ABSTRACT
Enteroendocrine cells (EECs) are specialized sensory cells in the intestinal epithelium that sense and transduce nutrient information. Consumption of dietary fat contributes to metabolic disorders, but EEC adaptations to high fat feeding were unknown. Here, we established a new experimental system to directly investigate EEC activity in vivo using a zebrafish reporter of EEC calcium signaling. Our results reveal that high fat feeding alters EEC morphology and converts them into a nutrient insensitive state that is coupled to endoplasmic reticulum (ER) stress. We called this novel adaptation “EEC silencing”. Gnotobiotic studies revealed that germ-free zebrafish are resistant to high fat diet induced EEC silencing. High fat feeding altered gut microbiota composition including enrichment of Acinetobacter species, and we identified an Acinetobacter strain sufficient to induce EEC silencing. These results establish a new mechanism by which dietary fat and gut microbiota modulate EEC nutrient sensing and signaling.
INTRODUCTION
All animals derive energy from dietary nutrient ingestion. The energy harvested through digestion and absorption of dietary nutrients in the intestine is consumed by metabolic processes or stored as fat in adipose tissues. Excessive nutrient intake leads to metabolic disorders such as obesity and type 2 diabetes. To maintain energy homeostasis the animal must constantly monitor and adjust nutrient ingestion in order to balance metabolic needs with energy storage and energy intake. To accurately assess energy intake, animals evolved robust systems to monitor nutrient intake and communicate this dynamic information to the rest of the body. However, the physiological mechanisms by which animals monitor and adapt to nutrient intake remain poorly understood.
The primary sensory cells in the gut epithelium that monitor the luminal nutrient status are enteroendocrine cells (EECs) (Furness, Rivera, Cho, Bravo, & Callaghan, 2013). These hormone-secreting cells are dispersed along the entire gastrointestinal tract but comprise only ~1% of gut epithelial cells (Sternini, Anselmi, & Rozengurt, 2008). However, collectively these cells constitute the largest, most complex endocrine network in the body. EECs synthesize and secrete hormones in response to ingested nutrients including carbohydrates, fatty acids, peptides and amino acids (Delzenne, Cani, & Neyrinck, 2007; Moran-Ramos, Tovar, & Torres, 2012). These nutrients directly stimulate EECs by triggering a cascade of membrane depolarization, action potential firing and voltage dependent calcium entry. Increase of intracellular calcium ([Ca2+]i) can trigger the fusion of hormone-containing vesicles with the cytoplasmic membrane and hormone release (Sternini et al., 2008). The apical surface of most EECs are exposed to the gut lumen allowing them to detect ingested luminal contents (Gribble & Reimann, 2016). However, some EECs are not open to the gut lumen and reside close to the basal lamina (Hofer, Asan, & Drenckhahn, 1999; Sternini et al., 2008). These different morphological types are classified as “open” or “closed” EECs respectively, and traditionally have been thought to reflect distinct developmental cell fates. However, the transition between open and closed EEC types has not been described.
Besides morphological characterization, EECs are commonly classified by the hormones they express. More than 15 different hormones have been identified in EECs which exert broad physiological effects on gut motility, satiation, food digestion, nutrient absorption, insulin sensitivity, and energy storage (Moran-Ramos et al., 2012). EECs communicate not only through circulating hormones, but also through direct paracrine and neuronal signaling to multiple systems including the intrinsic and extrinsic nervous system, pancreas, liver and adipose tissue (Bohorquez et al., 2015; Gribble & Reimann, 2016; Kaelberer et al., 2018; Latorre, Sternini, De Giorgio, & Greenwood-Van Meerveld, 2016). EECs therefore have a key role in regulating energy homeostasis and represent the first link that connects dietary nutrient status to systemic metabolic processes.
Energy homeostasis can be influenced by many environmental factors, although diet plays the most important role. Despite efforts to reduce dietary fat intake in recent decades, the percentage of energy intake from fat remains ~33% in the US (Austin, Ogden, & Hill, 2011). High levels of dietary fat have a dominant effect on energy intake and adiposity (Hu et al., 2018) and have been implicated in the high prevalence of human metabolic disorders worldwide (Ludwig, Willett, Volek, & Neuhouser, 2018; Oakes, Cooney, Camilleri, Chisholm, & Kraegen, 1997; Panchal et al., 2011). The effects of a high fat diet on peripheral tissues like pancreatic islets, liver and adipose tissue have been studied extensively (Green & Hodson, 2014; Kahn, Hull, & Utzschneider, 2006). It is also well appreciated that consumption of a high fat diet affects the microbial communities residing in the intestine, commonly refered to as the gut microbiota (David et al., 2014; Hildebrandt et al., 2009; Murphy et al., 2010; Turnbaugh, Backhed, Fulton, & Gordon, 2008; Wong et al., 2015). Gnotobiotic animal studies also demonstrated that gut microbiotas altered by high fat diet can promote adiposity and insulin resistance (Ridaura et al., 2013; Turnbaugh et al., 2008; Turnbaugh et al., 2006), but the underlying mechanisms are incompletely understood. Notably, despite the importance of EECs in nutrient monitoring and systemic metabolic regulation, it remains unknown how a high fat diet might impact EECs function and whether the gut microbiota play a role in this process.
A major problem in studying the effects of diet on EEC physiology has been the lack of in vivo techniques for studying these rare cells in an intact animal. Historically, in vivo EEC function has been studied by measuring hormone levels in blood following luminal nutrient stimulation (Goldspink, Reimann, & Gribble, 2018). However, many gastrointestinal hormones have very short half-lives and peripheral plasma hormone levels do not mirror real-time EEC function (Cuenco et al., 2017; Druce et al., 2009; Kieffer, McIntosh, & Pederson, 1995). EEC function has been measured in vitro via cell and organoid culture models using electrophysiological cellular recordings and fluorescence-based calcium imaging (Kaelberer et al., 2018; Kay, Boissy, Russnak, & Candido, 1986; Reimann et al., 2008). However, these in vitro models are not suited for modeling the effect of diet and microbiota on EEC function as they are unable to reproduce the complex in vivo environment that involves signals from neighboring cells like enterocytes, enteric nerves, blood vessels and immune cells. Moreover, in vitro culture systems are unable to mimic the dynamic and complex luminal environment that contains food and microbiota. Therefore, to fully understand the effects of diet and microbiota on EEC function, it is necessary to study EECs in vivo.
In this study, we utilized the zebrafish model to investigate the impact of dietary nutrients and microbiota on EEC function. The development and physiology of the zebrafish digestive tract is similar to that of mammals (Wallace, Akhter, Smith, Lorent, & Pack, 2005; Wallace & Pack, 2003). Zebrafish hatch from their protective chorions at 3 days post-fertilization (dpf) and microbial colonization of the intestinal lumen begins shortly thereafter (Rawls, Mahowald, Goodman, Trent, & Gordon, 2007). The zebrafish intestine becomes completely patent by 4 dpf and feeding and digestion begin around 5 dpf. The zebrafish intestine develops most of the same differentiated epithelial cell types as observed in mammals, including absorptive enterocytes, mucus-secreting goblet cells, and EECs (Ng et al., 2005; Wallace et al., 2005; Wallace & Pack, 2003). Digestion and absorption of dietary fat occur primarily in enterocytes within the proximal intestine of the zebrafish (Quinlivan & Farber, 2017) (yellow area in Figure 1D). These conserved aspects of intestinal epithelial anatomy and physiology are associated with a conserved transcriptional regulatory program shared between zebrafish and mammals (Lickwar et al., 2017). To monitor EEC activity in zebrafish, we used a genetically encoded calcium indicator (Gcamp6f) expressed under control of an EEC gene promoter. The excitability of EECs upon luminal stimulation could be measured using in vivo fluorescence-based calcium imaging. By combining this in vivo EEC activity assay with diet and gnotobiotic manipulations, we show here that specific members of the intestinal microbiota mediate a novel physiologic adaption of EECs to high fat diet.
RESULTS
Establishing methods to study enteroendocrine cell function using an in vivo zebrafish model
We first developed an approach to identify and visualize zebrafish EECs in vivo. Previous mouse studies have shown that the transcription factor NeuroD1 plays an essential role to restrict intestinal progenitor cells to an EEC fate (H. J. Li, Ray, Singh, Johnston, & Leiter, 2011; Ray & Leiter, 2007), and is expressed in almost all EECs without expression in other intestinal epithelial cell lineages (H. J. Li, Kapoor, Giel-Moloney, Rindi, & Leiter, 2012; Ray, Li, Metzger, Schule, & Leiter, 2014). We used transgenic zebrafish lines expressing fluorescent proteins under control of regulatory sequences from the zebrafish neurod1 gene, Tg(-5kbneurod1:TagRFP) (McGraw, Snelson, Prendergast, Suli, & Raible, 2012) and TgBAC(neurod1:EGFP) (Trapani, Obholzer, Mo, Brockerhoff, & Nicolson, 2009). We found that both lines labeled cells in the intestinal epithelium of 6 dpf zebrafish (Fig. 1A-B, Fig. S1A), and that these neurod1+ cells do not overlap with goblet cells and express the intestinal secretory cell marker 2F11 (Crosnier et al., 2005) (Fig. S1C-E). To further test whether these neurod1+ cells in the intestine label secretory but not absorptive cell lineages, we crossed Tg(-5kbneurod1:TagRFP) with the Notch reporter line Tg(tp1:EGFP) (Parsons et al., 2009). Activation of Notch signaling is essential to restrict intestinal progenitor cells to an absorptive cell fate (Crosnier et al., 2005; H. J. Li et al., 2012), suggesting tp1+ cells may represent enterocyte progenitors. In accord, we found that neurod1+ cells in the intestine do not overlap with tp1+ cells (Fig. S1B). Additionally, our results demonstrated that neurod1+ cells in the intestine do not overlap with the mature enterocyte marker ifabp/fabp2 (Kanther et al., 2011)(Fig. 1C). These results suggested that, similar to mammals, neurod1 expression in the zebrafish intestine occurs specifically in EECs. In addition, using EdU labeling at 5 dpf zebrafish larvae, we found that EECs in the intestine are post-mitotic and require about 30 hours to differentiate from proliferating progenitors (Fig. S2A-F).
Hormone expression is a defining feature of EECs, so we next evaluated the expression of four hormones in neurod1+ EECs in 6 dpf zebrafish larvae: pancreatic peptide YY (PYY), cholecystokinin (CCK), somatostatin (Tg(sst2:RFP),(Z. Li, Wen, Peng, Korzh, & Gong, 2009)) and glucagon (precursor to glucagon-like peptides GLP-1 and GLP-2; Tg(gcga:EGFP), (Ye, Robertson, Hesselson, Stainier, & Anderson, 2015)) (Fig. 1E-H). We found that PYY and CCK hormones, which are important for regulating fat digestion and feeding behavior, are only expressed in EECs in the proximal intestine where dietary fats and other nutrients are digested and absorbed (Carten, Bradford, & Farber, 2011; Farber et al., 2001) (Fig. 1I-J). In contrast, somatostatin expression occurred in EECs along the whole intestine and glucagon expressing EECs were found along proximal and mid-intestine but excluded from the distal intestine (Fig. S1F-G). The regionalization of EEC hormone expression may reflect the functional difference of EECs and other epithelial cell types along the zebrafish intestine (Lickwar et al., 2017).
EECs are specialized sensory cells in the intestinal epithelium that can sense nutrient stimuli derived from the diet such as glucose, amino acids and fatty acids. Upon receptor-mediated nutrient simulation, EECs undergo membrane depolarization that results in transient increases in intracellular calcium that in turn induce release of hormones or neurotransmitters (Goldspink et al., 2018). Therefore, the transient increase in intracellular calcium concentration is an important mediator and indicator of EEC function. To investigate EEC function in zebrafish, we utilized a neurod1:Gcamp6f transgenic zebrafish model (Rupprecht, Prendergast, Wyart, & Friedrich, 2016), in which the calcium-dependent fluorescent protein Gcamp6f is expressed in EECs under control of the −5kb neurod1 promoter (McGraw et al., 2012). Using this transgenic line, we established an in vivo EEC activity assay system which permitted us to investigate the temporal and spatial activity of EECs in vivo. Briefly, unanesthetized Tg(neurod1:Gcamp6f) zebrafish larvae were positioned under a microscope objective and a solution containing a stimulus was delivered onto their mouth. The stimulus was then taken up into the intestinal lumen and EEC Gcamp6f activity was recorded simultaneously (Fig. 2A; see Methods and Fig. S3 for further details). Using this EEC activity assay, we first tested if zebrafish EECs were activated by fatty acids. We found that palmitate, but not the BSA vehicle control, activated a subset of EECs (Fig. 2B-F, supplemental video 1). Similar patterns of EEC activation in the proximal intestine were induced by the fatty acids linoleate and dodecanoate; whereas, the short chain fatty acid butyrate did not induce EEC activity (Fig. 2D). The ability of EECs in the proximal intestine to respond to fatty acid stimulation is interesting because that region is the site of dietary fatty acid absorption (Carten et al., 2011). In this region EECs express CCK which regulates lipase and bile secretion and PYY which regulates food intake (Fig. 1IJ). Our results further establish that activation by fatty acids is a conserved trait in zebrafish and mammalian EECs.
High fat feeding impairs enteroendocrine cell nutrient sensing
The vast majority of previous studies on EECs in all vertebrates has focused on acute stimulation with dietary nutrients including fatty acids. In contrast, we have very little information on the adaptations that EECs undergo during the postprandial process. To address this gap in knowledge, we applied an established model for high fat meal feeding in zebrafish (Carten et al., 2011; Semova et al., 2012). In this high fat (HF) meal model, zebrafish larvae are immersed in a solution containing an emulsion of chicken egg yolk liposomes which they ingest for a designated amount of time prior to postprandial analysis using our EEC activity assay (Fig. 2G). Importantly, ingestion of a HF meal does not prevent subsequent nutrient stimuli such as fatty acids to be ingested and distributed along the length of the intestine (Fig. S4A-F). To our surprise, we found that the ability of EECs in the proximal intestine to respond to palmitate stimulation in our EEC activity assay was quickly and significantly reduced after 6 hours of HF meal feeding (Fig. 2H-J, supplemental video 2).
We next sought to test if HF feeding only impairs EEC sensitivity to fatty acids or if there are broader impacts on EEC nutrient sensitivity. First, we investigated EEC responses to glucose stimulation. Similar to fatty acids, glucose stimulation activated EECs only in the proximal intestine of the zebrafish under unfed control conditions (Fig. 3A, B, supplemental video 1). Previous mammalian cell culture studies reported that glucose-stimulated elevation of intracellular calcium concentrations and hormone secretion in EECs is dependent upon the EEC sodium dependent glucose cotransporter 1 (Sglt1), an apical membrane protein that is expressed in small intestine and renal tubles and actively transports glucose and galactose into cells (Song, Onishi, Koepsell, & Vallon, 2016). Similarly, we found that Sglt1 is expressed on the apical surface of zebrafish intestinal epithelial cells including enterocytes and EECs (Fig. 3E). In addition, co-stimulation with glucose and phlorizin, a chemical inhibitor of Sglt1, blocked the EEC activation induced by glucose (Fig. 3F-G). Consistently, the EEC response to glucose stimulation was significantly increased by the addition of NaCl in the stimulant solution which will facilate sodium gradient dependent glucose transport by Sglt1 (Fig. 3C). In addition, zebrafish EECs also responded to the other Sglt1 substrate, galactose, but not fructose (Fig. 3D). These results suggest that glucose can induce EEC activity in a Sglt1 dependent manner in the zebrafish intestine.
We then examined if HF feeding impaired subsequent EEC responses to glucose, as we had observed for fatty acids (Fig.2G-J). Indeed, HF feeding significantly reduced EECs’ response to subsequent glucose stimulation (Fig. 3H-J, supplemental video 3). We extended these studies to investigate zebrafish EEC responses to amino acids. Among the twenty major amino acids we tested, we only observed significant EEC activity in response to cysteine stimulation under control conditions (Fig. S5A-B, supplemental video 1). However, in contrast to the fatty acid and glucose responses, zebrafish EECs that respond to cysteine were located primarily in the mid intestine (Fig. S5A-B) and HF meal ingestion did not significantly impair subsequent EEC responses to cysteine (Fig. S5C-E). These results collectively indicate that HF feeding impairs the function of palmitate and glucose responsive EECs in the proximal intestine, the region where fat absorption take place.
High fat feeding induces morphological adaption in enteroendocrine cells
To further investigate how HF feeding impacts zebrafish EECs, we leveraged the transparency of the zebrafish to permit morphologic analysis of EECs. In zebrafish under control conditions, most EECs are in an open-type morphology (Fig. 1B-G) with an apical process that extends to the intestinal lumen, allowing them to directly interact with the contents of the intestinal lumen (Fig. 4A). When we examined the proximal zebrafish intestine after 6 hours of HF feeding, we discovered that most EECs had adopted a closed-type morphology that apparently lacked an apical extension and no longer had access to the lumenal contents (Fig. 4B, Fig.S6A-C). We first speculated this shift from open-type to closed-type EEC morphology may be due to cell turnover and loss of open-type EECs and replacement with newly differentiated closed-type EECs. To test this possibility, we created a new Tg(neurod1:Gal4); Tg(UAS:Kaede) photoconversion tracing system in which UV light can be used to convert the Kaede protein expressed in EECs from green to red emission (Fig. S7A-C). This allowed us to label all existing differentiated neurod1+ EECs by UV light photoconversion immediately before HF feeding (Fig. S7G), so that pre-existing EECs emit red and green Kaede fluorescence and any newly differentiated EECs emit only green Kaede fluorescence (Fig. S7D-E). However, we did not observe the presence of any green EECs following HF feeding (Fig. S7F-G). To test whether HF feeding induced EEC apoptosis, we used an in vivo apoptosis model in which Tg(ubb:sec5A-tdTomato) (Scott T. Espenschied, 2019) zebrafish were crossed with TgBAC(neurod1:EGFP) allowing us to determine if apoptosis occurred in EECs (Fig. S8A-B). However, we did not detect activation of apoptosis in closed-type EECs following high fat diet feeding (Fig. S8C). These results suggest that the striking change in EEC morphology during HF feeding is not due to EEC turnover but is instead due to adaptation of the existing EECs.
To analyze this adaptation of EEC morphology in greater detail, we used a new transgenic model TgBAC(gata5:lifActin-EGFP) together with the Tg(-5kbneurod1:TagRFP) line. In these animals, the apical surface of EECs and other intestinal epithelial cells can be labeled by gata5:lifActin-EGFP and the cytoplasmic extension of EECs to the apical lumen can be visualized and quantified through z-stack confocal imaging of the proximal intestine (Supplemental video 4). We measured the ratio of EECs with apical extensions to the total number of EECs, and defined that ratio as an “EEC morphology score”. In control embryos, most EECs are open-type and the morphology score is near 1 (Fig. 4E). We found that the EEC morphology score gradually decreased upon high fat feeding (Fig. 4E, supplemental video 5), indicating that EECs had changed from an open-type to closed-type morphology. To further analyze the dynamics of the EEC apical response, we generated a new transgenic line Tg(-5kbneurod1:lifActin-EGFP)(Fig. S9 A-C). Using in vivo confocal time-lapse imaging in Tg(-5kbneurod1:lifActin-EGFP) zebrafish, we confirmed that EEC apical processes undergo dynamic retraction after HF feeding (Fig. 4F), which was not observed in control animals (Fig. S9 C-D, supplemental video 6). Interestingly, EECs in the distal intestine retained their open-type morphology following HF feeding (Fig. S6 F-H), suggesting the adaptation from open- to closed-type EEC morphology is a specific response of EECs in the proximal intestine. This suggests that this EEC morphological adaption upon HF feeding is associated with impairment of EEC sensitivity to subsequent exposure to nutrients such as palmitate and glucose. We operationally define this novel EEC morphological and functional postprandial adaption to HF feeding as “EEC silencing”.
Activation of ER stress following high fat feeding leads to EEC silencing
We next sought to identify the mechanisms underlying HF feeding-induced EEC silencing. Quantitative RT-PCR assays in dissected zebrafish digestive tracts revealed that HF feeding broadly increased expression of EEC hormones (Fig. 5A). The largest increases were pyyb and ccka (Fig. 5A), both of which are expressed by EECs in the proximal zebrafish intestine (Fig.1) and are important for the response to dietary lipid. However, HF feeding did not significantly alter expression of EEC specific transcription factors (neurod1, pax6b, isl1), nor the total number of EECs per animal (Fig. 5A, C). These data suggested that HF feeding challenges the existing EECs to increase hormone synthesis and secretion, perhaps in response to depletion of pre-existing hormone granules. We speculated that this increase in hormone synthesis might place an elevated demand and stress on the endoplasmic reticulum (ER), the organelle where hormone synthesis takes place. ER stress is known to activate a series of downstream cell signaling responses called the Unfolded Protein Response (UPR) (Hetz, 2012; Xu, Bailly-Maitre, & Reed, 2005). Increased misfolded protein and induction of ER stress activates ER membrane sensors Atf6, Perk and Ire1 (Hetz, 2012; Xu et al., 2005). The activated ER stress sensor Ire1 then splices mRNA encoding the transcription factor Xbp1, which in turn induces expression of target genes involved in the stress response and protein degradation, folding and processing (Yoshida, Matsui, Yamamoto, Okada, & Mori, 2001). Using quantitative RT-PCR analysis in dissected zebrafish digestive tracts, we found that HF feeding increased expression of UPR genes including chaperone proteins Gpr94 and Bip (Fig. 5B). To investigate whether ER stress is activated in EECs, we took advantage of a transgenic zebrafish Tg(ef1α:xbp1δ-gfp) that permits visualization of ER stress activation by expressing GFP only in cells undergoing xbp1 splicing (J. Li et al., 2015). We crossed Tg(ef1α:xbp1δ-gfp) with Tg(-5kbneurod1:TagRFP) zebrafish and found that zebrafish larvae fed a HF meal, but not control larvae, displayed a significant induction of GFP in neurod1+ EECs (Fig. 5 J, K, O). Next, we tested if activation of ER stress in EECs is required for EEC silencing. Whereas HF feeding normally reduces the EEC morphology score, this did not occur in zebrafish treated with tauroursodeoxycholic acid (TUDCA), a known ER stress inhibitor (Uppala, Gani, & Ramaiah, 2017; Vang, Longley, Steer, & Low, 2014) (Fig. 5 L-N, R).
To further confirm the hypothesis that ER stress activation can lead to EEC silencing, we tested if induction of ER stress is sufficient to cause EEC silencing independent of HF feeding. We treated 6 dpf Tg(neurod1:Gcamp6f) zebrafish larvae with thapsigargin, a chemical compound commonly used to induce ER stress by interrupting ER calcium storage and protein folding (Samali, Fitzgerald, Deegan, & Gupta, 2010), and then performed the EEC response assay. Thapsigargin treatment reduced the EEC calcium response to both glucose and palmitate (Fig. 5D-I) and decreased the EEC morphology score, both key phenomena associated with EEC silencing (Fig. 5P). To confirm this result, we tested a second ER stress inducer brefeldin A (BFA), which inhibits anterograde ER export to Golgi and blocks protein secretion (Donaldson, Cassel, Kahn, & Klausner, 1992; Klausner, Donaldson, & Lippincott-Schwartz, 1992). Similar to thapsigargin, treatment with BFA significantly decreased the EEC morphology score (Fig. 5Q). These results support a working model wherein increased hormone synthesis and secretion following HF feeding induces ER stress in EECs which leads to EEC silencing.
Blocking fat digestion and absorption inhibits EEC silencing following high fat feeding
We next sought to explore the physiological mechanisms within the gut lumen that may lead to EEC silencing after HF feeding. We reasoned that induction of ER stress in EECs after a HF meal is likely caused by over-stimulation with fatty acids and other nutrients derived from the meal. Fatty acids are liberated from dietary triglycerides in the gut lumen through the activity of lipases, so we predicted that lipase inhibition would block EEC silencing normally induced by HF feeding. We therefore treated zebrafish larvae with orlistat, a broad-spectrum lipase inhibitor commonly used to treat obesity (Ballinger, 2000; Hill et al., 1999). We found that treatment of Tg(neurod1:Gcamp6f) zebrafish with orlistat during HF feeding significantly increased the ability of EECs to subsequently respond to glucose and palmitate (Fig. 6 A-F). Next, we investigated the effect of orlistat on EEC morphology during HF feeding in Tg(gata5:lifActin-EGFP); Tg(-5kbneurod1:TagRFP) zebrafish. We found that following 10 hours of HF feeding, EECs in control animals had switched from an open-type to a closed-type morphology and significantly reduced the EEC morphology score (Fig. 6 G, N). By contrast, treatment with orlistat prevented HF induced EEC morphological changes (Fig. 6 H, N), suggesting lipase activity is required for EEC silencing.
To investigate further how orlistat treatment inhibits EEC silencing, we analyzed the its effect on ER stress in EECs following HF feeding using Tg(ef1α:xbp1δ-gfp) zebrafish. We found that orlistat treatment significantly reduced the percentage of EECs that are ef1α:xbp1δ-gfp+ following HF feeding (Fig. 6 I, J, O). We next sought to test if additional pathways are activated in EECs by HF feeding, and if those EEC responses are dependent on lipase activity or ER stress. Induction of ER stress can lead to activation of the transcription factor NF-κB through release of calcium from the ER, elevated reactive oxygen intermediates or direct Ire1 activity (Kim et al., 2015; Pahl & Baeuerle, 1997). After crossing a transgenic reporter of NF-κB activity Tg(NFkB:EGFP) (Kanther et al., 2011) with Tg(-5kbneurod1:TagRFP), we found that HF feeding significantly increased the number of NF-κB+ EECs (Fig. 6 K, P), but that this effect could be significantly reduced by treatment with orlistat or the ER stress inhibitor TUDCA (Fig. 6 L, M, P). These results indicate that EEC silencing and associated signaling events that follow ingestion of a HF meal require lipase activity.
Lipases act on dietary triglycerides to liberate fatty acids and monoacylglycerols that are then available for stimulation of EECs (Hara, Hirasawa, Ichimura, Kimura, & Tsujimoto, 2011; Lauffer, Iakoubov, & Brubaker, 2009). To test if free fatty acids are sufficient to induce EEC silencing, we treated 6 dpf zebrafish larvae with palmitate, a major fatty acid component in our HF meal (Poureslami, Raes, Huyghebaert, Batal, & De Smet, 2012). Treatment with palmitate for 6 hours significantly reduced the ability of EECs to respond to subsequent palmitate stimulation, but did not influence the EEC morphology score, nor the EEC response toward subsequent glucose stimulation (Fig. S10). These results suggest that the fatty acid palmitate is sufficient to induce only a portion of the EEC silencing phenotype induced by a complex HF meal.
High fat feeding induces EEC silencing in a microbiota dependent manner
Using the same HF feeding model in zebrafish, we previously showed that the gut microbiota promote intestinal absorption and metabolism of dietary fatty acids (Semova et al., 2012), and similar roles for microbiota have been established recently in mouse (Martinez-Guryn et al., 2018). We therefore predicted that the microbiota may also regulate EEC silencing after HF feeding. Using our established methods (Pham, Kanther, Semova, & Rawls, 2008), we raised Tg(gata5:lifActin-EGFP); Tg(-5kbneurod1:TagRFP) zebrafish larvae to 6 dpf in the absence of any microbes (germ free or GF) or colonized at 3 dpf with a complex zebrafish microbiota (ex-GF conventionalized or CV). In the absence of HF feeding, we observed no significant differences between GF and CV zebrafish in their EEC morphology score or EEC response to palmitate (Fig. 7C,D,G,I). We then performed HF feeding in these 6 dpf GF and CV zebrafish larvae. In contrast to CV HF-fed zebrafish larvae, EECs in GF zebrafish did not show a change in morphology after HF feeding (Fig. 7A, B, I) and exhibited significantly greater responses to palmitate stimulation (Fig. 7E, F, H). In accord, the ability of HF feeding to induce reporters of ER stress and NF-κB activation was significantly reduced in GF compared to CV zebrafish (Fig.7J,K). These results indicate that colonization by microbiota mediates EEC silencing in HF fed zebrafish. EECs are known to express Toll-like receptors (TLRs) (Kanwal, Wiegertjes, Veneman, Meijer, & Spaink, 2014) (Palti, 2011), which sense diverse microbe-associated molecular patterns and signal through the downstream adaptor protein Myd88 leading to activation of NF-κB and other pathways (Kawasaki & Kawai, 2014). To test if EEC silencing requires TLR signaling, we evaluated myd88 mutant zebrafish (Burns et al., 2017). We found that EECs’ response to palmitate after HF feeding was equivalent to that of wild type fish (Fig. S11 A-B), suggesting microbiota promote EEC silencing in a Myd88-independent manner.
HF diets are known to significantly alter gut microbiota composition in human, mice and zebrafish (David et al., 2014; Hildebrandt et al., 2009; Wong et al., 2015). We therefore hypothesized that HF feeding might alter the composition of the microbiota, which in turn might promote EEC silencing. To test this possibility, we first analyzed the effects of HF feeding on intestinal microbiota density through colony forming unit (CFU) analysis in dissected intestines from CV zebrafish larvae. Strikingly, we found that intestinal microbiota abundance had increased ~20-fold following 6 hours of HF feeding (Fig. 8A). To determine if this increase in bacterial density was accompanied by alterations in bacterial community structure, we performed 16S rRNA gene sequencing. Since diet manipulations can alter microbiota composition in the zebrafish gut as well as their housing water media (Wong et al., 2015), we analyzed samples from dissected intestines of zebrafish larvae in control and HF fed groups as well as their respective housing medias (Fig. 8B). Analysis of bacterial community structure using the Weighted Unifrac method (Caporaso et al., 2010) revealed, as expected, relatively large differences between gut and media samples (PERMANOVA P<0.02 control gut vs. control media, P<0.005 HF gut vs HF media) (Fig. 8C). The addition of HF feeding had a relatively smaller but consistent effect on overall bacterial community structure in both gut and media (PERMANOVA P=0.2 control gut vs HF gut, P=0.094 control media vs HF media) (Fig. 8C). HF feeding caused a small reduction in within-sample diversity among media microbiotas as measured by Faith’s Phylogenetic Diversity (Kruskal-Wallis P=0.049), but no significant effects on gut microbiotas (P=0.29)(Faith & Baker, 2007). Taxonomic analysis of zebrafish gut and media samples revealed several bacterial taxa significantly affected by HF feeding (Table S2). Members of class Betaproteobacteria dominated the control media, but HF feeding markedly decreased their relative abundance (LDA effect size 5.45, P=0.049). Conversely, HF feeding increased the relative abundance of members of class Gammaproteobacteria (LDA effect size 5.49, P=0.049; Fig.8D) such as genera Acinetobacter (LDA effect size 5.13, P=0.049), Pseudomonas (LDA effect size 5.02, P=0.049) and Aeromonas (LDA effect size 4.78, P=0.049; Fig. 8E; Tables S2 and S3). HF feeding also increased the relative abundance in media of class Cytophagia from phylum Bacteroidetes (LDA effect size 4.66, P=0.049; Fig. 8D) due to increases in the genus Flectobacillus (LDA effect size 4.76, P=0.049; Fig. 8F; Tables S2 and S3). The increased relative abundances of Aeromonas sp. and Pseudomonas sp. in HF fed medias was not recapitulated in the gut microbiotas (Fig.8G; Table S2). However, similar to the media, HF feeding significantly increased abundance of class Cytophagia (LDA effect size 4.01, P=0.018; Fig.8D) due to enrichment of Flectobacillus (LDA effect size 4.01, P=0.004; Fig.8H). Additionally, HF feeding resulted in a 100-fold increase the relative abundance of Acinetobacter sp. in the gut (average 0.04% in control gut, 4.28% in HF gut; LDA effect size 4.31, P=0.001; Fig. 8G, Tables S2 and S4). These results establish that HF feeding has diverse effects on the bacterial communities in the zebrafish gut and media, and raise the possibility that members of these affected bacterial genera may regulate EEC silencing in response to HF feeding.
We next tested if EEC silencing could be facilitated by representative members of the zebrafish microbiota, including those enriched by HF feeding. We selected a small panel of bacterial strains that were isolated previously from the zebrafish intestine (Stephens et al., 2016) and used them to monoassociate separate cohorts of GF Tg(gata5:lifActin-EGFP); Tg(-5kbneurod1:TagRFP) zebrafish at 3dpf (Fig. 8I). These bacteria strains were from 9 different genera including Acinetobacter sp. ZOR0008, a member of the Acinetobacter calcoaceticus-Acinetobacter baumannii complex (Gerner-Smidt, Tjernberg, & Ursing, 1991) (Bouvet & Jeanjean, 1989). We did not observe significant differences in colonization efficiency among these bacteria strains that were inoculated into GF zebrafish (Fig. S12A-B). At 6 dpf, we performed HF feeding and examined the EEC morphology score. Strikingly, only Acinetobacter sp. ZOR0008 was sufficient to significantly reduce the EEC morphology score upon HF feeding (Fig. 8J) similar to conventionalized animals (Fig. 7A,B,I). These results indicate that the effects of microbiota on EEC silencing following HF feeding display strong bacterial species specificity, and suggest Acinetobacter bacteria enriched by HF feeding may mediate the effect of microbiota on HF sensing by EECs.
Discussion
In this study, we established a new experimental system to directly investigate EEC activity in vivo using a zebrafish reporter of EEC calcium signaling. Combining genetics, diet and gnotobiotic manipulations allowed us to uncover a novel EEC adaptation mechanism through which high fat feeding induces rapid change of EEC morphology and reduced nutrient sensitivity. We called this novel adaptation “EEC silencing”. Our results show that EEC silencing following a high fat meal requires lipase activity and it is coupled to ER stress. Furthermore, HF meal induced EEC silencing is promoted by certain microbial species (e.g., Acinetobacter sp.). As discussed below, we propose a working model (Fig. S13) that nutrient over-stimulation from high fat feeding increases EECs hormone synthesis burden, overgrowth of the gut bacterial community including enrichment of Acinetobacter sp., which in turn activates EECs ER stress response pathways and thereby induces EECs silencing. This study demonstrates the utility of the zebrafish model to study in vivo interactions between diet, gut microbes, and EEC physiology. In the future, the mechanisms underlying EEC silencing could be targeted in rational manipulations of EEC adaptations to diet and microbiota which could be used to reduce incidence and severity of metabolic diseases.
EEC physiology in zebrafish
Our studies here provide important new tools for studying EECs in the context of zebrafish intestinal epithelial development and physiology. Similar to mammals, fish EECs are thought to arise from intestinal stem cells through a series of signals that govern the differentiation process (Aghaallaei et al., 2016). Delta-Notch signaling appears to control the differentiation of stem cells into absorptive and secretory cell lineages in both zebrafish and mammalian models (Crosnier et al., 2005). Activation of Notch signaling can block the differentiation of EECs by inhibiting the expression of key EEC bHLH transcription factors (H. J. Li et al., 2011). In mammals, the bHLH transcription factor Neurod1 that has been shown to regulate EEC terminal differentiation (H. J. Li et al., 2011; Ray & Leiter, 2007). Our results indicate that Neurod1 is expressed by and important in EEC differentiation in zebrafish as it is in mammals. Moreover, this finding enabled us to use neurod1 regulatory sequences to label and monitor zebrafish EECs.
The hallmark of EECs is their expression of hormones. In this study, using transgenic reporter lines and immunofluorescence staining approaches to examine a panel of gut hormones in zebrafish EECs, we found that zebrafish EECs express conserved hormones as mammalian EECs. Interestingly, a subset of EECs express proglucagon peptide which can be processed to hormones Glucagon like peptide 1 (GLP-1) and 2 (GLP-2) (Sandoval & D’Alessio, 2015). GLP-1, one of the incretin hormones, is released by EECs in response to oral glucose intake that can faclilate insulin secretion and reduce blood glucose (Drucker, Habener, & Holst, 2017). Multiple studies suggest that the expression of Sglt1 is important for EEC glucose sensing (Gorboulev et al., 2012; Reimann et al., 2008; Roder et al., 2014). EECs in Sglt1 knockout mice fail to secrete GLP-1 in response to glucose and galactose (Gorboulev et al., 2012). In our studies, we identified similar Sglt1 mediated glucose sensing machinery in zebrafish EECs. This together suggest that zebrafish EECs may exhibit conserved roles in regulating glucose metabolism.
Our data also establish that zebrafish EECs develop striking regional specificity in the hormones they express along the intestine (Fig. S1). For example, the CCK and PYY hormones that are important for regulating food digestion and energy homeostasis (Beglinger & Degen, 2006; Liddle, 1997; Raybould, 2007) were only expressed in the proximal intestine. In addition to hormonal regional specificity, we found that EEC calcium response toward nutrients also display regional specificity. For example, glucose and long chain/medium chain fatty acids only stimulate EECs in proximal intestine, a region in zebrafish where digestion and absorption of dietary fats primarily occurs (Carten et al., 2011). This hormonal and functional regional specificity suggests that distinct developmental and physiological programs govern EEC function along the intestinal tract, and that EECs in the proximal zebrafish intestine may play key roles in monitoring and adapting to dietary nutrient experience.
EEC silencing
In this study, we discovered that high fat feeding can induce a series of functional and morphological changes in EECs we refer to as “EEC silencing”. EEC silencing includes (1) reduced EEC sensitivity to nutrient stimulation (e.g., fatty acids and glucose) and (2) conversion of EEC morphology from an open to a closed type. To our knowledge, EEC silencing has not been observed in previous studies of EEC in any vertebrate. This underscores the unique power of in vivo imaging in zebrafish to reveal new physiologic and metabolic processes. Our evidence suggests that EEC silencing is a stress response that EECs display following consumption of a high fat meal in the presence of specific microbes. However, EEC silencing may also serve to protect EECs against excessive stress following consumption of a high fat meal. In neurons for example, similar desensitization has been shown to protect nerve cells from excitatory neurotransmitter induced toxicity (Gainetdinov, Premont, Bohn, Lefkowitz, & Caron, 2004; Quick & Lester, 2002) and blocking desensitization of excitatory neuronal receptors induces rapid neuronal cell death (Walker et al., 2009). High dietary fat can also lead to excessive production of excitatory stimuli like long-chain fatty acids. We speculate that EEC silencing provides an adaptive mechanism for EECs to avoid excessive stimuli and protect against cell death.
The observation that EECs exhibited reduced sensitivity to oral glucose following high fat feeding is interesting and consistent with the finding in mice that high fat feeding reduces intestinal glucose sensing and glucose induces GLP-1 secretion in vivo (Bauer et al., 2018). In vitro, small intestinal cultures from high fat fed mice also exhibit reduced secretory responsiveness to nutrient stimuli including glucose when comparing with intestine cultures from control mice (Richards et al., 2016) but underlying mechanisms remained unclear. These studies, together with our results, suggest that high fat feeding impairs EEC function. However, how high fat feeding reduces EEC glucose sensitivity is still unclear as we did not detect changes in the EEC glucose sensor sglt1 expression in high fat fed dissected intestine (data not shown). One possibility is that high fat feeding affects EECs glucose sensing via altering Sglt1 activity (Ishikawa, Eguchi, & Ishida, 1997; Subramanian, Glitz, Kipp, Kinne, & Castaneda, 2009; Wright, Hirsch, Loo, & Zampighi, 1997). We also speculate that high fat feeding induced EEC morphological changes may contribute to EEC glucose insensitivity. Since Sglt1 is expressed on the brush border at the apical surface of the cell, as EECs switch from an open to closed type morphology they would lose their contact with the gut lumen and exposure to luminal glucose stimuli.
Our observation that EECs can change their morphology from an “open” to “closed” state upon high fat feeding was surprising. The majority of EECs in the intestinal tract are open with an apical extension and microvilli facing the intestinal lumen. In contrast, some EECs lie flat on the basement membrane and are “closed” to the gut lumen (Gribble & Reimann, 2016). The presence of open and closed EECs has been observed in both mammals and fish (Rombout, Lamers, & Hanstede, 1978). Previously, it was believed that the open and closed EECs were two differentiated EEC types that perhaps had different physiological functions (Gribble & Reimann, 2016). The open EECs were thought to sense and respond to luminal stimulation while although less clear the closed EECs were thought to respond to hormonal and neuronal stimulation from the basolateral side. However, our data reveal for the first time that EECs can convert from an open to a closed state. This indicates that EECs possess plasticity to actively prune their apical extension. The pruning of cellular process can be observed extensively in neurons. Studies from multiple organisms revealed that sensory neurons can eliminate their dendrites and axon during development and in response to injury through active pruning (Kanamori et al., 2013; Nikolaev, McLaughlin, O’Leary, & Tessier-Lavigne, 2009; Sagasti, Guido, Raible, & Schier, 2005; Williams, Kondo, Krzyzanowska, Hiromi, & Truman, 2006; Yu & Schuldiner, 2014). This process includes focal disruption of the microtubule cytoskeleton, followed by thinning of the disrupted region, severing and fragmentation and retraction in proximal stumps after severing events (Williams & Truman, 2005). In our system, the thinning and fragmentation in the EEC apical extension was also observed. It is well known that EECs possess many neuron-like features including neurotransmitters, neurofilaments, and synaptic proteins (Bohorquez et al., 2015). Whether EECs adopt the same mechanisms as neurons to prune their cellular processes in response to nutritional and microbial signals is interesting and requires future study.
The effects of diet and microbes on EEC silencing
In this study, we have shown that both diet and microbes play important roles in inducing EEC silencing. Dietary manipulations and changes in gut microbiota have been shown to affect EEC cell number and GI hormone gene expression in mice and zebrafish (Arora et al., 2018; Rawls, Samuel, & Gordon, 2004; Richards et al., 2016; Troll et al., 2018). However, it remains unclear from previous studies how environmental factors like diet and gut microbiota affect EEC function. We found that while the presence of microbiota did not influence EEC nutrient sensing under basal conditions, microbiota played an essential role in mediating high fat diet induced EEC silencing as germ free EECs were resistant to high fat diet induced silencing. We speculate that EEC silencing may temporarily attenuate the host’s ability to accurately sense ingested nutrients and thereby control energy homeostasis. Our finding that gut microbiota play an essential role in high fat diet induced EEC silencing may provide a new mechanistic inroad for understanding the effects of gut microbiota in diet induced metabolic diseases including obesity and insulin resistance (Backhed, Manchester, Semenkovich, & Gordon, 2007; Rabot et al., 2010).
There are several nonexclusive ways by which specific gut microbiota members such as Acinetobacter sp. might affect EECs in the setting of a high fat diet. First, microbiota could affect EEC development to increase production of EEC subtypes that are relatively susceptible to diet-induced EEC silencing. Previous transcriptome analysis in the ileum of GLP-1 secreting EECs showed that microbiota colonization increased transcript levels of genes associated with synaptic cycling, ER stress response and cell polarity was reduced in germ free mice (Arora et al., 2018). This suggests that EECs in colonized animals may be more prone to diet-induced ER stress and morphological changes including those associated with EEC silencing.
Second, high fat meal conditions induce bacterial overgrowth and alter the selective pressures within the gut microbial community to allow for enrichment and depletion of specific bacterial taxa. Such changes in microbial density and community composition may then acutely affect EEC physiology. Indeed, we found that high fat feeding altered the relative abundance of several bacterial taxa in the zebrafish gut and media, including a 100-fold increase for members of the Acinetobacter genus. Strikingly, a representative Acinetobacter sp. was the only strain we identified that was sufficient to mediate high fat induced alterations in EEC morphology. We speculated that bacterial overgrowth may also result in increased presentation of microbe-associated molecular patterns which could then hyper-activate TLR or other microbe-sensing pathways that could lead to EEC functional changes. However, our data from myd88 mutant zebrafish suggest that at least the Myd88-dependent microbial sensing pathways are not required for high fat induced EEC silencing. As described below, identification of the specific signals produced by Acinetobacter sp. and other bacteria that facilitate EEC silencing remain an important research goal.
Third, gut microbiota might affect EEC function through promoting lipid digestion and absorption. This is supported by our observations that blocking fat digestion and subsequent lipid absorption in enterocytes through orlistat treatment inhibited high fat diet induced EEC silencing. EEC function may be directly influenced by the products of lipolysis such as free fatty acids (Edfalk, Steneberg, & Edlund, 2008; Hirasawa et al., 2005; Katsuma et al., 2005). However, in our experiments, palmitate treatment was only sufficient to reproduce a portion of the EEC silencing phenotype (i.e. loss of nutrient sensitivity), suggesting that additional undefined signals from fat digestion in the intestine are required to fully induce EEC silencing. In the intestinal epithelium, EECs are surrounded by enterocytes and these two cell types exhibit complex bi-directional communication (Hein, Baker, Hsieh, Farr, & Adeli, 2013; Hsieh et al., 2009; Okawa et al., 2009; Shimotoyodome et al., 2009). Following ingestion of a complex high fat meal, free fatty acids and glycerol liberated from triglyceride digestion are taken up by enterocytes and assembled into lipid droplets and chylomicrons (Phan & Tso, 2001). The subsequent enlargement of enterocytes from lipid droplet accumulation may exert mechanical pressure on EECs that then induces EECs adaption through pruning of their apical protrusions. Besides mechanical pressure, lipoproteins and free fatty acids released from enterocytes may act on EECs basolaterally to alter their function (Chandra et al., 2013; Okawa et al., 2009; Shimotoyodome et al., 2009). Previous studies have shown that lipid digestion and absorption is impaired in germ free animals and enterocytes in germ free condition exhibit reduced lipid droplet accumulation (Martinez-Guryn et al., 2018; Semova et al., 2012). Therefore, reduced mechanical pressure or secondary signaling molecules from enterocytes in the germ free condition may lead to the resistance of EECs to high fat induced silencing. On the other hand, gut microbiota may promote EECs silencing by facilitating intestinal lipid digestion and absorption. Acinetobacter was the most highly enriched genus in the larval zebrafish intestine following high fat feeding in this study and was also enriched in adult zebrafish gut following a chronic high fat diet (Wong et al., 2015). Further, we identified a representative member of this genus that is sufficient to mediate EEC silencing under high fat diet conditions. However, the molecular mechanisms by which Acinetobacter spp. evoke this host response remains unknown. Studies suggest that A. baumannii, a closely related oportunitistic pathogen, can signal to host epithelial cells through secreted outer membrane vesicles (OMVs) and activation of downstream inflammatory pathways (Jha, Ghosh, Gautam, Malhotra, & Ray, 2017; Jin et al., 2011; Jun et al., 2013; March et al., 2010). In addition to OMVs, Acinetobacter strains are known to secrete phospholipase that can affect host cell membrane stability and interfere with host signaling (Lee et al., 2017; Songer, 1997). Members of the Acinetobacter genus are also known to possess potent oil degrading and lipolytic activities (Lal & Khanna, 1996; Snellman & Colwell, 2004). Moreover, species from Acinetobacter genus have the ability to produce emulsifiers which might enhance lipid digestion (Navon-Venezia et al., 1995; Toren, Navon-Venezia, Ron, & Rosenberg, 2001; Walzer, Rosenberg, & Ron, 2006). Interestingly, Acinetobacter spp. in the human gut are positively associated with plasma TG and total- and LDL-cholesterol (Graessler et al., 2013), and Acinetobacter spp. are also enriched in the crypts of the small intestine and colon in mammals (Mao, Zhang, Liu, & Zhu, 2015; Pedron et al., 2012; Saffarian et al., 2017). Therefore, it will be intertesting to determine whether Acinetobacter spp. also modulate EEC function in mammals under high fat diet conditions. Finally, considering the small scale of our monoassocation screen, we anticipate that additional members of the gut microbiota in zebrafish and other animals will be found to also affect EEC silencing and other aspects of EEC biology.
MATERIALS AND METHODS
Zebrafish strains and husbandry
All zebrafish experiments conformed to the US Public Health Service Policy on Humane Care and Use of Laboratory Animals, using protocol number A115-16-05 approved by the Institutional Animal Care and Use Committee of Duke University. Conventionally-reared adult zebrafish were reared and maintained on a recirculating aquaculture system using established methods (Murdoch et al., 2019). For experiments involving conventionally-raised zebrafish larvae, adults were bred naturally in system water and fertilized eggs were transferred to 100mm petri dishes containing ~25mL of egg water at approximately 6 hours post-fertilization. The resulting larvae were raised under a 14h light/10h dark cycle in an air incubator at 28°C at a density of 2 larvae/ml water. To ensure consistent microbiota colonization, 10mL filtered system water (5μm filter, SLSV025LS, Millipore) was added into 3 dpf zebrafish larva that were raised in 25mL egg water. All the experiments performed in this study ended at 6 dpf unless specifically indicated. The strains used in this study are listed in Table S1. All lines were maintained on a EKW background.
Gateway Tol2 cloning approach was used to generate neurod1:lifActin-EGFP and neurod1:Gal4 plasmid (Kawakami, 2007; Kwan et al., 2007). The 5kb pDONR-neurod1 P5E promoter was previously reported (McGraw et al., 2012) and generously donated by Dr. Hillary McGraw. The PME-lifActin-EGFP (Riedl et al., 2008) and the PME-Gal4-vp16 plasmids (Kwan et al., 2007) were also previously reported. pDONR-neurod1 P5E and PME-lifActin-EGFP was cloned into pDestTol2pA2 through an LR Clonase reaction (ThermoFisher,11791). Similarly, pDONR-neurod1 P5E and PME-Gal4-vp16 was cloned into pDestTol2CG2 containing a cmlc2:EGFP marker. The final plasmid was sequenced and injected into the wild-type Ekkwill (EKW) zebrafish strain and the F2 generation of allele Tg(neurod1:lifActin-EGFP)rdu70 and Tg(neurod1:Gal4; cmlcl2:EGFP)rdu71 was used for this study.
The construct used to generate the TgBAC(gata5:lifActin-EGFP) line was made by inserting lifeact-GFP at the gata5 ATG in the BAC clone DKEYP-73A2 using BAC recombineering as previously described (PMID: 12618378). The BAC was then linearized using I-SceI and injected to generate transgenic lines. Allele TgBAC(gata5:lifActin-EGFP)pd1007 was selected for further analysis. The construct used to generate the TgBAC(cd36-RFP) lines was made by inserting link-RFP before the cd36 stop codon in the BAC clone DKEY-27K7 using the same BAC recombineering as previously described (Navis et al PMID 23487313). Then, Tol2 sites were recombined into the BAC and the resulting construct was injected with transposase mRNA (cite Kawakami here) to generate the transgenic lines. Allele TgBAC(cd36-RFP)pd1203 was selected for further analysis.
Gnotobiotic zebrafish husbandry
For experiments involving gnotobiotic zebrafish, we used our established methods to generate germ-free zebrafish using natural breeding (Pham et al., 2008) with the following exception: Gnotobiotic Zebrafish Medium (GZM) with antibiotics (AB-GZM) was supplemented with 50 μg/ml gentamycin (Sigma, G1264). Germ free zebrafish eggs were maintained in cell culture flasks with GZM at a density of 1 larvae/ml. From 3 dpf to 6 dpf, 60% daily media change and ZM000 (ZM Ltd.) feeding were performed as described (Pham et al., 2008).
To generate conventionalized zebrafish, 15 mL filtered system water (5μm filter, SLSV025LS, Millipore, final concentration of system water ~30%) was inoculated to flasks containing germ-free zebrafish in GZM at 3 dpf when the zebrafish normally hatch from their protective chorions. The same feeding and media change protocol was followed as for germ free zebrafish. Microbial colonization density was determined via Colony Forming Unit (CFU) analysis. To analyze the effect of high fat feeding on intestinal bacteria colonization, dissected digestive tracts were dissected and pooled (5 guts/pool) into 1mL sterile phosphate buffered saline (PBS) which was then mechanically disassociated using a Tissue-Tearor (BioSpec Products, 985370). 100 µL of serially diluted solution was then spotted on a Tryptic soy agar (TSA) plate and cultured overnight at 30°C under aerobic conditions.
To generate mono-associated zebrafish, a single bacterial strain was inoculated into each flask containing 3dpf germ-free zebrafish. The respective bacterial stock was streaked on a TSA plate and cultured at 28°C overnight under aerobic conditions. A single colony was picked and cultured in 5mL Tryptic soy broth media shaking at 30°C for 16 hours under aerobic conditions. 250 µL bacterial culture was pelleted and washed 3 times with sterile GZM and inoculated into flasks containing germ-free zebrafish. OD600 and CFU measurements were performed in each mono-associated culture. The final innoculation density in GZM was 108-109 CFU/mL. The colonization efficiency was determined at 6 dpf by CFU analysis from dissected zebrafish intestines as described above.
EEC response assay and image analysis
This assay was performed in Tg(neurod1:Gcamp6f) 6 dpf zebrafish larvae. Unanesthetized zebrafish larvae were gently moved into 35mm petri dishes that contained 500µL 3% methylcellulose. Excess water was removed with a 200µL pipettor. Zebrafish larvae were gently positioned horizontal to the bottom of the petri dish right side up carefully avoiding touching the abdominal region and moved onto an upright fluorescence microscope (Leica M205 FA microscope equipped with a Leica DFC 365FX camera). The zebrafish larvae were allowed to recover in that position for 2 minutes. One hundred µL of test agent was pipetted directly in front of the mouth region without making direct contact with the animal. Images were recorded every 10 seconds. For fatty acid stimulation, 30 frames (5mins) were recorded. For glucose stimulation, 60 frames (10mins) were recorded. The Gcamp6f fluorescence was recorded with the EGFP filter. The following stimulants were used in this study: palmitic acid/linoleate/dodecanoate (1.6mM), butyrate (2mM), glucose (500mM), fructose (500mM), galactose (500mM), cysteine (10mM). Since palmitic acid/linoleate/dodecanoate was not water soluble by itself, 1.6% BSA was used as a carrier to facilitate solubility. Solutions were filtered with 0.22µm filter.
Image processing and analysis was performed using FIJI software. The time-lapse fluorescent images of zebrafish EEC response to nutrient stimulation were first aligned to correct for experimental drift using the plugin “align slices in stack.” Normalized correlation coefficient matching method and bilinear interpolation method for subpixel translation was used for aligning slices (Tseng et al., 2012). The plugin “rolling ball background subtraction” with the rolling ball radius=10 pixels was used to remove the large spatial variation of background intensities. The Gcamp6f fluorescence intensity in the proximal intestinal region was then calculated for each time point. The ratio of maximum fluorescence (Fmax) and the initial fluorescence (F0) was used to measure EEC calcium responsiveness.
High fat feeding
The high fat feeding regimen was performed in 6 dpf zebrafish larvae using methods previously described (Semova et al., 2012). ~25 zebrafish larvae were transferred into 6 well plates and 5mL egg water (for gnotobiotic studies, GZM was used). Replicates were performed in three wells for each treatment group in each experiment. Chicken eggs were obtained from a local grocery store from which 1mL chicken egg yolk was transferred into a 50mL tube containing 15mL egg water (for gnotobiotic studies, sterile GZM was used). Solutions were sonicated (Branson Sonifier, output control 5, Duty cycle 50%) to form a 6.25% egg yolk emulsion. 4 mL water from each well was removed and replenished with 4mL egg yolk. 4 mL egg water emulsion was used to replenish the control group. Zebrafish larvae were incubated at 28°C for the indicated time. The high fat meal was administered between 10am - 12pm to minimize circadian influences.
Chemical treatment
To block Sglt1, phloridzin (0.15mM, Sigma P3449) was used to pretreat zebrafish for 3 hours prior to glucose stimulation, and 0.15mM phloridzin was co-administered with the glucose stimulant solution. To induce ER stress, thapsigargin (0.75µM, Sigma T9033) and brefeldin A (9µM, Sigma B6542) were added to egg water and zebrafish were treated for 10 hours prior to performing the EEC activity assay. To block high fat meal induced EEC silencing, sodium tauroursodeoxycholic acid (TUDCA; 0.5mM, T0266) or orlistat (0.1mM, Sigma O4139) were added to the high fat meal solution and zebrafish were treated for the indicated time.
Quantitative RT-PCR
The quantitative real-time PCR was performed as described previously (Murdoch et al., 2019). In brief, 20 zebrafish larvae digestive tracts were dissected and pooled into 1mL TRIzol (ThermoFisher, 15596026). mRNA was then isolated with isopropanol precipitation and washed with 70% EtOH. 500ng mRNA was used for cDNA synthesis using the iScript kit (Bio-Rad, 1708891). Quantitative PCR was performed in triplicate 25 μl reactions using 2X SYBR Green SuperMix (PerfeCTa, Hi Rox, Quanta Biosciences, 95055) run on an ABI Step One Plus qPCR instrument using gene specific primers (Table S1). Data were analyzed with the ΔΔCt method. 18S was used as a housekeeping gene to normalize gene expression.
16S rRNA gene sequencing
Wild-type adult EKW zebrafish were bred and clutches of eggs from three distinct breeding pairs were collected, pooled, derived into GF conditions using our standard protocol (Pham et al., 2008), then split into three replicate flasks with 30ml GZM as described above. At 3 dpf 12.5ml 5μm-filtered system water was inoculated into each flask per our standard conventionalization method. ZM000 feeding and water changes were performed daily from 4 dpf to 5 dpf. At 6 dpf, zebrafish larvae from each flask were divided evenly into a control and a high fat fed group. High fat feeding was performed as described above for 6 hours. Then 1 ml water samples were collected from each flask and snap frozen on dry ice/EtOH bath. For intestinal samples, individual digestive tracts from 6dpf zebrafish were dissected and flash frozen (3-4 larvae/flask, 3 flasks/condition). All samples were stored in −80°C for subsequence DNA extraction.
The Duke Microbiome Shared Resource (MSR) extracted bacterial DNA from gut and water samples using a MagAttract PowerSoil DNA EP Kit (Qiagen, 27100-4-EP) as described previously (Murdoch et al., 2019). Sample DNA concentration was assessed using a Qubit dsDNA HS assay kit (ThermoFisher, Q32854) and a PerkinElmer Victor plate reader. Bacterial community composition in isolated DNA samples was characterized by amplification of the V4 variable region of the 16S rRNA gene by polymerase chain reaction using the forward primer 515 and reverse primer 806 following the Earth Microbiome Project protocol (http://www.earthmicrobiome.org/). These primers (515F and 806R) carry unique barcodes that allow for multiplexed sequencing. Equimolar 16S rRNA PCR products from all samples were quantified and pooled prior to sequencing. Sequencing was performed by the Duke Sequencing and Genomic Technologies shared resource on an Illumina MiSeq instrument configured for 150 base-pair paired-end sequencing runs. Sequence data are deposited at SRA under Bioproject accession number PRJNA532723.
Subsequent data analysis was conducted in QIIME2 (Caporaso et al., 2010)(https://peerj.com/preprints/27295/). Paired reads were demultiplexed with qiime demux emp-paired, and denoised with qiime dada2 denoise-paired (Callahan et al., 2016). Taxonomy was assigned with qiime feature-classifier classify-sklearn (Scikit-learn: Machine Learning in Python. Journal of Machine Learning Research), using a naive Bayesian classifier, trained against the 99% clustered 16S reference sequence set of SILVA, v. 1.19 (Quast et al., 2013). A basic statistical diversity analysis was performed, using qiime diversity core-metrics-phylogenetic, including alpha- and beta-diversity, as well as relative taxa abundances in sample groups. The determined relative taxa abundances were further analyzed with LEfSe (Linear discriminant analysis effect size) (Segata et al., 2011), to identify differential biomarkers in sample groups.
Immunofluorescence Staining and Imaging
The whole mount immunofluorescence staining was performed as previously described (Ye et al., 2015). In brief, ice cold 2.5% formalin was used to fix zebrafish larvae overnight at 4°C. The samples were then washed with PT solution (PBS+0.75%Triton-100). The skin and remaining yolk was then removed using forceps under a dissecting microscope. The deyolked samples were then permeabilized with methanol for more than 2 hours at −20°C. The samples were then blocked with 4% BSA at room temperature for more than 1 hour. The primary antibody was diluted in PT solution and incubated at 4°C for more than 24 hours. Following primary antibody incubation, the samples were washed with PT solution and incubated overnight with secondary antibody with Hoechst 33342 for DNA staining. The imaging process was performed with a Zeiss 780 inverted confocal and Zeiss 710 inverted confocal microscopes with the 40× oil lenses. The following primary antibodies were used in this study: rabbit anti PYY (custom, aa4-21, 1:100 dilution) (Chandra, Hiniker, Kuo, Nussbaum, & Liddle, 2017), goat anti-CCK (Santa Cruz SC-21617, 1:100 dilution), rabbit anti-Sglt1 (Abcam ab14686, 1:100 dilution). The secondary antibodies used in this study are from Alexa Fluor Invitrogen. All the secondary antibodies were used at a dilution of 1:250.
To quantify EEC morphology score, chick anti-GFP (Aves GFP1010, 1:500 dilution) and rabbit anti-mcherry (TAKARA 632496, 1:250 dilution) antibodies were used in the fixed Tg(gata5:lifActin-EGFP);Tg(neurod1:TagRFP) samples to perform immunofluorescence staining. The region following intestine bulb were imaged with a Zeiss 780 inverted confocal and Zeiss 710 inverted confocal microscopes with the 40× oil lenses. Images were processed with FIJI. The gata5:lifActin-EGFP only stains the apical brush border of the intestine. Total EECs number was assessed via counting RFP+ cell bodies. The number of EECs with intact apical protrusion was assessed via counting the number of RFP+ cells with attachment to GFP staining brush border. EEC morphology for each sample were quantified as ratio between EECs with intact apical protrusion and total EEC number.
For live imaging experiments, zebrafish larvae were anesthetized with Tricane and mounted in 1% low melting agarose in 35mm petri dishes. The live imaging was recorded with Zeiss 780 upright confocal with a 20× water lens.
Statistical Analyses
The appropriate sample size for each experiment was suggested by preliminary experiments evaluating variance and effects. Using significance level of 0.05 and power of 80%, a biological replicate sample number 10 was suggested for EEC calcium response analysis and a biological replicate sample number 13 was suggested for EEC morphology analysis. For each experiment, wildtype or indicated transgenic zebrafish embryos were randomly allocated to test groups prior to treatment. In some EEC calcium response experiments, less than 10 biological replicate samples were used due to technical limitations associated with live sample imaging. In EEC morphology analysis, each experiment contained 8-15 biological replicates or individual fish samples. Individual data points, mean and standard deviation are plotted in each figure.
The raw data points in each figure are represented as solid dots. The data was analyzed using GraphPad Prism 7 software. For experiments comparing just two differentially treated populations, a Student’s t-test with equal variance assumptions was used. For experiments measuring a single variable with multiple treatment groups, a single factor ANOVA with post hoc means testing (Tukey) was utilized. Statistical evaluation for each figure was marked * P<0.05, ** P<0.01, *** P<0.001, **** P<0.0001 or ns (no significant difference, P>0.05). Statistical analyses for 16S rRNA gene sequencing data can be found in in the corresponding Methods section above.
Competing interests
The authors declare no competing interests.
Acknowledgements
We thank Dr. Hillary McGraw for the 5kb pDONR-neurod1 P5E plasmid, Dr. Joachim Berger for the pMElifActin-EGFP plasmid, Dr. David Raible for the Tg(neurod1:TagRFP) transgenic fish and Dr. Clair Wyart for the Tg(neurod1:Gcamp6f) transgenic fish. We also thank the Duke Light Microscopy Core Facility for equipment access and technical support, the Duke Zebrafish Core Facility for assisting zebrafish husbandry and the Duke Microbiome Shared Resource for 16S rRNA gene sequencing. This work was supported by grants from the National Institutes of Health R01-DK093399 (to J.F.R. and R.A.L.), R01 DK109368 (to R.A.L.), and R01-DK081426 (to J.F.R.); the Department of Veterans Affairs I01BX002230 (to R.A.L.); and an Innovation Grant from the Pew Charitable Trusts (to J.F.R. and R.A.L.). L.Y. was supported by the Digestive Disease and Nutrition Training Program at Duke University (T32-DK007568).