Mice that lack activity of αvβ6- and αvβ8-integrins reproduce the abnormalities of Tgfb1- and Tgfb3-null mice

The arginine-glycine-aspartate (RGD)-binding integrins αvβ6 and αvβ8 activate latent TGFβ1 and TGFβ3 in vivo, but it is uncertain whether other RGD-binding integrins such as integrins αvβ5 and αvβ3 activate these TGFβ isoforms. To define the combined role of αvβ6- and αvβ8-integrin in TGFβ activation, we analyzed mice lacking function of both integrins by means of gene deletion and/or pharmacologic inhibition. Most Itgb6–/–;Itgb8–/– embryos die at mid-gestation; those that survive develop cleft palate–as observed in Tgfb3–/– mice. Itgb8–/– mice treated with an anti-αvβ6-integrin antibody develop severe autoimmunity and lack Langerhans cells–similar to Tgfb1-null mice. These results support a model in which TGFβ3-mediated palate fusion and TGFβ1-mediated suppression of autoimmunity and generation of Langerhans cells require integrins αvβ6 and αvβ8 but not other RGD-binding integrins as TGFβ activators.

Processing of the TGFβ pro-protein produces a noncovalent complex between the TGFβ pro-peptide (called latency-associated peptide, LAP) and TGFβ (Annes et al., 2003). The association of LAP with TGFβ has to be altered or eliminated before TGFβ can engage TGFβ receptors. TGFβ activators include proteases that degrade LAP and molecules that nonproteolytically disrupt the latent complex, such as thrombospondin-1 and integrins αvβ6 and αvβ8 (Kojima et al., 1993;Yu and Stamenkovic, 2000;Annes et al., 2003;Mu et al., 2002;Munger et al., 1999).
Mice with a knock-in mutation of Tgfb1 that inactivates the integrin-binding site in TGFβ1-LAP (RGD changed to RGE) duplicate the phenotype of Tgfb1 -/mice . Thus, RGD-binding integrins are required for TGFβ1 activation in early life. In addition to integrins αvβ6 and αvβ8, the RGD-binding integrins αvβ1, αvβ3, αvβ5 and α8β1 bind TGFβ1-LAP and, in the case of integrin αvβ3, also TGFβ3-LAP (Sheppard, 2005;Munger et al., 1999;Lu et al., 2002;Ludbrook et al., 2003;Munger et al., 1998). Evidence suggests that, under certain circumstances, integrins αvβ3 and αvβ5 activate latent TGFβ1 (Asano et al., 2005a;Asano et al., 2005b;Wipff et al., 2008). Therefore, it is not clear which RGD-binding integrins are relevant to the phenotype of Tgfb1 RGE/RGE mice. To better define the RGD-binding integrins required for TGFβ1 and TGFβ3 activation, we analyzed mice that lack the function of the two best-characterized TGFβ-activating integrins, αvβ6 and αvβ8. We did this by generating mice homozygous for null mutations in both Itgb6 and Itgb8, and by pharmacologically inhibiting integrin αvβ6 in Itgb8 -/mice.
The arginine-glycine-aspartate (RGD)-binding integrins αvβ6 and αvβ8 activate latent TGFβ1 and TGFβ3 in vivo, but it is uncertain whether other RGD-binding integrins such as integrins αvβ5 and αvβ3 activate these TGFβ isoforms. To define the combined role of αvβ6and αvβ8-integrin in TGFβ activation, we analyzed mice lacking function of both integrins by means of gene deletion and/or pharmacologic inhibition. Most Itgb6 -/-;Itgb8 -/embryos die at mid-gestation; those that survive develop cleft palate -as observed in Tgfb3 -/mice. Itgb8 -/mice treated with an anti-αvβ6-integrin antibody develop severe autoimmunity and lack Langerhans cells -similar to Tgfb1-null mice. These results support a model in which TGFβ3-mediated palate fusion and TGFβ1-mediated suppression of autoimmunity and generation of Langerhans cells require integrins αvβ6 and αvβ8 but not other RGDbinding integrins as TGFβ activators.

Results and Discussion
The originally described Itgb8 -/mice (C57BL/6J + 129/Sv genetic background) died at or before birth (Zhu et al., 2002). After being backcrossed to the ICR background, however, Itgb8 -/mice had a median survival of ~2 months despite being born with brain hemorrhage, making it possible to assess the effect of complete loss of integrin αvβ8 in adult mice. At birth, these Itgb8 -/mice were present at the expected Mendelian ratio (data not shown). The external appearance of hemorrhage disappeared after the first week, and brains of 2-week-old mice that were examined histologically showed normal vasculature and no hemorrhage. Itgb8 -/mice arẽ 60% smaller than control littermates (Fig. 1A). Cleft palate was observed once among ~100 newborn knockout mice.
We observed 17 adult Itgb8 -/mice until death occurred naturally or until their morbidity required sacrifice (see Table 1). In most of mice, weakness (particularly in the hindlimbs) developed at 7-10 weeks of age, and the mice became moribund as they lost mobility. This is similar to mice with a conditional deletion of Itgav in the central nervous system; those mice are born with brain hemorrhage but survive and develop axonal deterioration in the spinal cord and cerebellum, leading to ataxia and loss of hindlimb coordination (McCarty et al., 2005). Two Itgb8 -/mice developed rectal prolapse, and two Itgb8 -/mice developed localized infections (pyometra and periorbital abscess in one mouse each); occasional abscesses were also reported in Smad3-null mice (Yang et al., 1999). Histological examination of adult Itgb8 -/mice (lung, heart, liver, spleen, kidney, bladder, brain, thymus, pancreas, stomach, colon) showed no abnormalities aside from the above sporadic abnormalities. Mice with a conditional deletion of Itgb8 in hematopoietic cells or, specifically, in dendritic cells develop a late-onset autoimmune syndrome, characterized by splenomegaly, hepatitis and colitis with death between 4 and 10 months of age (Travis et al., 2007). However, we found no splenomegaly, hepatitis or colitis in adult Itgb8 -/mice, which all died before 4 months of age (Fig. 1B).
To assess the effect of the lack of integrins αvβ6 and αvβ8 on immune function in older mice, we treated Itgb8 -/mice with murine mAb 6.3G9, a specific inhibitor of integrin αvβ6 (Weinreb et al., 2004). To avoid interfering with palate fusion, we started treatment after palate fusion, which occurs at E15, and protocols differed regarding the time the treatments were started (see Table 2). All injections were 10 mg/kg, a dose that produces an essentially complete inhibition of integrin αvβ6 (Horan et al., 2007;Puthawala et al., 2008).
Journal of Cell Science 122 (2)  Weakness in hind limbs S Itgb8 -/mice were observed until death or until sacrifice was required because of morbidity (which was most often due to poor mobility and inability to feed). D, natural death; S; sacrificed.
When integrin αvβ6 inhibition was initiated before birth (E16.5) or at birth [referred to as postnatal day 0 (P0)] (see Table 2, groups 1 or 2, respectively), mice became moribund or died between days 11 and 26 (Fig. 3A). In all cases, littermates that were non-knockout (non-KO), i.e. Itgb8 +/or Itgb8 +/+ , and had been treated with the same antibody regimen appeared healthy at the time of the demise of the 6.3G9-treated Itgb8 -/mice. All 6.3G9-treated Itgb8 -/mice had extensive mononuclear cell infiltrates in heart, lung, liver, stomach and pancreas (Fig. 3B). Similarly to Tgfb1 -/and Tgfb1 RGE/RGE mice, they also lacked LCs, a subtype of dendritic cell in the epidermis (Fig. 3C) (Borkowski et al., 1996;Yang et al., 2007). By contrast, non-KO littermates that had been treated with 6.3G9 and were sacrificed in parallel had epidermal LCs and showed no signs of inflammation except in the lungs of half the treated animals (consistent with loss of integrin αvβ6 function). The survival of these 6.3G9-treated Itgb8 -/mice ( Fig. 3A) was slightly shorter than that reported for Tgfb1 -/and Tgfb1 RGE/RGE mice (Kulkarni et al., 1993;Yang et al., 2007), most of which die between 3 and 4 weeks of age.
In four additional Itgb8 -/mice, 6.3G9 treatment was started on P3 (Table 1, group 3). The survival of these mice was indistinguishable from that of untreated Itgb8 -/mice (Fig. 3A). Inflammatory changes were less marked: whereas all four mice had lung inflammation, consistent with inhibition αvβ6 integrin, only two of four had gastric inflammation or inflammation of the liver, and none had inflammation of the pancreas or heart.
Mice that lack Foxp3, a transcription factor that specifies regulatory T cell lineage (Treg), develop autoimmunity similar to that in Tgfb1 -/mice, which suggests a role for TGFβ1 in Treg development and/or function (Fontenot et al., 2003). Thymusderived Tregs are generated in a TGFβ1-independent manner because they are present in Tgfb1 -/mice (Marie et al., 2005) and in mice whose type 2 TGFβ receptor (TGFβR2) is conditionally deleted in CD4+ cells (Marie et al., 2006). By contrast, TGFβ1 deficiency causes increased proliferation and decreased survival of peripheral or 'adaptive' Tregs (Sakaguchi, 2004;Wan and Flavell, 2007;Marie et al., 2006;Rubtsov and Rudensky, 2007).
We quantified Tregs in 6.3G9-treated Itgb8 -/mice and in non-KO control mice at P8, when significant numbers of thymic Tregs are found in wild-type mice (Fontenot et al., 2005). Inhibition of αvβ6 integrin was initiated before birth, as in Group 1 (see Table  2). Thymic and splenic Tregs in 6.3G9-treated Itgb8 -/mice were not reduced, as also observed in Tgfb1 RGE/RGE mice (Fig. 3D) . The 6.3G9-treated Itgb8 -/mice did not show histological evidence of inflammation at P8 (data not shown). P17 mice were evaluated in studies that demonstrated reduced numbers of splenic Tregs in TGFβ1-deficient mice (Marie et al., 2006;Rubtsov and Rudensky, 2007), which might explain the different result in our study.
Travis and colleagues have shown that conditional loss of αvβ8 integrin in all leukocytes or, specifically, in dendritic cells causes autoimmunity (Travis et al., 2007). These experiments were performed by crossing mice with a floxed Itgb8 allele with mice that expressed Cre recombinase under the control of either the Vav1 or CD11c (Itgax) promoter. In both cases, a wasting syndrome with inflammation in the colon and the portal triads of the liver began at ~4-5 months of age. Death occurred naturally or when their morbidity required sacrifice, in most cases after 6 months of age. In unpublished work, this group also generated Itgb6 -/-;(Vav1cre)Itgb8 fl/fl mice. These mice had a similar wasting and autoimmune syndrome, but with an accelerated course, with death occurring by 23 weeks (personal communication, Dean Sheppard, UCSF, San Francisco, CA, and Mark Travis, University of Manchester, UK). Neither of these inflammatory phenotypes is as severe as in TGFβ1deficient or 6.3G9-treated Itgb8 -/mice. Together, these observations and our findings suggest that expression of αvβ8 integrin by both dendritic cells and non-immune cells, in combination with αvβ6 integrin, is needed to prevent excessive  Mice groups 1, 2 and 3, three protocols used to treat mice with either mAb 6.3G9 against αvβ6 integrin or mAb 1E6 (control). Group 1 mice were treated prenatally by injecting pregnant females with 10 mg/kg of 6.3G9 (or control antibody) on E16.5 and E18.5, and then postnatally beginning on P3 and continuing weekly. Group 2 mice were first injected on the day of birth (P0) and then weekly. Group 3 mice were first injected on P3 and then weekly. lymphocyte activation. αvβ8 integrin is widely expressed on many non-immune cell types, such as neurons, fibroblasts and epithelial cells (Zhu et al., 2002;Araya et al., 2007;Nishimura et al., 1998;Proctor et al., 2005).
6.3G9-treated Itgb8 -/mice have a shorter survival time than Tgfb1 -/and Tgfb1 RGE/RGE mice. The reasons for this are not known. The genetic background might be responsible. In addition, the mice might be physiologically impaired because of brain hemorrhage and low body weight. It is also possible that TGFβ3, which is activated by αvβ6and αvβ8-integrins, subtly ameliorates the immune phenotype of TGFβ1-deficient mice. If so, this rescue effect might be absent in 6.3G9-treated Itgb8 -/mice, because αvβ6 and αvβ8 can activate TGFβ3. In regard to this hypothesis, it is of interest to consider mice with a T-cell-specific deletion of the type II TGFβ receptor, because in these mice T cells would be unresponsive to all three TGFβ isoforms. In one of two reports on such mice, (CD4-Cre)Tgfbr2 fl/fl mice appear to have a slightly more aggressive inflammatory process than do Tgfb1-null mice (Li et al., 2006).
The finding that the inhibition of integrin αvβ6 in Itgb8 -/mice initiated at P3 results in a much milder inflammatory syndrome than when inhibition is initiated at P0 or before birth remains unexplained, but might relate to dynamics of Treg development in the first few days of life. Classic experiments have shown that thymectomy on the third day of life (P2) leads to autoimmunity, whereas thymectomy on the first or seventh day (P0 or P6, respectively) does not, suggesting suppressor cells are released from the thymus later than autoreactive cells (reviewed by Shevach et al., 2001). Rudensky's group has demonstrated a rapid increase in thymic Tregs in the first few days of life, with the biggest increase occurring at days 3-4 (Fontenot et al., 2005). Previous experiments have indicated that thymic Treg-cell development is largely independent of TGFβ (Marie et al., 2005;Marie et al., 2006). However, a recent report that describes mice in which TGFβ signaling is conditionally ablated in T cells by means of floxed Tgfbr1 alleles and Cre recombinase expressed under control of the CD4 promoter, has demonstrated that thymic Treg development is dramatically blocked at days P2-P4; by 1 week, thymic Treg-cell levels had normalized owing to expansion of Tregs in an IL2dependent manner (Liu et al., 2008). CD4+CD25+Foxp3+ Tregs from these mice showed an activated phenotype and were defective in in-vitro suppressor assays. Thus, our results might indicate that sufficient numbers of normally functioning Tregs are released by P3 to suppress inflammation, but if TGFβ signaling is eliminated earlier, normal Tregs are not released and severe inflammation ensues.
The present findings complement our recent report on the phenotype of mice that lack integrin-responsive TGFβ1 and totally lack TGFβ3 (Tgfb1 RGE/RGE ;Tgfb3 -/-). These mice showed the expected abnormalities seen in Tgfb1 RGE/RGE or Tgfb3 -/mice, as well as one new abnormality not present in any TGFβ mutant mouse: Journal of Cell Science 122 (2)  (n=17), Itgb8 -/mice with 6.3G9 treatment starting at P3 (n=4), Itgb8 -/mice with 6.3G9 treatment starting at P0 (n=8), and Itgb8 -/mice with 6.3G9 treatment starting at E16.5 (n=12). (B) Hematoxylin and eosin staining of liver, lung, stomach and heart of representative Itgb8 -/and littermate non-KO control mice treated with 6.3G9 starting at E16.5 and sacrificed at 19 days. (C) Itgb8 -/mice treated with 6.3G9 starting at E16.5 lack Langerhans cells (LCs). Representative examples of epidermal sheets stained with an antibody that recognizes LCs. (D) Tregs isolated from thymus and spleen were identified by staining for FOXP3. Cells are from control mice, Itgb8 -/mice, and control and Itgb8 -/mice treated with 6.3G9 starting at E16.5. (n=3-4 for thymus, n=3 for spleen). Error bars represent the s.e.m. Scale bar: 100 μm. abnormal brain vascular morphogenesis that appears to be identical to that in mice that lack αvβ8 integrin (Zhu et al., 2002;Bader et al., 1998). Here, we show that mice that lack the function of both αvβ6and αvβ8-integrin display abnormalities not seen in either of the two single-integrin-knockout mice that recapitulate the major abnormalities in Tgfb3-null mice (cleft palate) and Tgfb1-null mice (severe inflammation and lack of LCs). Taken together, these findings reveal a system in which two TGFβ isoforms and two TGFβ activators act cooperatively in several developmental processes.

Mice
Itgb8 -/mice were from Louis Reichardt (UCSF, San Francisco, CA) and had been backcrossed for at least six generations onto the ICR background (Charles River Laboratories, Boston, MA). Itgb6 -/mice on a C57BL/6J background were obtained from Dean Sheppard (University of California, San Francisco, CA) and were backcrossed three generations onto the ICR background before crossing with Itgb8 -/mice.

Treatments
Group 1 mice were treated at E16.5 and E18.5 by injecting pregnant females with 10 mg/kg 6.3G9 (or the control IgG1 monoclonal antibody 1E6) (Puthawala et al., 2008;Hahm et al., 2007;Weinreb et al., 2004). At birth, Itgb8 -/mice were identified by the presence of brain hemorrhage (confirmed by genotyping). Injections of 10 mg/kg into Itgb8 -/mice and littermate controls continued at P3 and then once a week. Group 2 mice were first injected on P0 and then once a week. Group 3 mice were first injected on P3 and then once a week. Mice were observed daily and sacrificed if moribund. Tissues were collected at death for histology. Two types of control mice were used for this experiment: control non-KO littermates that received the same injections of 6.3G9 as the Itgb8 -/mice, and Itgb8 -/mice treated with control IgG1 mAb. Non-KO mice receiving 6.3G9 were sacrificed when their 6.3G9treated KO littermates died; no non-KO mice that had received 6.3G9 died before similarly treated KO littermates. The NYU School of Medicine animal care committee approved all procedures, which conformed to NIH guidelines.

LC immunostaining
LCs in epidermal sheets were detected as described previously (Thomas et al., 2001;Yang et al., 2007). LC staining was performed on epidermal sheets isolated from backs of 6.3G9-treated Itgb8 -/mice at time of death and compared with samples from littermate non-KO control mice sacrificed at the same time. Epidermal sheets from seven Itgb8 -/and eight non-KO mice treated with 6.3G9 were stained. 6.3G9 treatment was started at E16.5.

Flow cytometric analysis
Single-cell suspensions from spleen or thymus (after RBC lysis) were incubated with anti-CD4-FITC antibody with Mouse BD Fc Block (BD Biosciences), washed, incubated in permeabilization/fixation buffer, re-incubated with Fc Block and anti-Foxp3-PE antibody (eBioscience), washed, and analyzed using a FACscan flow cytometer (Becton Dickinson). This work was funded by NIH grant R01 HL063786 and an Irma T. Hirschl Scholar Award from the Irma T. Hirschl/Monique Weill-Caulier Trusts (both to J.S.M.). We thank Paola Mita for help with Langerhans cell staining and Ezra Dweck for helpful discussions and technical assistance.