Boosting apoptotic cell clearance by colonic epithelial cells attenuates inflammation in vivo

BAI1 deficiency affects apoptotic cell clearance after acute tissue injury in multiple tissues

To examine the role of BAI1 in apoptotic cell clearance in vivo, we generated mice with Bai1 gene disruption (Figure S1A). The Bai1^−/− mice appeared grossly normal, but bone marrow derived macrophages from Bai1^−/− mice showed a defect in the engulfment of apoptotic targets as well as an increased release of inflammatory cytokines (Figure 1A). The lack of a global phenotype at steady state was likely due to the continued expression of Bai1 homologous genes or the known redundancy in engulfment pathways, as gene deletion of individual engulfment genes often elicit a less pronounced global phenotype (Devitt et al., 2004; Elliott et al., 2010; Park et al., 2011; Scott et al., 2001).

Defects and inflammatory sequelae due to disruption of individual engulfment genes are often better revealed when mice are challenged with insults that induce apoptosis of a large number of cells within a tissue (Devitt et al., 2004; Elliott et al., 2010; Park et al., 2011; Scott et al., 2001). We tested this possibility in three different tissue contexts. First, when BAI1-deficient mice were challenged with dexamethasone to induce apoptosis in the thymus, the clearance of apoptotic thymocytes was compromised, with many remaining uncleared corpses (as shown by cleaved caspase 3 staining) (Figure 1B). Second, when we tested the BAI1-deficient mice in a model of testicular torsion, a form of ischemia-reperfusion injury where the apoptotic germ cells get cleared by Sertoli cells that line the seminiferous epithelium (Elliott et al., 2010; Park et al., 2011), the number of uncleared apoptotic corpses was greatly increased in mice lacking BAI1 (Figure 1C). Third, we assessed apoptotic cell clearance in a model of colonic inflammation. We chose the dextran sodium sulfate (DSS)-mediated colitis model, as this has been shown to mimic some aspects of the inflammatory features of human ulcerative colitis (Perse and Cerar, 2012) and apoptosis of colonic epithelial cells clearly contributes to DSS-induced colitis in this model (Qiu et al., 2011). The colonic tissue, both epithelial cells and the lamina propria fraction, primarily expressed Bai1, with minimal expression of the homologues Bai2 and Bai3, or the two other direct phosphatidylserine-binding receptors, Timd4 and Stab2 (Figure 1D and Figure S1B–S1D). DSS given orally via drinking water induced colitis over several days (Figure 1E). When we analyzed the presence of apoptotic cells within the colon after DSS treatment, the number of TUNEL⁺ nuclei in the colonic epithelium of Bai1^−/− mice was significantly increased relative to control mice (Figure 1F). When we stained for cleaved caspase3, an executioner caspase involved in apoptosis but not necroptosis, Bai1^−/− mice showed significantly increased cleaved caspase 3⁺ cells in their colonic epithelium (Figure 1G), consistent with previous studies demonstrating that colitis induction after DSS treatment involves apoptosis (Gunther et al., 2011; Pasparakis and Vandenabeele, 2015; Qiu et al., 2011). Collectively, these data suggested that genetically removing BAI1 could lead to accumulation of uncleared apoptotic cells after induction of apoptosis in three different tissue contexts in vivo.

The deficiency of BAI1 augments acute colitis severity

We then assessed how the expression of endogenous Bai1 might be modulated in the wild type mice after DSS treatment. Endogenous Bai1 mRNA in colonic tissue and in the gut epithelial cells progressively decreased over time after DSS administration (Figure 2A and Figure S1E). Furthermore, when we tested HCT-116 colonic epithelial cells, BAI1 mRNA in HCT-116 cells was also decreased by DSS treatment (Figure S1F). Moreover, in various transcriptomic analyses, Bai11 expression as well as the downstream signaling intermediates ELMO1, ELMO2, DOCK1, and RAC1 have been reported to be downmodulated in conditions of colonic inflammation in both mouse and human (Carey et al., 2008; Costello et al., 2005; Gao et al., 2013a; Kugathasan et al., 2008; Noble et al., 2008; Swan et al., 2013). These observations suggested that the complete loss of BAI1 in the gene-deleted mice, from the beginning of DSS administration, might contribute to greater to colonic inflammation, and we addressed this further.

When given 3% or 5% DSS in drinking water, the Bai1^−/− mice showed greater weight loss at days 2, 3, 4, and 5 compared to control mice and also displayed an increase in disease severity index, which is a score that encompasses quantification of weight loss, bloody stools, and diarrhea (Figure 2B and 2C and Figure S1G and S1H). Bai1^−/− mice showed decreased survival after 8 days of 5% DSS treatment with only 13% survival rate for Bai1^−/− mice compared to 83% survival for co-housed controls (Figure 2D). Histologically, the DSS-treated Bai1^−/− mice showed more destruction of the colonic epithelial structure, increased span of the inflamed region, and a thicker muscularis layer (Figure 2E). We also performed ‘continuous’ sections of the proximal region (which generally had more lesions) from multiple mice (~14 sections at 200 µm intervals per proximal region of mouse colon) and scored the number of TUNEL⁺ cells from these sections. TUNEL staining of the last six continuous sections of the proximal colon from two control and two Bai1^−/− littermates are shown in Figure S2, with Bai1^−/− mice showing much more TUNEL⁺ cells compared to their control littermates. Without DSS treatment, the Bai1^−/− mice showed no increase in apoptotic cells in the colon and did not show any obvious colonic pathology compared to control mice (data not shown). The colon length (a disease parameter) was comparable in untreated control and Bai1^−/− mice (at 8 and 40 weeks) (data not shown), but the DSS-treated Bai1^−/− mice had significantly shorter colons (Figure S1I and S1J).

Consistent with prior studies that demonstrated acute DSS-induced colitis is minimally influenced by the adaptive immune system (Dieleman et al., 1994), the mesenteric lymph nodes and spleen did not show significant difference in lymphocyte numbers between Bai1^+/+ and Bai1^−/− mice after DSS treatment (Figure 2F). Also, the CD4⁺CD25⁺Foxp3⁺ regulatory T (T_reg) cell subpopulations of the mesenteric lymph node and spleen were similar (Figure 2G). Furthermore, there was no difference in the T helper-17 (Th17) cell population (Figure 2H) isolated from the lamina propria of control and Bai1^−/− mice. To minimize environmental influences that might affect colitis severity (Brinkman et al., 2013), the experimental animals and the appropriate littermate controls were always housed in the same cages and exposed to the same bedding, food, and water in all experiments. In addition, treatment of control and Bai1^−/− littermates with an antibiotic cocktail (known to affect the bacterial population of the gut and the adaptive immune response (Diehl et al., 2013)) prior to initiation of DSS-induced colitis did not affect the enhanced colitis severity in BAI1-deficient mice (Figure 2I–2K). Mertk, a member of the TAM family of engulfment receptors continued to be expressed in the colonic tissue (Figure S1D), but was not able to overcome the loss of BAI1 under these DSS-treated conditions (note: expression of other two TAM members, Tyro3 and Axl, was low). Collectively, these data suggested that mice lacking the phagocytic receptor BAI1 were more susceptible to acute injury to the colon in the context of DSS-induced colitis.

Transgenic overexpression of BAI1 attenuates DSS-induced colitis in vivo

We next asked whether forcibly increasing the expression of BAI1 would be beneficial in this colonic inflammation model. We engineered transgenic mice capable of conditionally overexpressing a HA-tagged human BAI1 construct (BAI1^Tg). The human and mouse BAI1 are highly homologous and function comparably in the in vitro engulfment assays (Park et al., 2007), but using the cDNA for human BAI1 as transgene allowed distinction from the endogenous mouse Bai1. A single copy of the wild type BAI1 transgene was introduced via homologous recombination into the non-essential Rosa26 locus in C57BL/6 mouse embryonic stem cells, and used to generate chimeric founder transgenic mice (BAI1^{flox-STOP-flox}). The expression of the BAI1^Tg was kept silent by a transcriptional and translational STOP cassette flanked by loxP sites, with cre-mediated deletion of the STOP cassette allowing expression of the HA-BAI1 and the bicistronic eGFP marker (Figure 3A). Initially, we expressed transgenic BAI1 globally by crossing the BAI1^{flox-STOP-flox} mice to mice carrying E2Acre, which mediates ubiquitous deletion in most tissues (denoted as E2A-cre, BAI1^Tg mice) (Lakso et al., 1996).

We first confirmed the cre-dependent expression of transgenic human HA-tagged BAI1 and eGFP in the colon of the E2A-cre, BAI1^Tg mice (Figure 3B and 3C and Figure S3A and S3B). The peritoneal macrophages from E2A-cre, BAI1^Tg mice showed greater engulfment (about 18 %) compared to macrophages from control mice (Figure S3C). When the BAI1^Tg mice were tested in the testicular torsion model (Elliott et al., 2010; Park et al., 2011), the testis of BAI1^Tg mice had significant reduction in the number of apoptotic corpses compared to controls (Figure 3J). When tested in the dexamethasone-induced thymic apoptosis model, there was also a trend toward reduction in the number of apoptotic corpses in the BAI1^Tg mice but this did not achieve statistical significance due to the magnitude of apoptosis in this model (data not shown). When we assessed how the expression of transgenic human BAI1 versus the endogenous mouse Bai1 compared after DSS treatment, the transgenic human BAI1 mRNA expression was unaffected in E2A-cre, BAI1^Tg mice by DSS administration, while the endogenous mouse Bai1 mRNA was decreased as in control mice (Figure 3D and 3E). This provided a system to test whether forcibly enhancing BAI1 expression could be beneficial in colonic inflammation.

We tested the E2A-cre, BAI1^Tg mice in the DSS colitis model, and made several observations: (i) The E2A-cre, BAI1^Tg mice had less weight loss after DSS treatment compared to controls (Figure 3F and Figure S3D and S3E, the controls being either BAI1^Tg mice that still had the STOP cassette, or mice carrying cre without the transgene); (ii) The E2A-cre, BAI1^Tg mice had an overall less disease severity index (Figure 3G and Figure S3F); (iii) The E2A-cre, BAI1^Tg mice had less colonic disease as measured by the colon length (Figure S3G and S3H); (iv) Histological analysis from multiple sections of DSS-treated mice showed that E2A-cre, BAI1^Tg mice had reduced relative area of inflamed colonic region, and thinner muscularis layer relative to controls, despite a similar baseline colonic architecture (Figure 3H); (v) The E2A-cre, BAI1^Tg mice had fewer TUNEL⁺ cells in their colonic epithelium compared to control littermates (Figure 3I). Collectively, these data suggested that increasing BAI1 expression could provide a ‘gain of function’ in an acute inflammation model of the colon.

Overexpression of a signaling deficient BAI1 mutant does not attenuate colitis severity

We asked whether this attenuated disease seen in response to increasing BAI1 expression was simply an effect of higher quantities of a membrane protein that can engage PtdSer or if it required BAI1 engagement of its downstream signaling components. BAI1 signals via the intracellular engulfment protein ELMO and its binding partner DOCK1 (either ELMO1 or ELMO2 depending on the cell type, with ELMO2 being more prominently expressed in epithelial cells) (Park et al., 2007). The ELMO-DOCK1 complex, in turn, leads to the activation of the small GTPase RAC, and the resulting cytoskeletal rearrangements within the phagocytes facilitate corpse internalization (Figure 4A) (Park et al., 2007). Of note, in unbiased transcriptome analysis, the expression of ELMO2 and DOCK1 have been shown to be lower in patients with inflammatory bowel disease (IBD) (Bjerrum et al., 2010; Burczynski et al., 2006; Costello et al., 2005) and in colitis models in animals (Schmidt et al., 2010). Mutation of three contiguous intracellular residues within the cytoplasmic tail of BAI1 (denoted as BAI1-AAA mutant) can abrogate the binding to ELMO and phagocytosis (Figure 4A) (Park et al., 2007). We engineered transgenic mice expressing the BAI1-AAA construct (essentially similar to the wild type BAI1^Tg mice above). For generating the BAI1-AAA transgenic founder mice, we used ES cell clones that could induce BAI1-AAA on the membrane equivalent to ES cells carrying the wild type BAI1 transgene (data not shown). To induce mutant transgene expression, the BAI1-AAA mice were crossed to E2A-cre mice (E2A-cre, BAI1-AAA^Tg) and the BAI1-AAA expression was confirmed (Figure 4B and Figure S4A). When tested in the DSS inflammation model, the E2A-cre, BAI1-AAA^Tg mice did not show an attenuation of colitis severity compared to the control littermates, with comparable weight loss, disease severity index, colon length, and inflammatory region (Figure 4C–4F and Figure S4B–S4E). Also, there were comparable TUNEL⁺ nuclei in DSS-treated conditions in the control and BAI1-AAA^Tg mice (Figure 4G). These data demonstrated that BAI1 requires the engagement of downstream signaling components for attenuating colonic inflammation.

BAI1 overexpression in the myeloid lineage fails to attenuate colitis severity

Since professional phagocytic cells of the myeloid lineage within the lamina propria are considered one of the key mediators in intestinal homeostasis and inflammation (Peterson and Artis, 2014), we first crossed the transgenic BAI1^{flox-STOP-flox} mice with mice that drive cre expression under the lysozyme M promoter in multiple cells of myeloid lineage (Lyz2-cre) (Clausen et al., 1999). Total lamina propria, as well as macrophages from the lamina propria of colons (in addition to the peritoneal macrophages) from Lyz2-cre, BAI1^Tg mice confirmed myeloid expression of transgenic human BAI1 and/or eGFP (Figure 5A and Figure S5A and S5B). In ex vivo analysis, peritoneal macrophages from Lyz2-cre, BAI1^Tg mice showed enhanced the uptake of apoptotic cells (Figure S5C). Lyz2-cre, BAI1^Tg mice did not show a benefit in colitis after DSS treatment compared to control mice, as determined by the lack of a significant difference in the bodyweight change, disease severity index, colon length, and colonic inflammatory regions (Figure 5B–5D and Figure S5D–S5H). Thus, forcibly increasing BAI1 expression in myeloid cells is insufficient to reduce DSS-induced colonic inflammation. This prompted us to ask whether boosting BAI1 in other cell types would recapitulate the benefit seen in the E2A-cre, BAI1^Tg mice.

BAI1 overexpression in colonic epithelial cells is sufficient to attenuate colitis severity

Non-professional phagocytes, such as epithelial cells, can engulf apoptotic cells in other tissues including lung and mammary gland, and also produce anti-inflammatory mediators to dampen tissue inflammation (Bagci et al., 2014; Juncadella et al., 2013; Monks et al., 2008; Sandahl et al., 2010). Therefore, we asked whether boosting BAI1 expression within colonic epithelial cells contribute to dampening inflammation in the DSS colitis model. We crossed the transgenic BAI1^{flox-STOP-flox} mice with mice carrying the cre under the villin 1 promoter (Vil1-cre) to induce the BAI1^Tg specifically in the intestinal epithelial cells (Madison et al., 2002). The Vil1-cre, BAI1^Tg mice (but not the littermate controls) showed specific expression of HA-BAI1 in the EpCAM-positive intestinal epithelial cells and colon (Figure 5E and Figure S6A–S6C), and transgenic BAI1 was undetectable in the EpCAM-negative cell types (Figure 5E and Figure S6C). Targeting BAI1 overexpression to colonic epithelial cells was sufficient to dampen inflammation in the DSS colitis model. Compared to control littermates, the Vil1-cre, BAI1^Tg mice showed relatively little weight loss, lower overall disease severity index, a thinner muscularis layer, increased colon length, and reduced inflammatory region (Figure 5F–5I and Figure S6D–S6F). Although we have not assessed BAI1 overexpression in other myeloid subtypes (such as dendritic cells), the data with Vil1-cre, BAI1^Tg mice implies that overexpression of BAI1 in colonic epithelial cells alone can reduce colitis severity.

We then asked how the improved disease parameters in Vil1-cre, BAI1^Tg mice correlate with apoptotic cell clearance within the colonic epithelium. First, the colonic epithelial cell line HCT-116 expressed BAI1 mRNA and readily internalized apoptotic targets in vitro; since epithelial cells can also bind each other readily, the corpse internalization in this assay was confirmed by using as targets apoptotic epithelial cells labeled with a pH sensitive dye cypHer 5E (which fluoresces brighter after reaching the lower pH of the phagolysosome) (Figure 6A and 6B). Moreover, decreasing BAI1 expression in HCT-116 cells via siRNA reduced the engulfment of apoptotic targets (Figure 6A and 6B), while overexpression of wild type BAI1, but not the signaling deficient BAI1-AAA mutant, enhanced uptake by HCT-116 cells (Figure 6C). Of note, silencing or overexpression of BAI1 did not affect the survival, apoptosis or proliferation rate of the colonic epithelial cells (data not shown). Another colonic epithelial cell line Caco-2 could also engulf apoptotic BEAS-2B or MLE-12 epithelial cells (Figure 6D).

We next asked whether the BAI1 overexpressing colonic epithelial cells in the Vil1cre, BAI1^Tg mice would show increased engulfment. We used a detection method that can score both the apoptotic cells and also the ‘engulfed targets’ within phagocytes. While the TUNEL assay in vivo detects DNase I cleavage within apoptotic nuclei (type I break) (Loo, 2011), the DNA derived from the engulfed targets undergoes DNase II-mediated cleavage (type II break) within the phagolysosomes (Figure 6E). We quantified both apoptotic (type I DNA cleavage) and engulfed cells (by monitoring type-II DNA cleavage) within the colonic epithelium (Evans and Aguilera, 2003). The colon sections of Vil1-cre, BAI1^Tg mice showed increased numbers of engulfed apoptotic cells (Figure 6F and 6G) with concomitantly fewer TUNEL-positive apoptotic nuclei in the same sections (Figure 6H and Figure S6G). In comparison, the control mice had more TUNEL positive apoptotic nuclei and relatively fewer engulfed cells (Figure 6F–6H). Co-staining for EpCAM and z-stack analysis of the sections showed that the engulfed targets were internalized, and resided within other EpCAM⁺ epithelial cells (Figure 6F and 6I and Figure S6H). These data suggested that increasing the expression of BAI1 within the intestinal epithelial cells could enhance cell clearance within the colonic epithelium, and in turn, attenuated colitis severity in this model.

BAI1 signaling modulates inflammatory cytokines in the colitis model

While the clearance of early stage apoptotic cells is anti-inflammatory, failure or delay in corpse clearance can lead these early apoptotic cells to become secondarily necrotic and also pro-inflammatory (Henson, 2005; Nagata et al., 2010). Since alarmins such as interleukin 1α (IL-1α) and IL-33 have been known to be released from damaged epithelial cells and linked to human colitis (Moussion et al., 2008; Suwara et al., 2014), we assessed their expression. Prior to the DSS treatment, the expression of inflammatory cytokine genes, Il1a and tumor necrosis factor (Tnf) were barely detectable in either the control or the Bai1^−/− mice, consistent with no obvious basal disease (Figure 7A). However, in the colonic tissue of Bai1^−/− mice, the qunatity of Il1a and Tnf were increased at day 5 of DSS treatment (Figure 7A and 7B). In addition, quantity of IL-33 mRNA and protein were significantly increased in Bai1^−/− mice (Figure 7A and 7B). The mRNA of other inflammatory cytokines, Il6, Il12b, and Il22 (but not Il17a) were also increased in the Bai1^−/− mice compared to control mice (Figure S7A–S7D). In contrast, in mice globally overexpressing BAI1 (E2A-cre, BAI1^Tg mice) or specifically in the intestinal epithelial cells (Vil1-cre, BAI1^Tg mice), the mRNA for Il1a and Tnf were decreased compared to littermate controls, correlating with reduced colitis severity and fewer apoptotic cells in BAI1^Tg mice (Figure 7C and 7D). Also noteworthy is that mice transgenic for the mutant form of BAI1 (BAI1-AAA^Tg mice), did not display a reduction in Il1a and Tnf (Figure 7E).

The increased inflammation seen in the mice lacking BAI1 could have occurred due to accumulation of corpses and/or the lack of signaling via BAI1 that dampens inflammation. However, this is not readily distinguishable in vivo, as corpse recognition, engulfment, and anti-inflammatory signaling are closely linked. Therefore, we tested the contribution of BAI1 in vitro in regulating inflammatory cytokine release. We measured the inflammatory cytokines in BAI1-silenced cells (HCT-116) and peritoneal macrophages of Bai1^−/− mice after incubation with apoptotic cells. We found that silencing of BAI1 in HCT-116 resulted in higher amounts of the inflammatory cytokine IL1A and a modest reduction of the anti-inflammatory cytokine TGFB1 and Bai1^−/− mice in peritoneal macrophages also had increased amount of inflammatory cytokines (Il1a, Il1b, Il33, and Tnf) (Figure S7E–S7J). In addition, in vitro, treatment of HCT-116 cells with DSS could partially recapitulate the release of IL-1α as the cells became progressively more necrotic over a time course (Figure S7K–S7M). (Note: IL-33 was not detectable under these conditions from HCT-116 cells). These data suggest that the severity of colonic inflammation seen in vivo could have arisen from both the presence of uncleared apoptotic cells in the colonic tissue, as well as the reduction of anti-inflammatory signaling by BAI1.

Generation of Bai1 deficient and BAI1 transgenic mice

ES cells (GST_4419_G3) with a gene-trap mutation between exon 2 and 3 of Bai1 were obtained from TIGM. To generate conditional BAI1 transgenic mice, HA-tagged human BAI1 cDNAs (encoding wild type BAI1 and BAI1-AAA mutant) were inserted into the previously described CAG-STOP-eGFP-ROSA26TV (CTV) vector (Xiao et al., 2007). CTV-BAI1 vectors were transfected into C57BL/6 embryonic stem cells (JM8A3) and screened for homologous recombination into Rosa26 locus. Transgene encoded BAI1 and eGFP were confirmed in the ES cells in vitro via cre transfection. Selected ES clones were used to obtain transgenic mice. Tg-BAI1^{flox-STOP-flox} mice were crossed with E2A-cre mice (Jackson laboratory), for global transgene expression (Lakso et al., 1996), Lyz2-cre mice (Jackson laboratory), (Clausen et al., 1999) for myeloid expression, Vil1-cre mice (Jackson laboratory) for intestinal epithelial cell expression (Madison et al., 2002). All animal experiments were performed according to protocols approved by the animal care and use committee (ACUC) at the University of Virginia.

DSS-induced colitis

Mice (littermates, 8 ~ 10 weeks old) were given 3 % or 5 % dextran sulfate sodium (DSS; MP Biomedicals) in drinking water for 5 or 8 days and analyzed on d5 or d8. Changes in body weight and disease severity index were checked daily. Disease severity was determined based on weight loss, blood in the stool, and stool consistency, as previously described (Owen et al., 2011). The total scores were calculated as follows: weight loss (0: < 1 %, 1: 1~5 %, 2: 6~10 %, 3: 11~15 %, 4: > 15 %), stool blood (0: negative, 2: positive, 4: gross bleeding), and stool consistency (0: normal, 2: loose stools, 4: diarrhea). All mice were randomly allocated to experimental groups and not recognized during the experiment.

Staining for apoptotic and engulfed corpses

Mice were euthanized on day 5 after DSS treatment. Proximal colon region was cut by microtome at 200 µm intervals and TUNEL stained with the In situ cell death detection kit, TMR red (Roche) according to the manufacturer’s instructions. Five random fields were captured per section for each mouse. Apoptotic indices from each field per mouse were integrated and shown as the mean ± s. e. m. Apoptotic cells and engulfed apoptotic cells were stained by labeling of DNA breaks due to DNase type I within apoptotic cells and DNase type II within phagolysosome using Apoptag ISOL dual fluorescence apoptosis detection kit (Millipore). Subsequently, the samples were incubated overnight with EpCAM antibody (APC conjugated; eBioscience; Catalog number 17-5791-80). The nuclei were counterstained with Hoechst 33342. The samples were analyzed using AxioImager Z2 equipped with Apotome for optical sectioning. About 10 fields of the proximal colon were captured for each mouse. Signals of DNA breaks induced by DNase type II were quantitatively analyzed by Image J (NIH) program. Engulfment index, the percentage of DNase type II versus the total nuclei signals (Hoechst 33342 staining) was calculated for each field. Engulfment indices from each field per mouse were integrated and depicted as mean ± s. e. m.

Statistical analysis

Sample numbers were determined based on optimal size sufficient statistical power. There were no inclusion/exclusion criteria for the analysis. Data are shown as means ± s.e.m. from multiple independent experiments. Student’s two-tailed t-test was applied for analysis of statistical significance, defined as P values <0.05.