DSS-induced damage to basement membranes is repaired by matrix replacement and crosslinking

Basement membranes are omnipresent extracellular structures in multicellular animals. They are strong, thin sheets of extracellular matrix underlying epithelia, enveloping muscles and organs, and separating tissue layers. Basement membranes function as mechanical scaffolds to distribute cellular and tissue-level forces, and when they cannot distribute these forces, diseases such as muscular dystrophy and skin blistering result (Nyström et al., 2017). In addition to mechanical roles, basement membranes signal to cells and can tether other signaling molecules, directly and indirectly modulating cell differentiation, survival, migration and polarity of epithelial cells (Kleinman and Martin, 2005; Li et al., 2017; Wang et al., 2008; Bunt et al., 2010). Evolutionarily ancient structures, basement membranes are conserved from hydra to humans (Fidler et al., 2017). They are composed of four main components: laminin, collagen IV, perlecan, and nidogen. Super-resolution imaging of a glomerular basement membrane in vivo indicates that it has a laminar structure with components spatially segregated into layers (Suleiman et al., 2013), a finding consistent with in vitro observations that laminin and collagen IV can polymerize independently to form sheet-like polymers (Yurchenco and Furthmayr, 1984; Yurchenco et al., 1985, 1992).

The mechanical strength of basement membranes comes mainly from collagen IV, which assembles non-covalently after undergoing a conformational change mediated by high concentrations of extracellular chloride (Cummings et al., 2016). After extracellular assembly, the collagen IV network is reinforced by covalent crosslinking, the extent of which determines the stiffness of the basement membrane (Bhave et al., 2017). Crosslinking can occur at three distinct sites on a triple-helical collagen IV molecule: at the N-terminal 7S domain (Risteli et al., 1980; Langeveld et al., 1991), at lateral sites along the triple-helical domain (Yurchenco and Furthmayr, 1984), and at the C-terminal NC1 domain (Vanacore et al., 2009). The best understood of these crosslinks is between NC1 domains, which are variably crosslinked head-to-head by sulfilimine bonds, covalent linkages catalyzed by the enzyme peroxidasin using bromide as a cofactor (Bhave et al., 2012; McCall et al., 2014). There are two possible sulfilimine crosslink sites within each NC1–NC1 dimer, and up to six per NC1 hexamer that joins two triple-helical structures. However, only approximately two to four of these six sites are occupied per hexamer on average (Bhave et al., 2012; McCall et al., 2014), and the sulfilimine occupancy ratio appears to be a tissue-specific property. Like basement membranes themselves, sulfilimine crosslinks and the peroxidasin enzyme are both conserved throughout the animal kingdom (Fidler et al., 2014).

Despite their fundamental importance, questions remain about how basement membranes are assembled, and even less is known about how they are repaired after damage. As with all biological systems, basement membranes routinely get damaged. Traumatic damage like skin wounds are one obvious context for basement membrane repair. Damage and repair are also endemic processes; for example, leukocytes migrate through basement membranes and the resulting lesions are repaired to maintain mechanical integrity (Huber and Weiss, 1989). Damaged basement membrane may also play a causative role in the progression of diseases such as asthma, kidney failure and diabetes, in which altered basement membrane structures may be caused by faulty repair programs (Flood-Page et al., 2003; Sugimoto et al., 2006; Tsilibary, 2003). Further, there are diseases of the basement membrane itself (e.g. Alport syndrome, Goodpasture's syndrome, thin basement membrane disease), and understanding repair is central to treating these conditions.

There is a paucity of information on basement membrane repair, stemming from a lack of models of basement membrane damage. Questions about de novo basement membrane assembly have been addressed in vitro, in cell culture and in embryos. For example, the self-assembling nature of basement membrane components was originally discovered in vitro with proteins purified from EHS sarcomas (the origin of matrigel) (Yurchenco and Furthmayr, 1984; Kleinman et al., 1986, 1982). Despite their power, these systems may not be suitable for analyzing repair because in vitro analysis removes cells from the architectural and mechanical environment of the tissue, both of which are likely important for matrix repair. More recently, biochemical studies on matrix assembly (Fox et al., 1991; Ries et al., 2001; Takagi et al., 2003; Hopf et al., 1999) have been complemented by genetic analysis of animals with mutant basement membrane proteins, and these analyses were critical in identifying an order of assembly: laminin first, as collagen IV and nidogen each require laminin, and perlecan requires collagen IV (Pöschl et al., 2004; Urbano et al., 2009; Pastor-Pareja and Xu, 2011; Ramos-Lewis et al., 2018; Wolfstetter et al., 2019). Whole-animal mutants for basement membrane proteins are embryonic lethal, how­ever, and the analysis of repair requires conditional mutants that can be induced temporally, after assembly has been completed. We recently published a study analyzing repair of basement membrane after an epidermal pinch wound in which the basement membrane was torn, generating a region devoid of basement membrane ∼100 µm in diameter, and found that it repaired with a scar, incorporating the known basement membrane proteins slightly differently from the order of incorporation for assembly (Ramos-Lewis et al., 2018). This assay mimics trauma wounds, with new matrix filling the breach as the cell layer migrates in; however, it is not suitable for biochemical analysis or genetic screening because the wounds are painstaking to administer and are variable and small in size. Thus, we sought a complementary assay for analyzing basement membrane repair.

Previously, we reported that when either Peroxidasin or its cofactor bromide is limiting during Drosophila development, basement membranes become measurably expanded and the muscles they support become deformed (McCall et al., 2014). This phenotype was first evident around the larval midgut, suggesting that this tissue has a heightened requirement for sulfilimine crosslinking. This loss-of-crosslinking phenotype appeared surprisingly similar to published images of the deformation caused by feeding flies the intestinal irritant dextran sodium sulfate (DSS) (Amcheslavsky et al., 2009), a polyanionic derivative of the polysaccharide dextran. Stemming from this observed similarity, this study investigates whether and how DSS damages the basement membrane, and whether and how crosslinking is important for basement membrane repair. In this report, we introduce a new DSS-based experimental model to probe basement membrane damage and repair, one which is reproducible and suited to microscopy. Using this model, we determine that replacement and crosslinking are essential processes of basement membrane repair.

We focused on DSS as a potential basement membrane damaging agent. DSS is a 36–50 kDa negatively charged derivative of the carbohydrate polymer dextran, and administering DSS to mice in drinking water has long been used to induce a condition like ulcerative colitis. Because of its use in damaging the mouse intestine, DSS has also been utilized in adult Drosophila to damage the fly gut epithelium (Amcheslavsky et al., 2009; Ren et al., 2010, 2013; Karpowicz et al., 2010; Cordero et al., 2012; Tian and Jiang, 2014; You et al., 2014; Tian et al., 2015). In the first report of treating flies with DSS (Amcheslavsky et al., 2009), it was noted that DSS appeared to alter basement membrane around the gut. The basement membrane is a sheet-like extracellular matrix that gets its mechanical strength from covalent crosslinking of the Collagen IV polymer (Bhave et al., 2017). The DSS-induced basement membrane phenotype reported by Amcheslavsky et al. (2009) appeared similar to a phenotype we reported in larvae when we mutated or inhibited Peroxidasin (Pxn), a Collagen IV crosslinking enzyme (McCall et al., 2014). Thus, it seemed possible that DSS was interfering with Collagen IV function.

To examine these similar phenotypes, we compared the treatments of DSS-feeding and ubiquitous Pxn knockdown (using TubP-Gal4) in the midguts of adult females, which are larger than the midguts of males. The Drosophila gut comprises an epithelial monolayer, exposed to the lumen on its apical side and abutting the basement membrane on its basal side. Several basement membranes are expected to lie in close apposition outside the midgut. The epithelial basement membrane lies between the epithelial cells and the underlying visceral muscles responsible for peristalsis, which run circumferentially and longitudinally along the gut. Typically, muscles are wrapped in basement membrane (Yurchenco, 2011). Further, basement membranes separate organs from the body cavity and hemolymph. Fig. 1A shows a schematic of these tissues and their associated basement membranes, based on our imaging in this study.

To visualize the basement membrane, we used viking (vkg)-GFP⁴⁵⁴ (Vkg–GFP hereafter), which has a fully-functional GFP-trapped Collagen IV α2 protein transcribed from the endogenous genomic locus. In these flies, Collagen IV can be easily visualized by GFP fluorescence with standard epifluorescence microscopy (Fig. 1B). We fed flies 3% DSS in a 5% sucrose solution over the course of 2 days and confirmed the previous findings that the tissue appeared altered compared to controls fed only sucrose (Fig. 1B,C; Fig. S1). On closer inspection, we found that the obvious DSS-induced morphological changes were caused by the peristalsis muscles, which appeared as dark areas surrounded by basement membrane. In optical cross-sections along the long axis of the midgut after DSS treatment, the circumferential muscles appeared contracted inward, protruding toward the epithelia and lumen, giving a rounder appearance in cross-section that was also evident by phalloidin staining of actin (Fig. 1E–F′). A similar rounding of circumferential muscles was observed when Pxn was knocked down ubiquitously in adults (Fig. 1D). This rounded muscle phenotype was quantified by measuring the aspect ratio of the circumferential muscles in cross-section (Fig. 1D′), which was significantly different from controls (Fig. 1G). The altered appearance of the basement membrane and contracted morphology of the muscles was also evident when we visualized the basement membrane components Laminin (LanB1–GFP; Fig. 1H,I) or Perlecan (Trol–GFP; Fig. 1J,K).

As demonstrated by Pxn knockdown, defects in muscle morphology can result from the loss of basement membrane function. A mechanistic relationship between basement membrane function and muscle morphology and function has arisen from muscular dystrophy studies. Normally, muscle contraction force is resisted by the basement membrane, as the contraction force is transmitted by covalent linkages connecting the contractile machinery to the basement membrane (schematic in Fig. 2A). The genes that are defective in muscular dystrophy disease encode proteins comprising components of these mechanical linkages, including the basement membrane protein laminin (Mercuri and Muntoni, 2013). In muscular dystrophy patients, these linkages are disrupted so that muscle-generated contractile forces pull on the unsupported plasma membrane, resulting in membrane tears, calcium entry, muscle hypercontraction and eventual muscle damage (Nyström et al., 2017; Vila et al., 2017). The visceral muscles in Drosophila are striated and multinucleate, more similar to mammalian skeletal muscle than mammalian visceral smooth muscles. Striations are caused by repeating sarcomeres, units of actomyosin contractile machinery, which have multi-protein covalent linkages extending from the Z-bands to the basement membrane (Maartens and Brown, 2015). Previous work has shown that the reduction of Laminin in Drosophila ovarian muscles decreases sarcomere size, recapitulating a muscular dystrophy phenotype (Andersen and Horne-Badovinac, 2016). As in ovarian muscles, Drosophila gut muscles decreased sarcomere size by about 10% upon reduction of Laminin (LanB1-RNAi driven by TubP-Gal4, Gal80^ts; Fig. 2B–D). We reasoned that if DSS damaged muscles indirectly via the basement membrane, DSS should also decrease sarcomere size. In the DSS-fed flies, an even more pronounced decrease in sarcomere size was observed (∼30%; Fig. 2E–G), although incomplete knockdown of Laminin may have contributed to the bigger effect of DSS. To test whether sarcomere size was reduced by active contraction, we treated guts with relaxation buffer ex vivo. Although treatment with relaxation buffer did somewhat increase sarcomere size of both DSS- and control-fed flies, the DSS-treated sarcomeres remained significantly shorter than controls (Fig. S2). Thus the decrease in sarcomere size in DSS-treated guts was not caused by DSS-induced contraction. These results are consistent with the interpretation that DSS damages the basement membrane of peristalsis muscles, leading to muscle damage and altered muscle morphology.

Basement membranes have traditionally been imaged by electron microscopy, so we used transmission electron microscopy (TEM) to image the gut in cross sections perpendicular to the long axis of the gut. In control samples, basement membrane was observed on the basal surface of the enterocytes and also on the outer surface of the muscles; between these layers and in the inter-muscle regions, the organization of the extracellular matrix was unclear (Fig. 3A,A′; Fig. S3). After DSS feeding, the most obvious change to the tissue was in the peristalsis muscles, which were irregular and shredded (Fig. 3B,B′, yellow arrows), consistent with our epifluorescence analysis. Importantly, the basement membrane itself appeared thicker after DSS feeding. Measuring the expansion of the basement membrane on the basal surface of the enterocytes (pseudo-colored green, Fig. 3A′–B′), the basement membrane was 140±60 nm (mean±s.e.m.) thick in controls, whereas after DSS treatment it was 330±250 nm thick (Fig. 3C). Thus, by TEM, the gut peristalsis muscle basement membrane appeared more than twice as expanded in flies fed DSS than in control flies.

We were concerned about possible dehydration artifacts associated with fixing samples for EM, and we were also unsure how to interpret the uncharacterized matrix between the apparent basement membrane and the muscles. Thus, we repeated our experiments using super-resolution structured illumination microscopy (SIM), which gave ∼2× increase in resolution compared to diffraction-limited techniques such as laser-scanning confocal (Gustafsson et al., 2008) with two advantages over TEM: a standard formaldehyde-based immunohistochemistry fixation protocol was utilized, and the basement membrane was identified with fluorescent Vkg–GFP labelling rather than as an unlabeled electron-dense structure. By SIM imaging, the control basement membrane was 260±50 nm thick, whereas DSS-treated basement membrane expanded to 390±100 nm thick (Fig. 3D–F). Thus, two independent fixation and visualization techniques determined that basement membranes expanded after DSS treatment.

Basement membrane expansion could indicate an increase in levels of basement membrane proteins, resulting in a thicker structure; alternatively, it could indicate a fracturing or delaminating mechanical failure of the basement membrane sheet. To address whether increased levels of basement membrane proteins were present after DSS feeding, we measured the total fluorescence of the GFP-labeled basement membrane proteins within the basement membrane around the gut. We imaged the fluorescence of Vkg–GFP (Fig. 3G), LanB1–GFP (Fig. 3H) and Trol–GFP (Fig. 3I), all functional GFP fusions expressed from endogenous regulatory sequences. For each, the total fluorescence levels were unchanged after DSS feeding. Independently, we found no difference in gut length or diameter after DSS feeding (Fig. S4). We conclude that the expansion in basement membrane is not caused by an increase in basement membrane deposition; rather, the basement membrane expansion suggests mechanical failure.

To gain insight into how DSS expands the basement membrane, we fed fluorescently-labeled FITC–DSS to flies and visualized its localization in unfixed gut tissue after a 6-h chase (Fig. 4A–F). For controls, we fed flies unconjugated FITC (no DSS treatment), or we fed flies unconjugated FITC and unconjugated DSS, as separate molecules, to control for any stickiness of FITC to DSS-damaged tissues (Fig. 4A–C). Basement membrane was labeled with a functional protein-trapped Trol–RFP, rendering the basement membrane red. Labeled FITC–DSS localized reproducibly to the basement membrane of the damaged guts, whereas no fluorescence was observed in the basement membrane of either control (Fig. 4D–F″). These results were obtained in unfixed tissue, as FITC–DSS washed out during the fixation procedure because it has no amine groups to fix it in place. The localization of FITC–DSS to the basement membrane indicates that DSS is trans­ported from the gut lumen through the entero­cytes (cells that specialize in nutrient transport) to the adjacent basement membrane where it accumulates; indeed, punctae of FITC–DSS were observed within the epithelial layer (Fig. S5I). The basement membrane morphological expansion could be caused by negatively charged DSS creating osmotic pressure, leading to swelling. This pathological swelling of basement membrane could cause mechanical weakening, as implied by the presence of dysmorphic muscles. Because DSS has been used extensively in mouse models, we performed a similar experiment, feeding FITC–DSS to mice and examining its intestinal localization. In mice, FITC–DSS localizes to punctae along the epithelial layer, co-localizing with intercellular junctions, and not to the underlying basement membrane (Fig. S5), indicating that murine enterocytes do not transport DSS across the epithelial barrier as do Drosophila enterocytes.

We wanted to directly test the stiffness of the basement membrane after DSS treatment; however, the gut basement membrane cannot be separated from the peristalsis muscles, which would significantly alter tissue stiffness independently of the basement membrane. As an alternative, we analyzed the Malpighian tubules, part of the Drosophila excretory system. These simple tubules have only a tube-shaped epithelial monolayer surrounded on the outside by basement membrane, without muscle or exoskeleton. We did not expect DSS to diffuse to the Malpighian tubules at high levels after feeding in vivo because it gets trapped around the gut; instead, we treated the Malpighian tubules with DSS ex vivo by soaking them in a DSS solution for 20 min. As observed for guts after DSS feeding, conjugated FITC–DSS accumulates in the basement membrane surrounding the tubule, whereas unconjugated FITC does not, either in the presence or absence of DSS (Fig. 4G–I).

To measure the stiffness of the Malpighian tubule basement membrane directly ex vivo, we used a glass micro-cantilever system to measure the tensile stress-strain response (Fig. 4J). This technique is used to estimate the stiffness of the basement membrane, reported as the elastic modulus (Young's modulus). To evaluate the contribution of the basement membrane to tubule stiffness, we compared the elastic modulus of intact tubules to those that were decellularized by the addition of detergent after dissection. Decellularization did not significantly change the stiffness of the tissue (Fig. 4K), confirming that we were measuring the stiffness of the basement membrane. In contrast, the Malpighian tubule stiffness was significantly decreased after soaking in DSS (Fig. 4L,M), indicating that DSS disrupts the mechanical properties of the basement membrane. We considered the possibility that basement membrane stiffness was altered by DSS only during the time that DSS resided in it, asking whether the stiffness returned immediately upon DSS wash-out. To test this possibility, we used labeled FITC–DSS to determine conditions for washing out DSS after it became lodged in the basement membrane, and then tested Malpighian tubules soaked in DSS and then washed. We found that the elastic modulus of the basement membrane was virtually unchanged after DSS removal (Fig. 4L,M). Thus, ex vivo DSS inflicts irreversible mechanical damage on the basement membrane.

One possibility for how DSS could access the basement membrane in vivo is by damaging the enterocyte barrier, thus exposing the basement membrane to the contents of the gut lumen including DSS. If this were the case, loss of gut epithelial barrier function would precede or be simultaneous with basement membrane damage. Epithelial barrier function can be evaluated using a blue food dye, which spreads throughout the body of the fly when the intestinal barrier is breached, termed a ‘smurf’ phenotype (Rera et al., 2011) (Fig. 5A). To investigate the function and loss of the intestinal epithelial barrier, we fed flies with blue dye for 6 days, with or without DSS, and scored the flies regularly for body color (Fig. 5B,C). Dead flies with blue bodies were excluded from the count, as barrier integrity is always lost on death; dead flies were much more prevalent with DSS feeding than without (Fig. 5D,E; Fig. S6). The majority of smurf (blue) flies were observed after 5 days, with the first instance recorded after 3 days, well after the onset of basement membrane damage, assayed after 2 days of DSS feeding when all samples show the effects. Therefore, DSS damages basement membranes directly, rather than after the loss of epithelial barrier.

Taken together, the above data indicate that on feeding, DSS accumulates in the basement membrane of the gut and mechanically damages it, causing the gut muscles to become dysmorphic. Although this damage is irreversible ex vivo, we asked whether the basement membrane could be repaired after DSS damage was inflicted in vivo. The muscle aspect ratio was surveyed at various times after transferring the flies from DSS treatment back to normal food. Although there was a short-lived reversal of muscle shape within 3 h, lasting recovery was accomplished more slowly, returning to normal 48 h after termination of DSS feeding (Fig. 6A). To confirm that the basement membrane itself had repaired in addition to the muscles, we examined the basement membrane by TEM at 48 h after terminating DSS feeding and found that the basement membrane sheet had returned to its previous undamaged width (Fig. 6B–E). When analyzed by SIM as well as by TEM, the basement membrane of control flies returned to its normal width by 48 h after termination of DSS (Fig. 7G). At 48 h after DSS termination, the levels of basement membrane proteins Vkg, LanB1 and Trol were not significantly different than in undamaged midguts as measured by GFP-fusion protein fluorescence (Fig. S7). Thus, both muscle morphology and direct measurement indicated that basement membrane repaired within two days of DSS treatment withdrawal.

Having established the time course for repair, we analyzed biological requirements for repair. The role of Collagen IV crosslinking in repair was investigated by knocking down the NC1-crosslinking enzyme Pxn ubiquitously in adults using TubP-Gal4, Gal80^ts for 7 days before feeding with DSS for two days, then removing from DSS to allow repair for two more days. When basement membrane morphology was analyzed by SIM, the basement membranes of Pxn-knockdown flies did not recover and remained expanded even 48 h after they were removed from DSS to normal food, in contrast to sibling controls, in which basement membranes repaired (Fig. 7A–G). We next analyzed muscle aspect ratio as a readout of basement membrane repair in these Pxn-knockdown flies. As noted in Fig. 1, even before DSS feeding, the Pxn-knockdown muscles have an increased aspect ratio; this ratio is exacerbated by DSS feeding, and it does not recover after removal of DSS (Fig. 7N); we confirmed the specificity of this phenotype with a second Pxn RNAi line (not shown). As an independent method of assessing the role of crosslinking, we fed flies phloroglucinol (PHG), an irreversible chemical inhibitor of Pxn (Bhave et al., 2012). PHG has two advantages over genetic knockdown: first, there is no pre-treatment because it is immediately effective, eliminating complications from long-term loss of Pxn before DSS feeding; second, it is likely to give a more penetrant phenotype than RNAi-based knockdown, which is usually incomplete. PHG was administered with the DSS or vehicle, and it caused a modest increase in muscle aspect ratio even without DSS during the course of the 4-day experiment (2 days treatment, 2 days recovery); when combined with DSS, the muscle aspect ratio became severely increased, and this ratio did not recover even 48 h after DSS withdrawal when PHG was maintained in the food (Fig. 7H–N). We conclude that crosslinking is required for repairing the basement membrane after DSS-induced mechanical damage.

We considered two mechanisms by which collagen crosslinking could be required for repair. First, crosslinking might be increased to stabilize the pre-existing collagen within damaged basement membranes, akin to stapling broken fragments together. Theoretically, up to six sulfilimine crosslinks can bridge every NC1 hexamer, but an average of only 2–4 crosslinks per hexamer is detected in bulk fly and vertebrate tissues (Fig. S8A) (Bhave et al., 2012; McCall et al., 2014). Thus, it seemed possible that crosslinking is increased in response to damage within the basement membrane. Alternatively, crosslinking might be required after the incorporation of new collagen IV that replaces damaged proteins, to maintain the original extent of tissue crosslinking. We evaluated these models both genetically and biochemically. We reasoned that if we knocked down Collagen IV with RNAi, collagen replacement would be reduced; simultaneously, we could feed flies PHG to inhibit Pxn. If crosslinking were required only for the replaced collagen in the basement membrane, then the phenotype would be no worse when both were inhibited together. In contrast, if crosslinking and collagen replacement were separate mechanisms of repairing basement membrane, then the phenotype would be worse when both were inhibited (see model in Fig. S8B). We found that the phenotype of inhibiting both processes was no worse than inhibiting each alone (Fig. S8C), arguing that Pxn is required for maintaining the extent of Collagen IV crosslinking after collagen is replaced during repair. In a separate biochemical experiment, we measured the amount of crosslinking in isolated guts by examining NC1 dimer electrophoretic mobility, which changes with crosslinking status, allowing us to calculate the number of crosslinks per hexamer. We found no significant change before, during, or after DSS treatment, with ∼3.7 crosslinks/hexamer in the gut on average, arguing that Collagen IV does not become hyper-crosslinked in response to basement membrane damage (Fig. S8D,E).

These results indicate that an important mechanism of basement membrane repair is through direct replacement of the matrix components. We tested this model by knocking down two critical components of basement membrane, Collagen IV (vkg) and Laminin (LanB1), initiating knockdown ubiquitously after the animals reached adulthood. Under the knockdown conditions tested, no effect was observed before damage, but each was required for repairing the basement membrane after damage, as assayed by muscle aspect ratio (Fig. 8). Similar results were observed with second RNAi lines, ruling out off-target effects (not shown). Thus, basement membrane is repaired by replacement and crosslinking of the newly incorporated matrix.

In this study, we determine that feeding DSS to flies damages the basement membrane around the midgut. Multiple assays were used to reach this conclusion. Specifically, FITC–DSS accumulates in the basement membrane around the midgut after feeding, and FITC–DSS binds to the basement membrane of Malpighian tubules upon soaking. The binding of DSS to basement membrane is relatively weak, as FITC–DSS washes out quickly when either tissue is incubated ex vivo. Upon DSS incorporation, the basement membrane expands as measured either by TEM of basement membrane electron density or by super-resolution microscopy of Vkg–GFP. Electron micrographs also show that after DSS incorporation the basement membrane becomes visibly damaged, appearing tattered and less dense. Despite these apparent structural changes, we have not detected changes in the protein composition of the basement membrane, as Laminin, Collagen IV and Perlecan are still present, and the fluorescence levels of Collagen IV, Laminin, and Perlecan GFP protein traps are the same as in undamaged tissue. Basement membrane tensional stiffness is altered upon DSS incorporation into Malpighian tubules, which contain only epithelial cells surrounded by basement membrane, with no other components such as muscles or cuticle to alter their response to tension. The tensional stress-strain assay reveals that DSS irreversibly damages the basement membrane ex vivo, rather than temporarily changing stiffness during its incorporation, as the basement membrane does not recover its stiffness after the DSS is washed out. Finally, after removal of DSS from food, repair of basement membrane occurs over 48 h in vivo. This repair process requires basement membrane proteins, including Collagen IV, Laminin and the Collagen IV crosslinking enzyme Pxn.

Although DSS clearly damages basement membrane, it is not possible to conclude that DSS damages no other tissue component. Interestingly, mouse enterocytes do not transport DSS across the epithelial barrier and DSS accumulates along extracellular junctions in the mouse intestinal lumen; although DSS is usually administered to mice for 7 days or longer (Chassaing et al., 2014), the level of tight junction protein ZO-1 begins to decrease after one day, with barrier permeability first observed at 3 days (Poritz et al., 2007), which is the same time as the earliest barrier permeability in flies. However, in flies the onset of basement membrane damage occurs before barrier permeability, so basement membrane damage is not secondary to barrier loss. Several lines of evidence support the conclusion that the substantial muscle damage induced by DSS feeding is secondary to basement membrane damage: similar muscle damage is recapitulated by Pxn knockdown, muscle morphology is not restored unless new basement membrane proteins can be incorporated, and basement membrane is damaged independently of muscles in the Malpighian tubules. We conclude that DSS causes direct damage to basement membranes, easily assayed by measuring muscle aspect ratio. Damage can also be measured by SIM, EM or mechanical stiffness assays.

Importantly, upon termination of DSS feeding, the basement membrane around the gut was repaired within 48 h, as assayed by both width measurements and muscle morphology. With repair cleanly separated from damage, DSS represents an excellent system for analysis of basement membrane repair. Feeding adult flies DSS is easy and creates reproducible damage in the gut basement membrane, which can be scaled up for biochemical analysis or genetic screening. The Drosophila gut itself is well suited to genetic analysis and microscopy, and unlike mechanical wounding, feeding DSS does not create cellular debris and/or clotting, which can be problematic for imaging. Further, Drosophila offers an excellent system for genetic analysis of basement membrane, as it has the same four basic protein components as mammals (Collagen IV, Laminin, Perlecan, and Nidogen) but there are far fewer genes encoding them (Ramos-Lewis and Page-McCaw, 2018) and the enzymes that modify basement membrane are also conserved but fewer in number, with one peroxidasin to promote Collagen IV crosslinking (Soudi et al., 2012) and two matrix metalloproteinases to cleave basement membrane proteins (Page-McCaw et al., 2003). We found some evidence for a fast-acting and temporary stabilization within 3 h of withdrawing DSS, but this reversal was short-lived; repair was a slower process, leading to a gradual improvement over 48 h.

Using this system, we began an analysis of basement membrane repair. Production of new Laminin and Collagen IV are required for repair, indicating that the basement membrane is repaired through replacement of original matrix components, rather than by crosslinking damaged components or replacing with different proteins. The crosslinking enzyme Pxn is also required. Because of the temporary improvement in aspect ratio observed at 3 h, we asked whether NC1 sulfilimine crosslinking acted independently of protein replacement to stabilize basement membranes, but both biochemical and genetic analysis indicated that the role of Pxn is limited to crosslinking the newly incorporated replacement Collagen IV molecules.

The Drosophila gut is a widely used model of stem cell-mediated homeostasis and regeneration. In this self-renewing tissue, cellular repair appears to use the same mechanisms as homeostasis, in that enterocytes and entero-endocrine cells are regularly replaced during adult life. Interestingly, our results show that in the gut basement membrane, Pxn is required for both repair of damage and also for basement membrane maintenance. The maintenance requirement suggests that even without damage, there is turnover of either sulfilimine crosslinks or collagen IV during homeostasis, suggesting that the gut basement membrane may be a more dynamic matrix than previously suspected. Similar to what is seen for cellular homeostasis and repair, basement membrane repair may also utilize the same mechanisms as homeostasis.

Drosophila melanogaster stocks were maintained at 25°C on standard cornmeal-molasses food unless otherwise indicated. See Table S4 for a complete description of fly lines used. For the temperature-sensitive RNAi experiments under Gal80^ts control, crosses between Tub-Gal4, Tub-Gal80^ts and UAS-(gene)^RNAi were performed at 18°C and progeny were allowed to grow to adults at 18°C. 3–5 day old (mated) female flies were then transferred to 29°C for indicated times. For Pxn knockdown, adult females remained at 29°C for 14 days before dissection. To minimize basement membrane damage prior to DSS exposure in repair assays, Pxn knockdown or LanB1 knockdown adult females were transferred to 29°C for 7 days prior to DSS/sucrose feeding, whereas vkg knockdown flies were transferred to 29°C for 11 days prior to DSS/sucrose feeding.

As described in Amcheslavsky et al. (2009), a 2.5 cm×3.75 cm piece of chromatography paper (Whatman, 3030-861, Grade: 3 MM CHR) was placed in an empty vial. 500 μl of a 5% sucrose solution with or without 3% 36–50 kDa DSS (dextran sulfate sodium salt colitis grade, MP Biomedicals, CAS number 9011-18-1; 36,0000–50,000 MW) was added directly to the chromatography paper in the vial. Anesthetized flies were carefully added to the vial so as to prevent them contacting the liquid media. Flies were transferred to a new vial with fresh media daily. Flies were fed according to this regimen for 48 h at 29°C. For recovery experiments, flies were allowed to recover for 48 h at 29°C on standard cornmeal molasses food. When indicated, 100 µM phloroglucinol (Sigma-Aldrich, 108736) was added to the DSS or sucrose solution before spotting onto the chromatography paper. For recovery in the presence of phloroglucinol, flies were transferred to molasses food (no cornmeal) containing 100 µM of phloroglucinol in solid media. Cornmeal was excluded to prevent the flies from eating around the phloroglucinol.

Adult females were placed in cold 1× PBS (137 mM NaCl, 2.7 mM KCl, 10 mM Na₂HPO₄, 1.8 mM KH₂PO₄) and were pinched with sharp #5 dissecting forceps (Dumont) between the abdomen and the thorax to separate the abdomen from the rest of the body. The abdominal cuticle was peeled off, leaving the guts and Malpighian tubules. Dissecting guts this way prevents them from being torn or pulled. Guts were dissected and immediately transferred with a Pasteur pipette into a 4% paraformaldehyde (Ted Pella Inc., 18505) in PBS fixative for 10 min at room temperature, and washed 3×5 min with PBS. For SIM analysis and for knockdown experiments in Fig. 8, guts were immunostained for GFP: blocked at 4°C overnight or room temperature for 2 h in blocking buffer (5% BSA, 5% normal goat serum, 0.05% NaN₃), then incubated with rabbit anti-GFP (Torrey Pines, TP401) at 1:200 in blocking buffer for 12 h at 4°C or 2 h at room temperature, then washed 3×10 min at room temperature in PBS, and then incubated with FITC donkey anti-rabbit (Jackson ImmunoResearch, 711-095-152) diluted 1:150 in blocking buffer for 12 h at 4°C or 2 h at room temperature, washed and mounted. When indicated, Alexa Fluor 647-conjugated phalloidin (Life Technologies), diluted 1:20 in blocking buffer, was added to secondary antibodies and incubated overnight at 4°C. All samples were mounted in DAPI-containing mounting media (Vectashield, Vector Laboratories, H1200). Analysis was confined to the posterior midgut, identified by its location anterior to the Malpighian tubules and posterior to the copper cell region, as depicted in the graphical abstract cartoon in Li et al. (2013). For gut dimensions, length was measured on intact guts from the crop to the posterior-most end of the gut; circumference was measured at the thickest part of the posterior midgut region.

For standard epifluorescence imaging, single optical sections were captured using a Zeiss Apotome mounted to an Axio Imager M2 with a 63×/1.3 oil objective. Images were taken with an AxioCam MRm camera (Zeiss), X-Cite 120Q light source (Excelitas Technologies), and AxioVision 4.8 software (Zeiss). ImageJ (version 1.48v, National Institutes of Health) 16-bit, grayscale, ZVI files were used for image analysis as well as for the images shown. For structured illumination microscopy (SIM), samples were mounted under a #1.0 coverglass (FisherBrand) and imaging and processing were performed on a GE Healthcare DeltaVision OMX equipped with a 60× Plan-apochromat N/1.42 NA oil objective lens and sCMOS camera.

For fluorescence intensity measurements of Vkg–GFP, LanB1–GFP, and Trol–GFP, samples to be compared were prepared on the same day and imaged at the same exposures. For each midgut, a representative field of the posterior midgut was imaged in a single optical cross section with a Nikon Apotome and a 63× objective. Within the field, a region of the midgut was selected for measurement based on its morphology, as only straight regions were analyzed. The ImageJ Measure tool was used to measure the total fluorescence intensity within a standard box (uniform length and area) that enclosed all the basement membrane of the enterocytes and muscles along a straight section of the midgut. An unpaired t-test was performed (GraphPad Prism 7.0). As an independent method to verify the fluorescence intensity of Vkg–GFP, we imaged three dimensions of the entire posterior midgut region of 40 sucrose- and 54 DSS-treated guts using a Nikon Spinning Disk microscope (Yokogawa CSU-X1 spinning disk head with Andor DU-897 EMCCD camera). After excluding regions that were out of focus, the maximum-intensity fluorescence projections were measured across the midgut, and data were compared using an unpaired t-test (GraphPad Prism 7.0). Like the first method, this second method of determining fluorescence intensity showed no difference between sucrose- and DSS-fed flies.

Samples were processed for TEM and imaged in the Vanderbilt Cell Imaging Shared Resource-Research EM facility. Samples were dissected in and fixed with 2.5% glutaraldehyde in 0.1 M cacodylate buffer, pH 7.4 at room temperature for 1 h then left at 4°C for 10 days. The samples were washed 3×5 min in 0.1 M cacodylate buffer, then incubated 1 h in 1% osmium tetroxide at room temperature, then washed with 0.1 M cacodylate buffer (1% calcium chloride, 0.1 M sodium cacodylate, and pH adjusted to 7.4). Subsequently, the samples were dehydrated through a graded ethanol series: 30%, 50%, 75%, 85%, 95%, 100%, 100%, 100% each for 15 min followed by a 1:1 solution of 100% ethanol and propylene oxide (PO) for 5 min. Samples were then infiltrated with 25% Epon 812 resin:75% PO for 35 min at room temperature. Next, they were infiltrated with 50% Epon 812 resin:50% PO for 1 h at room temperature then exchanged with new 50% Epon 812 resin:50% PO and incubated overnight at room temperature. Next day, the samples went through a 75%: 25% (resin: PO) exchange, were then exchanged into pure epoxy resin for 3–4 h, then incubated with pure epoxy resin overnight. Next, the resin was exchanged with freshly made, pure epoxy resin and incubated for 3 h, then embedded in epoxy resin and polymerized at 60°C for 48 h. For sectioning and imaging, 70–80 nm ultra-thin sections were then cut from the block and collected on 300-mesh copper grids. The copper grids were post-section stained at room temperature with 2% uranyl acetate (aqueous) for 15 min and then with Reynold's lead citrate for 10 min. Samples were subsequently imaged on the Philips/FEI Tecnai T12 electron microscope at 6500, 11,000, 15,000, 21,000 and 30,000× magnification.

Transmission electron microscopy (TEM) images of the basement membrane underlying either the gut epithelial layer or surrounding the Malpighian tubules were acquired at 30,000× magnification and analyzed in ImageJ. Measurements of basement membrane thickness were taken at regular intervals across the field of view from at least 10 images per experimental condition only where the basement membrane was clearly defined. A t-test was performed to compare DSS-fed versus sucrose-fed flies, using GraphPad Prism 7.0. Alternatively, SIM was used to measure basement membrane thickness as labeled using Vkg–GFP fluorescence. A rotation student, unfamiliar with the experiment and blinded to sample identity, chose the locations and made the measurements. GraphPad Prism 7.0 was used to perform statistical analysis on the sample set.

The sarcomere size was measured in gut muscles stained with phalloidin by drawing a line in ImageJ (version 1.48v) from one phalloidin-stained actin region to the next phalloidin-stained actin region. When indicated, guts were dissected in relaxation buffer (20 mM sodium phosphate, 5 mM MgCl₂, 5 mM EGTA, 5 mM ATP solution, 5 mM DTT, 10 mM 100× Halt Protease Inhibitor) and allowed to incubate for 30 min before further processing (Xiao et al., 2017).

To quantify muscle morphology, the aspect ratio (height:width) of the muscles surrounded by basement membrane was determined from the basement membrane staining (Vkg–GFP or LanB1–GFP) surrounding the muscle. Height was measured at the tallest part of the muscle, and a line perpendicular to the height line was measured as the width, as shown in Fig. 1D′. Aspect ratio was calculated as height:width.

Malpighian tubule basement membrane stiffness was measured in a similar manner to that described previously (Bhave et al., 2017). Briefly, measurement cantilevers were fabricated from pulled hollow glass capillary tubes and cantilever spring constants were measured in a manner similar to Shimamoto and Kapoor (2012). Malpighian tubules were attached to the measurement cantilever and a holding pipette (10 μm inner diameter), as shown in Fig. 4J, by applying vacuum. Both the measurement cantilever and holding pipettes were attached to micromanipulators. Imaging was performed on an inverted microscope (VWR) at 4× magnification with an attached digital camera for image acquisition. The holding pipette was translated in 40 μm increments to stretch the tubule and deflect the measurement cantilever. Images acquired at each deflection increment were used to calculate the deflection of the measurement cantilever and the displacement of the holding pipette. The change in the length of the tubule (Δl) was calculated as the difference between the distance traveled by the holding pipette and the deflection of the measurement cantilever (d_m). Initial length of the tubule was measured from an image acquired prior to translating the holding pipettes. After the experiment was complete, tubules were observed to return to their original length, indicating that only elastic strain had been recorded.

For feeding experiments, 3–5-day-old female flies were fed for 48 h on 5% sucrose/3% DSS, with the DSS prepared as follows: FITC–DSS (Sigma Aldrich, 78331-1G) diluted 1:10 with regular DSS; DSS and unconjugated FITC in the same molar ratio as in the first condition (0.86 mg FITC and 300 mg DSS in 10 ml water); FITC alone at a matched molar concentration. After feeding, flies were placed on normal cornmeal-molasses food for 6 h as a chase, then guts were dissected in PBS and immediately mounted in Grace's insect media (BioWhittaker, 04649F) for imaging without fixation.

For Malpighian tubule soaking experiments, tubules were dissected from 3–5-day-old female flies in PBS and immediately transferred to PBS containing one of the following: either FITC–DSS diluted 1:10 with regular DSS; DSS and unconjugated FITC in the same molar ratio as in the first condition (0.86 mg FITC and 300 mg DSS); or FITC alone at a matched molar concentration. Tubules were soaked for 20 min, then washed 3×2 min each with PBS, and mounted and imaged without fixing in PBS. To wash out the FITC–DSS or control FITC after soaking as above, tubules were washed 5×5 min washes in 1× PBS, which was determined by visual inspection under epifluorescence microscopy as sufficient to remove the fluorescence signal.

For mouse DSS experiments, mouse animal experiments were performed under protocols approved by the Vanderbilt University Animal Care and Use Committee and in accordance with NIH guidelines. C57BL6/J mice (Jackson Laboratory) were administered a 2.5% DSS solution in the drinking water consisting of 10% FITC-conjugated DSS/90% unconjugated DSS and were euthanized 24 h later. Control mice received a mix of 10% FITC+DSS (unconjugated)/90% unconjugated DSS. Upon euthanasia, intestinal tissues were removed, washed with 4% PFA, and spread longitudinally onto Whatman paper. Tissues were swiss-rolled, embedded in optimal cutting temperature medium (OCT), and frozen immediately at −80°C. For microscopy, tissues were sectioned at 5 mm thick onto glass slides. Slides were washed once in PBS and incubated at room temperature overnight in primary antibodies against EpCam (rat monoclonal G8.8, Santa Cruz Biotechnology, 53532) or Laminin (rabbit polyclonal, Sigma-Adrich, L9393) at 1:100 dilution, followed by three washes in PBS, and 1 h incubation in Alexa Fluor 647-labeled secondary antibodies at 1:500 dilution (Life Technologies, donkey anti-rabbit-IgG A31573 or goat anti-rat-IgG A21247) and Hoechst 33342 (Invitrogen, H3570) (1:100). Slides were mounted in Prolong Gold and viewed using fluorescent microscopy.

60 w¹¹¹⁸ flies were fed with DSS or sucrose as above but also including 0.5% erioglaucine disodium salt (aka Brilliant Blue; Sigma-Aldrich, 861146). The flies were examined three times daily (8:00 am, 2:00 pm, 8:00 pm) to assess barrier integrity (appearing blue or ‘smurf’ when barrier integrity is lost). The flies were deemed smurf only if their bodies were blue and they were alive. Every smurf fly was dead by the next time point.

Each sample contained roughly 200 w¹¹¹⁸ adult fly guts, dissected in Grace's insect media. Guts were transferred in 25-gut cohorts into a 2 ml pre-tared tube on ice, repeated until all samples were collected. Media was removed and samples were weighed, as 200 mg was a minimum starting mass. Samples were snap-frozen with liquid N₂. Using a cold mortar and pestle, frozen guts were ground into a fine powder, which was weighed then refrozen in liquid N₂. Samples were resuspended in deoxycholate solubilization buffer (1% sodium deoxycholate, 10 mM Tris pH 7.4, 1 mM EDTA) with 100× protease inhibitors (Thermo Scientific, 78430) and 50 µM phloroglucinol, at a concentration of 5 µl buffer per mg of powdered guts. Samples were sonicated (1 s pulse separated by 1 s pause for a total of 20 s) then incubated for 15 min on ice. To enrich for Collagen IV, samples were spun in a microfuge at 16.1 g for 30 min at 4°C, the pellet was washed (1 ml/200 mg sample) with chilled high salt wash (1 M NaCl, 50 mM Tris pH 7.4, 100 µM phloroglucinol), vortexed and incubated on ice for 15 min, then spun at 4°C for 30 min. The pellets were then washed (1 ml/200 mg sample) in a hypotonic wash (10 mM Tris pH 7.4, 50 µM phloroglucinol), tubes were inverted to rinse pellet, and then spun for 5 min at 4°C. Bacterial collagenase [1 mg/ml bacterial collagenase CLSPA (Worthington, LS005273), 5 mM CaCl₂, 10 mM Tris pH 7.4, 0.1 mM phloroglucinol, 10 mM 100× Halt Protease Inhibitor] was added to the pellet at a concentration of 1 ml/200 mg sample, and the tube was inverted multiple times before being wrapped in foil and incubated for 24 h in a 37°C water bath. Precipitate was removed by spinning down at 4°C for 30 min, and the supernatant containing the Collagen IV NC1 domains was immediately collected and filtered through a 0.2 µm pore mini-syringe filter. Using a Thermo Scientific NanoDrop 1000 spectrophotometer, the protein amount was determined, then samples were lyophilized and resuspended in water at the standardized concentration of 10 mg/ml. 200 ng of sample was run per lane. Samples were resuspended in non-reducing 4× sample buffer (200 mM Tris-HCl pH 6.8, 8% SDS, 40% glycerol, 0.2% Bromophenol Blue), heated for 5 min at 95°C, and loaded onto a Bio-Rad non-reducing 10% gel (Mini-PROTEAN TGX, 456-1034), and run at 100 volts. After removal from the apparatus and before transfer, the gel was reduced in 1× transfer buffer (25 mM Tris base, 192 mM glycine) with 2% β-mercapthoethanol (Fisher Scientific, 60-24-2) for 1 h at room temperature. A rabbit anti-NC1 primary antibody (1:500; McCall et al., 2014) and a 680 anti-rabbit secondary antibody (1:8000; LI-COR Biosciences, 926-32223) were used for imaging on an Odyssey CLx imaging system (LI-COR Biosciences).

We thank A. Scott McCall for his original recognition that Pxn loss of function phenocopied DSS treatment. We thank Lauryn Luderman and James O'Connor for assisting with blinded experiments and Dhiraj Peddu for help assessing smurf flies at night. We are grateful to Vincent Mirouse for the generous gift of the Trol–RFP fly line. Fluorescence intensities were measured with the help of Bryan Millis at the Nikon Center of Excellence. Indrayani Waghmare, James O'Connor, and Will Ramos-Lewis were essential for discussions and comments along the way. Electron microscopy was performed in part through the use of the Vanderbilt Cell Imaging Shared Resource (supported by National Institutes of Health grants CA68485, DK20593, DK58404, DK59637 and EY08126). We are grateful to the Bloomington Drosophila Stock Center and the Vienna Drosophila RNAi Center for fly lines. A huge thank you to A.M.H.’s three fur children: Charles contributed great positivity and excitement, and Megs and Benjamin ‘contributed’ to editing.