appeared equally sensitive to knockout of either Rab protein and appeared to be re-sorted into subapical compartments. In contrast to apical transport, our analyses of Rab8a and Rab11a double-knockout mice revealed that these two Rab proteins might redundantly traffic certain basolateral proteins, including integrin β1 and Na+/K+-ATPase. Integrin and E-cadherin trafficking were shown in cultured cells to be regulated by a diverse protein network including kinases, sorting nexins and small GTPases of Arf and Rab families (Caswell et al., 2008; Krishnan et al., 2013; Wang et al., 2015; Weber et al., 2011; Woichansky et al., 2016). Upon mitogen stimulation, recycling of integrin was regulated by Arf6 and Rab11a in cultured cells (Caswell and Norman, 2006; Powelka et al., 2004), and integrin β1 basolateral localization was affected by Rab25 deficency (Nam et al., 2010). E-cadherin was trafficked via the recycling endosome and active Rab11 (Desclozeaux et al., 2008; Woichansky et al., 2016). In Drosophila embryos, reducing Rab11 function disrupted integrity of the ectoderm and led to loss of adherens junction (Roeth et al., 2009). Basolateral trafficking defects were described in Myo5b-knockout mouse IECs (Cartón-García et al., 2015; Schneeberger et al., 2015), but were rarely detected in Rab8a or Rab11a single-knockout mouse IECs (Knowles et al., 2015; Sato et al., 2007). Our data suggested that IECs retaining a single allele of Rab8a or Rab11a were still able to transport basolateral cargos, whereas IECs with complete loss of Rab8a or Rab11a showed somewhat severe impairment in integrin β1 and Na+/K+-ATPase localization. Of important note, E-cadherin involved in adherens junction assembly remained in its basolateral localizations. In addition, based on electron microscopic analyses, tight junctions appeared to be properly formed in the absence of both Rab8a and Rab11a. These data suggest that compound loss of Rab8a and Rab11a was not sufficient to perturb the formation of tight and adherent junctions. The establishment of these important junctional structures might be under even more redundant and stringent control owing to their fundamental importance to the epithelial integrity.
Our study establishes that the morphology of the apical brush border in enterocytes is differentially regulated by the Rab8a and Rab11a GTPases. Previously, it was suggested, and we confirmed here with multiple lines of evidence, that Rab11a deficiency causes formation of basolateral microvilli without formation of microvillus inclusions (Knowles et al., 2015). Rab8a-deficient IECs exhibited prevalent cytoplasmic inclusions (Sato et al., 2007) and similarly Cdc42-deficient IECs developed typical cytoplasmic microvillus inclusions (Melendez et al., 2013; Sakamori et al., 2012). An independent study of a villin-specific Rab11a deletion model reported the detection of microvillus inclusions in neonates (Sobajima et al., 2015); however, the frequency of inclusion formation was not reported. Thus, we anticipated detecting cytoplasmic inclusions when we inducibly deleted Rab11a from neonatal IECs. However, 3 days after inducible Rab11a deletion from Rab11afl/fl;Villin-CreER neonates, we only detected extensive formation of lateral microvilli by TEM. Our results suggest that loss of Rab11a traffic may be causal to the aberrant lateral microvilli formation.
Upon Rab11a loss there results abnormal trafficking of Stx3 to lateral membranes, where it might initiate the ectopic formation of lateral microvillus growth (Knowles et al., 2015). DKOΔIEC embryonic IECs started to develop lateral microvilli, suggesting that loss of Rab11a produces a microvillus morphogenetic defect that is independent of Rab8a. Consequently, formation of cytoplasmic microvillus inclusions in Rab8a-, Cdc42- or Myo5b-deficient IECs, all of which had wild-type Rab11a, suggested that formation of cytoplasmic microvillus inclusions might require Rab11a-directed trafficking. Previous studies showed that the actin and microtubule coordinate at the growing end of microtubule for cell remodeling (Coles and Bradke, 2015; Preciado Lopez et al., 2014), and that E-cadherin and Par-3 proteins are transported by Rab11 endosomes that are apically restricted in a microtubule dynein-dependent manner (Le Droguen et al., 2015). Whether the reported induction of basolateral microvilli and microvillus inclusions upon disassembly of microtubules (Achler et al., 1989; Gilbert and Rodriguez-Boulan, 1991) played a role in our model remains to be studied. Lateral microvilli have, so far, been infrequently detected in MYO5B-mutant patients. The effect of Myo5b loss appears to be much more penetrant than either Rab11a or Stx3 loss. This is likely due to the multiple Rab proteins that can interact with Myo5b. Thus, it would be expected that loss of Myo5b might affect trafficking by Rab11a, Rab11b, Rab25, Rab10 and Rab6a at least. The combination of these altered trafficking pathways may explain the differences in manifestation of lateral microvilli as seen with Rab11a loss and Stx3 mutation.
Overall, our data indicate that apical sorting of membrane, cytoskeletal and enzymatic components rely upon both overlapping and distinct Rab-dependent pathways. Further, there appear to be at least two distinct microvillus defects driven by different molecular mechanisms resulting in enteropathy.
MATERIALS AND METHODS
Rab8afl/fl, Rab11afl/fl, Villin-Cre and Villin-CreER mice have been described previously (Das et al., 2015; el Marjou et al., 2004; Madison et al., 2002; Sato et al., 2007; Yu et al., 2014a). The mice were bred and maintained in an AAALAC-accredited animal facility at Rutgers University, Newark, with all experimental procedures approved by the Institutional Animal Care and Use Committee (IACUC). Due to the early postnatal lethality of DKOΔIEC mice, newborn pups were weighed after birth and killed to collect intestinal tissues for fixation and histological analysis. To induce Rab11a deletion from neonatal Rab11afl/fl;Villin-CreER mice, 3-day-old pups were injected once with 0.5 mg per gram of body weight tamoxifen dissolved in corn oil, and killed 3 days after injection.
Immunofluorescence and immunohistochemistry
Immunostaining was performed as previously described (Das et al., 2015; Sakamori et al., 2012; Yu et al., 2014a). Briefly, the small intestinal tissues were fixed in 4% PFA at 4°C overnight, embedded in paraffin, and sectioned (5 μm). After rehydration and antigen retrieval (citrate acid buffer, pH 6.0), the slides were blocked in blocking buffer (PBS-T with 2% normal serum, 2% BSA) at room temperature for 1 h and incubated with primary antibodies diluted in blocking buffer, at 4°C overnight. Primary antibodies used were as follows: villin (Santa Cruz Biotechnology, sc-7672, 1:50), ezrin (Cell Signaling Technology, 3145, 1:1000), p-ERM (Cell Signaling Technology, 3726, 1:200), Mst4 (Abcam, ab52491, 1:200), PKCζ (Santa Cruz Biotechnology, sc-216-G, 1:50), Lamp2 [Developmental Studies Hybridoma Bank (DSHB), ABL-93-S, 1:200], LC3A/B (Abcam, ab58610, 1:200), syntaxin 3 (Abcam, ab4113, 1:200), syndapin 2 (Sigma-Aldrich, HPA049854, 1:200), NHE3 (DSHB, AFFN-SLC9A3R1-9B6, 1:50), sucrase-isomaltase (Santa Cruz Biotechnology, sc-27603, 1:50), E-cadherin (BD Biosciences, 610182, 1:500), β-catenin (Cell Signaling Technology, 19807, 1:1000), Na+/K+-ATPases (DSHB, a5, 1:50), integrin β1 (Santa Cruz Biotechnology, sc-8978, 1:50) and laminin (Abcam, ab11575, 1:500). After washing with PBS, the slides were incubated with a fluorescently conjugated secondary antibody (Thermo Fisher Scientific, 1:1000) in the dark for 1–2 h. The slides were counterstained with TOPRO-3 (Thermo Fisher Scientific, T3605, 1:300) for nuclei, and mounted with ProLong Gold anti-fade mounting medium (Thermo Fisher Scientific, P36930). VECTOR Red Alkaline Phosphatase Substrate Kit (Vector Laboratories, SK-5100) was used to detect endogenous alkaline phosphatase activity in the intestinal brush border. Images were taken using a Zeiss LSM 510 confocal microscope. For integrin β1 immunohistochemistry, slides were quenched for endogenous peroxidase after antigen retrieval. After incubation with primary integrin β1 antibody (Santa Cruz Biotechnology, sc-8978, 1:500) at 4°C overnight, slides were washed and incubated with biotin-labeled secondary antibody (Vector Laboratories, 1:500) for 1–2 h. Staining was developed using an ABC Kit (Vector Laboratories, PK-6100), and then slides were counterstained with Hematoxylin (Vector Laboratories, H-3404) and mounted with Cytoseal mounting medium (Thermo Fisher Scientific, 8310-4). Images were taken under a Nikon (TE2000-U) microscope.
Transmission and SEM
Procedures for TEM and SEM have been described (Das et al., 2015; Gao and Kaestner, 2010; Knowles et al., 2015). Duodenal or ileal tissues were freshly dissected from wild-type and DKOΔIEC (n=3) mice at P1. For inducible Rab11a deletion, wild-type and Rab11afl/fl;Villin-CreER mice received tamoxifen injection at P3 and were killed at P6. Segments of duodenum or ileum (0.5 mm) were freshly dissected and fixed in fixative solution containing 2.5% glutaraldehyde and 2% paraformaldehyde in 0.1 M sodium cacodylate at 4°C overnight. After postfixing with 1% OsO4, samples were dehydrated and embedded following standard procedures. Ultrathin sections were mounted on copper grids, contrasted with uranyl acetate/lead citrate double staining and examined under an electron microscope. For SEM, after dehydration, the samples were incubated with hexamethyldisilazane (HMDS). After the HMDS treatment, the samples were mounted on stubs, coated with gold in a sputter coater and analyzed using an FEI Quanta 250 scanning electron microscope (Vanderbilt Cell Imaging Shared Resource, Vanderbilt University).
Mouse duodenums from 1-day-old wild-type and double-knockout mice (n=4) were dissected and lysed in lysis buffer (50 mM Tris-HCl pH 7.5, 150 mM NaCl, 10 mM EDTA, 10 mM EGTA, 25 mM NaF, 3 mM Na3VO4, 1% Triton X-100, 0.5% sodium deoxycholate, 0.1% SDS, 1 mM PMSF, 0.5 mM DTT and protease inhibitors) at 4°C by sonication. Lysates were centrifuged at 17,000 g, at 4°C, for 15 min. Supernatants were collected and analyzed for protein concentration. After adding 1× lithium dodecyl sulfate (LDS) and 50 mM DTT (final concentration), the samples were incubated at 70°C for 10 min. Samples were loaded on SDS-PAGE gels, transferred to PVDF membranes and blocked in 5% skim milk in Tris-buffered saline with 0.1% Tween 20 (TBST) for 1 h at room temperature. The membrane was incubated with the following primary antibodies: ezrin (Cell Signaling Technology, 3145, 1:1000), villin (Santa Cruz Biotechnology, sc-7672, 1:500), integrin β1 (Santa Cruz Biotechnology, sc-8978, 1:500), Mst4 (Abcam, ab11575, 1:1000), PKCζ (Santa Cruz Biotechnology, sc-216-G, 1:1000), p-PKCζ (Santa Cruz Biotechnology, sc-271962, 1:1000), LKB1 (Abcam, ab58786-100, 1:1000), NHE3 (DSHB, AFFN-SLC9A3R1-9B6, 1:200) and β-actin (Santa Cruz Biotechnology, sc-47778, 1:2000), at 4°C, overnight. After washing, membranes were incubated with secondary antibody (GE Amersham, NA931V and NA934V, 1:2000) for 1 h at room temperature and developed in ECL detection reagents (GE Amersham, RPN2209 and PRN2232).
Organoids were isolated and cultured from the proximal small intestines of 1-day-old wild-type (Cre-negative), double-knockout (Rab8afl/fl;Rab11afl/fl;Villin-Cre) and single-knockout (Rab8afl/fl;Rab11afl/+;Villin-Cre or Rab8afl/+;Rab11afl/fl;Villin-Cre) mice. Existing culture protocols were modified for optimal culturing of neonatal intestines (Sato and Clevers, 2013; Sato et al., 2009). Briefly, organoids were maintained in EGF (50 ng/ml, R&D Systems, 2028-EG-200), noggin (100 ng/ml, Peprotech, 250-38) and R-spondin-1 (1 μg/ml, R&D Systems, 3474-RS-050) (ENR) containing medium supplemented with Wnt3a (100 ng/ml, Peprotech, 315-20). Organoid growth was monitored by imaging and counted on a daily basis. EdU labeling and staining (Thermo Fisher Scientific, C10638) was performed according to the manufacturer's instructions. Images were taken using a Nikon (TE2000-U) light microscope and a Zeiss LSM 510 confocal microscope.
All experiments were conducted on littermate animals, and comparisons were made within the same experiment. At least three independent animals were used in experiments for each genotype unless otherwise noted. Data represent mean±s.e.m. from independent experiments. Statistical analyses were performed to evaluate the differences between groups using one-way or two-way ANOVA. Student's t-test (unpaired and two-tailed) was used for TEM quantifications. P<0.05 was considered significant. ImageJ version 1.49v (National Institutes of Health) was used for measurements and GraphPad Prism (7.02) was used to create graphs.
Conceptualization: Q.F., E.M.B., L.Z., M.J.T., J.R.G., N.G.; Methodology: Q.F., E.M.B., A.C.E., L.Z., M.J.T., J.R.G., N.G.; Software: N.G.; Validation: Q.F., E.M.B., A.C.E., L.Z., J.R.G., N.G.; Formal analysis: Q.F., E.M.B., A.C.E., L.Z., M.J.T., J.R.G., N.G.; Investigation: Q.F., E.M.B., A.C.E., L.Z., M.J.T., J.R.G., N.G.; Resources: J.R.G., N.G.; Data curation: Q.F., E.M.B., A.C.E., L.Z., J.R.G., N.G.; Writing - original draft: Q.F., N.G.; Writing - review & editing: Q.F., E.M.B., A.C.E., L.Z., M.J.T., J.R.G., N.G.; Visualization: N.G.; Supervision: E.M.B., J.R.G., N.G.; Project administration: N.G.; Funding acquisition: J.R.G., N.G.
This study was supported by the National Institutes of Health (DK102934 and CA178599 to J.R.G.), American Cancer Society (RSG-15-060-01-TBE to N.G.), New Jersey Commission on Cancer Research (DFHS16PPC045 to Q.F.), National Science Foundation [NSF/BIO/IDBR (1353890)], and Rutgers, the State University of New Jersey [Initiative for Multidisciplinary Research Teams (IMRT) Award]. Deposited in PMC for release after 12 months.
The authors declare no competing or financial interests.