Lack of evidence for a role of anthrax toxin receptors as surface receptors for collagen VI and for its cleaved-off C5 domain/endotrophin.

The microfibril-forming collagen VI is proteolytically cleaved and it was proposed that the released C-terminal Kunitz domain (C5) of the α3 chain is an adipokine important for tumor progression and fibrosis. Designated "endotrophin," C5 is a potent biomarker for fibroinflammatory diseases. However, the biochemical mechanisms behind endotrophin activity were not investigated. Earlier, anthrax toxin receptor 1 was found to bind C5, but this potential interaction was not further studied. Given the proposed physiological role of endotrophin, we aimed to determine how the signal is transmitted. Surprisingly, we could not detect any interaction between endotrophin and anthrax toxin receptor 1 or its close relative, anthrax toxin receptor 2. Moreover, we detect no binding of fully assembled collagen VI to either receptor. We also studied the collagen VI receptor NG2 (CSPG4) and confirmed that NG2 binds assembled collagen VI, but not cleaved C5/endotrophin. A cellular receptor for C5/endotrophin, therefore, still remains elusive.

Introduction

Collagen VI is a microfibrillar protein expressed in most tissues and is there involved in anchoring large interstitial structures and cells (Cescon et al., 2015). It is an unusual collagen in that it is predominantly composed of von Willerbrand factor type A (VWA) domains and contains only a short collagenous domain. The VWA domains are crucial for collagen VI assembly (Lamandé et al., 1998) and for the interaction with other extracellular matrix proteins (Wiberg et al., 2003). The major form of collagen VI consists of α1, α2, and α3 chains, but in some cases the α3 chain is replaced by an α4, α5 or α6 chain (Maaß et al., 2016). The α3-α6 chains are larger than the α1 and α2 chains and possess up to 10 tandem VWA domains in their N-terminal region.

The assembly of collagen VI microfibrils is a complex multi-step process. After the folding of heterotrimeric monomers (α1α2α3) these form antiparallel dimers (Knupp and Squire, 2001) that finally assemble laterally to tetramers and are then secreted. Microfibrils form by intercalating end-to-end associations that are accompanied by proteolytic processing of a large C-terminal globular region of the α3 chain (Beecher et al., 2011) which is absent in mature microfibrils (Beecher et al., 2011; Heumuller et al., 2019). This results in the release of the short most C-terminal C5 domain, a Kunitz domain, which was also termed endotrophin and proposed to be an adipokine (Park and Scherer, 2012; Sun et al., 2014). Recently it was shown that endotrophin can be released by BMP1 (Heumuller et al., 2019) or MMP14 (Li et al., 2020). However, the free endotrophin fragment is much less abundant than larger C5-containing fragments (Heumuller et al., 2019).

Collagen VI is linked to several diseases and, most prominently, mutations in COL6A1, COL6A2, or COL6A3 lead to myopathies (Bonnemann, 2011). Bethlem myopathy and Ullrich congenital muscular dystrophy are the most frequent, inherited both in an autosomal dominant and recessive manner and represent, respectively, the mild and severe end of the disease spectrum (Bonnemann, 2011). Moreover, collagen VI mutations can cause autosomal dominant limb-girdle muscular dystrophy (Scacheri et al., 2002) and autosomal recessive myosclerosis (Merlini et al., 2008). Recessive mutations in COL6A3 have been also linked to isolated dystonia (Zech et al., 2015).

The pathomechanisms leading to collagen VI-related myopathies have been studied in mouse models. Mice lacking Col6a1 display a phenotype reflecting Bethlem myopathy owing to an abnormal opening of the mitochondrial pore and disturbed autophagy in skeletal muscle (Grumati et al., 2010; Irwin et al., 2003). However, it is still unclear how the lack of extracellular collagen VI leads to an intracellular phenotype. Integrins (Pfaff et al., 1993), the cell surface proteoglycan NG2 (Nishiyama and Stallcup, 1993), and the anthrax toxin receptors 1 (Nanda et al., 2004) and 2 (Burgi et al., 2017) have been reported to be collagen VI receptors. Interestingly, the latter two have been linked to collagen VI in fibroproliferative diseases. Mutations in the anthrax toxin receptors cause GAPO syndrome (anthrax toxin receptor 1, ANTXR,1 also known as TEM8) (Stranecky et al., 2013) or infantile hyaline fibromatosis (anthrax toxin receptor 2, ANTXR2, also known as CMG2) (Lee et al., 2005) that are both accompanied by excessive deposition of collagen VI (Besschetnova et al., 2015; Glover et al., 1991). Primary fibroblasts isolated from Antxr1 deficient mice display an up-regulated expression of collagen VI and collagen I pointing to the antiproliferative function of this receptor (Besschetnova et al., 2015). It is, however, unclear if the loss of binding of collagen VI to Antxr1 interrupts a feedback loop. Moreover, even a conditional knock-out of Antxr1 in vascular endothelial cells causes severe skin fibrosis and increased expression of collagen VI (Besschetnova et al., 2015), indicating a complex cross-talk between tissues also involving VEGF signaling (Besschetnova et al., 2015). The fibrotic condition could also be the consequence of a reduced degradation, as the activity of the matrix-degrading enzyme MMP2 is strongly decreased (Besschetnova et al., 2015). Mice deficient in Antxr2 show a massive deposition of collagen VI in the uterus leading to infertility which is rescued in Antxr2/Col6a1 double deficient mice. In contrast to in Antxr1 deficient mice, collagen VI accumulation is probably not caused by a reduced degradation by metalloproteinases in Antxr2 deficient mice. However, there is evidence that Antxr2 is involved in endocytosis and lysosomal degradation of collagen VI (Burgi et al., 2017). In contrast to Antxr1, which was shown to bind to the C5 domain of the α3 chain (Nanda et al., 2004), it was proposed that Antxr2 binds to the triple helical domain of collagen VI (Burgi et al., 2017).

For endotrophin, the cleaved-off C5 domain, growing evidence has been provided that points to important roles in tumor progression, fibrosis, inflammation, and insulin resistance (Park and Scherer, 2012; Sun et al., 2014). However, the cellular receptor/s responsible for mediating these effects are still not known. A proposed interaction between the Antxr1 and the C5 domain was originally based on a yeast two-hybrid screen and co-immunoprecipitation of overexpressed Antxr1 and C5, but was not studied in detail (Nanda et al., 2004). Recently, serum levels of endotrophin have been shown to be valuable biomarkers for several pathological conditions (Nielsen et al., 2021). The combination of a proposed important physiological role of endotrophin and its growing importance as a biomarker led us to study the interaction with its proposed receptors in more detail. Moreover, the second anthrax toxin receptor, Antxr2, which shares high homology with Antxr1, was shown to act as a receptor for collagen VI, even if it remains unknown which part of the collagenous domain is mediating such an interaction (Burgi et al., 2017). Therefore, we revisited the interaction of the C5 domain (endotrophin) and of collagen VI with the anthrax toxin receptors and present data that substantially challenge current concepts.

Results

The VWA domains of the anthrax toxin receptors do not bind to the proteolytically released C5 domain of the collagen VI α3 chain

As the cellular receptor/s mediating the biological activity of endotrophin — the cleaved-off C5 domain of collagen VI — is/are not known, we revisited the well-accepted binding of C5 to Antxr1 (Nanda et al., 2004) by performing direct protein-protein interaction studies. Such studies, to our knowledge, have not been performed earlier. For that purpose, we expressed and purified recombinant murine C5 and the VWA domains of both Antxr1 and Antxr2 in HEK293T cells (Figures 1A and 1B). We decided to use the anthrax toxin receptor VWA domains as they are proposed to be the major extracellular ligand-binding regions of the two receptors and known to be sufficient to bind to the anthrax toxin (Bell et al., 2001; Sergeeva and van der Goot, 2019). In our analysis, we included the VWA domain of Antxr2 even if it was proposed that collagen VI binds to this receptor via its collagenous domain (Burgi et al., 2017). As a positive control, we used the protective antigen (PA) subunit of the anthrax toxin, the pathogenic and well-characterized ligand of the anthrax toxin receptors (Bradley et al., 2001). To be sure that recombinant C5 was correctly produced and purified we performed several quality controls on the protein. First, we excluded that the purified protein may form aggregates and confirmed that C5 was exclusively present in solution in a monomeric form by performing size exclusion chromatography (Figure S1A). To further prove that recombinant C5 was correctly folded, we measured CD spectra and analyzed the mobility of the protein using SDS-PAGE under reducing and non-reducing conditions. Both assays confirmed the proper folding of recombinant C5 (Figures S1B and S1C) (Nagy et al., 2003). Finally, we performed deglycosylation experiments and found that, in line with an in silico prediction (Steentoft et al., 2013), C5 is O-glycosylated, as it can be deglycosylated by trifluoroacetic acid and O-glycosidase but not by PNGase F (Figures S1D and S1E). By performing direct interaction studies using surface plasmon resonance, we confirmed that both VWA domains coupled with the sensor chip can bind to PA in a cation-dependent manner, with the VWA domain of Antxr2 showing a higher binding affinity than the same domain of Antxr1 (KDs: 6.45 ± 1.08 nM and 182 ± 64.3nM, respectively) as previously reported (Liu et al., 2009). Surprisingly, using the same assay under the same conditions, we could not detect any direct interaction between the VWA domain of Antxr1 and the C5 domain. Similarly, also the VWA domain of Antxr2 did not show any binding to C5 (Figure 1C). Lack of binding was confirmed when C5 was immobilized on the sensor chip and the VWA domains were used as analytes (Figure S2).

Figure 1

PA, but not the C5 domain of the collagen VI α3 chain binds to the VWA domains of anthrax toxin receptors

(A) Schematic representation of the domain organization of collagen VI ⍺ chains. C5 is found at the most C-terminal part of the ⍺3 chain, from which it can be cleaved-off by the proteases BMP1 or MMP14.

(B) Coomassie Blue stained SDS/polyacrylamide gel with the recombinant proteins used for the binding experiments. The upper C5 band is owing to O-glycosylation of the recombinant protein (see Figure S1C).

(C) SPR sensorgrams obtained either in presence of EDTA or CaCl2, of the interaction between murine Antxr1 or Antxr2 VWA-domains (immobilized on the sensor chips in similar quantities) and Protective Antigen (PA) or murine C5 (flown over as soluble ligands) in a dilution series from 0 to 320 nM. Ligand binding to the immobilized proteins on the chip is shown as response units on the y axis (n = 3).

Plasma membrane located, full-length anthrax toxin receptors do not mediate cell binding to the recombinant C5 domain

The fact that we could not detect direct binding of the C5 domain to the principal binding domain of the Antxr1 was surprising and in conflict with the results obtained by others (Nanda et al., 2004). Our contradictory findings made it necessary to unequivocally confirm this result by using alternative methods employing a biologically more relevant setting. Therefore, we generated HEK293T cell lines stably overexpressing full-length Antxr1 or -2 on the plasma membrane to test if they support cell binding to the C5 domain. To first validate the system, we treated the cells for 1 h at 37 °C with 5 μg/mL PA in a serum-free medium. When using a PA-specific antibody, staining of control HEK293T cells that had been transfected with an empty vector did not reveal any immunoreactivity. In contrast, when PA was added to Antxr1 or Antxr2 expressing cells we could detect a strong PA signal colocalizing with the two receptors on the cell surface (Figure 2A). However, when the same cells were treated with an equivalent concentration of the C5 domain and stained with a C5-specific antibody, we could not detect any positive signal in control cells or on cells overexpressing the two anthrax toxin receptors (Figures 2B and 2C). To further confirm that Antxr1 and Antxr2 act as surface receptors for PA but not for C5, we once more treated HEK293T cells with the two ligands and analyzed the amount of cell-bound protein by immunoblotting. Consistent with our immunofluorescence results, we could recover an increased amount of PA from the cell lysate fractions upon Antxr1 or Antxr2 overexpression, while we found that the C5 domain was only present in the culture medium and not in the cell layer of the three cell lines, irrespective of the expression of the anthrax toxin receptors (Figure 2D). Altogether, these results show that expression of the full-length anthrax toxin receptors mediates the binding of PA, but not of the C5 domain of the collagen VI α3 chain, to HEK293T cells. To exclude the possibility that C5 does not bind to the cell surface of HEK293T cells because of the lack of a co-receptor, we performed similar experiments using CHO-K1 cells, as it is known that these can bind and uptake the anthrax toxin (Bradley et al., 2001). For this purpose, we generated CHO-K1 cell lines overexpressing full-length Antxr1 and Antxr2 receptors and found that, while the overexpression of the two receptors greatly increased the capacity of PA to bind to the cells, they were completely unable to mediate the binding of C5 to the cell surface (Figure S3). Importantly, using the same assays (immunostaining or immunoblot) we could not detect any interaction of the recombinant C5 domain with the cell surface of mouse primary dermal fibroblasts, MC3T3-E1, RPE-1, U2OS, MCF-7 or MDA-231-MB cells (not shown).

Figure 2

Anthrax toxin receptors support the binding of PA, but not of C5 to the surface of HEK293T cells

(A and B) Confocal microscopy pictures of control and His-tagged Antxr1 or Antxr2 overexpressing HEK293T cells labeled with the indicated antibodies. Cells were treated with 5 μg/mL PA (A) or human C5 (B) in a serum-free medium for 1 h at 37 °C, washed, and then fixed. Similar experiments were also carried out using recombinant murine C5 with identical results (not shown).

(C) Quantification of the mean fluorescence intensity of the immunolabeled cells as in A and B. One-way ANOVA, a p value of ≤0.05 was considered significant.

(D) Control and anthrax toxin receptors overexpressing HEK293T cells were treated with 1 μg/mL PA or murine C5 for 1 h at 37 °C and then washed with PBS before collecting the culture supernatants (S) and harvesting the cell layers (CL). Samples were immunoblotted using the indicated antibodies. Each experiment was performed at least three times with similar results. Scale bars: 10 μm.

The cell surface proteoglycan NG2, and not the anthrax toxin receptors, supports cell binding to collagen VI, but not to the isolated C5 domain

Our previous experiments did not provide evidence supporting the role of anthrax toxin receptors as cell-surface receptors for the recombinant C-terminal C5 domain (endotrophin) of the collagen VI α3 chain. We, therefore, performed further experiments with a more physiological source of C5. For this purpose, we collected the conditioned medium of MC3T3-E1 osteoblasts grown in the presence of ascorbate, as we have previously found that these cells secrete collagen VI tetramers and form an abundant collagen VI microfibrillar network (Schiavinato et al., 2021). Next, we treated control and anthrax toxin receptor overexpressing HEK293T cells with the conditioned medium and stained them with the C5 antibody. As a positive control, we also generated HEK293T cells stably expressing the integral plasma membrane proteoglycan NG2 (CSPG4), an established collagen VI receptor (Stallcup et al., 1990). C5-containing collagen VI molecules present in the osteoblast medium bound scarcely to control HEK293T cells and, similar to what we observed using the recombinant C5 domain, overexpression of the two anthrax toxin receptors did not increase the interaction of collagen VI with the cells. In contrast, using the same antibody, we detected an intense signal on the cell surface of NG2-expressing cells (Figures 3A and 3B). The C5 domain is progressively cleaved-off from the α3 chain after the secretion of collagen VI tetramers (Aigner et al., 2002; Heumuller et al., 2019). Therefore, the signal detected on the NG2-expressing cells could represent either the released C5 domain, C5-containing collagen VI molecules, or fragments thereof. To distinguish between these possibilities, we stained the same cells with an antibody (α3N) specific for the N-terminal region of the α3 chain that is not cleaved from the rest of the molecule. Using this antibody, we observed no or very scarce reactivity on control cells and on cells expressing the anthrax toxin receptors, while NG2-expressing cells were intensively labeled with a signal that colocalized completely with that obtained with an antibody against C5 (Figures 3A and 3B). Another co-staining experiment with the different cell lines, using the α3N antibody together with an antibody specific to the α1 chain of collagen VI (α1C), gave similar results (Figure S4). Next, we analyzed extracts of the cells incubated with the conditioned medium by immunoblotting. Using an antibody specific for the α1 chain we found that only NG2-expressing cells showed an increased binding. Strikingly, using the C5 antibody on the same samples, we could confirm that the cells expressing NG2, but not Antxr1 or Antxr2, showed enhanced binding to the fully assembled C5-containing collagen VI tetramers, while the soluble, cleaved-off C5 domain did not bind to any of the cell lines (Figure 3C).

Figure 3

NG2, but not anthrax toxin receptors, mediates the binding of collagen VI to the cell surface

(A) Confocal microscopy pictures of control, His-tagged Antxr1, Antxr2 or NG2 overexpressing HEK293T cells that were incubated for 1 h at 37 °C with collagen VI-containing serum-free conditioned medium derived from MC3T3-E1 osteoblasts. After washing, cells were fixed and immunolabeled with the indicated antibodies.

(B) Quantification of the mean fluorescence intensity of the immunolabeled cells as in A.

(C) Control HEK293T cells and cells expressing the indicated receptors were treated with the conditioned medium as in A. After washing, cell lysates were harvested and analyzed by immunoblotting using the indicated antibodies to detect collagen VI moieties bound to the cell layers. For the detection of the ⍺1 chain (top panel) by immunoblot, samples were resolved on a 10% acrylamide gel under reducing conditions; for the cleaved-off C5 domain (middle panel) a 15% acrylamide gel was used under non-reducing conditions. For immunoblotting of collagen VI tetramers (bottom panel), samples were submitted to composite agarose/acrylamide gel electrophoresis under non-reducing conditions and immunoblotted with the C5 antibody. β-actin was used as a loading control.

(D) Cell lysates of control HEK293T cells and of cells expressing His-tagged Antxr1 or Antxr2 were incubated with Ni-NTA agarose beads and with either collagen VI containing conditioned medium or PA spiked into the same medium. After washing the beads, bound proteins were eluted in the SDS-containing sample buffer and analyzed by immunoblotting with the indicated antibodies.

(E) Strep-Tactin-coupled Sepharose beads were incubated with recombinant strep-tagged NG2-ectodomain (Ecto) or without any protein (Ctrl). The beads were afterward incubated with collagen VI-containing conditioned medium. After washing, the proteins bound to the beads were eluted in the SDS-containing sample buffer and immunoblotted with the indicated antibodies.

(F) Confocal microscopy pictures of NG2-expressing HEK293T cells incubated for 1 h with either recombinant C5 or collagen VI-rich conditioned medium derived from MC3T3-E1 osteoblasts, or conditioned medium derived from collagen VI α2-chain deficient MC3T3-E1 osteoblast cultures. After washing, cells were labeled with the indicated antibodies.

(G) Quantification of the mean fluorescence intensity of labeled cells as in (E). One-way ANOVA, a p value of ≤0.05 was considered significant. CM, conditioned medium; NT, not treated. Scale bars: 10 μm.

As these results are in disagreement with previous studies proposing a role for Antxr2 as the collagen VI receptor, we sought to further verify our findings. Toward this aim, we performed pull-down assays using full-length anthrax toxin receptors and a conditioned medium of MC3T3-E1 cultures. However, despite the fact that we performed the assay multiple times under different conditions, we could not pull-down collagen VI or the cleaved-off C5 domain with the anthrax toxin receptors, while PA was consistently pulled down (Figure 3D). On the other hand, we could co-precipitate collagen VI from MC3T3-E1 conditioned medium in a similar assay by using the recombinant ectodomain of NG2 (Tillet et al., 1997) (Figure 3E). Our results indicate that collagen VI tetramers bind only scarcely to HEK293T cells and that NG2, but neither Antxr1 nor −2, is able to endow the cells with increased binding activity. It has been previously shown that NG2 is able to interact with both pepsin-solubilized collagen VI, lacking its globular, non-collagenous domains, and the recombinant collagen VI α2 chain while the N-terminal region of the α3 chain showed no binding (Burg et al., 1996). However, it is not known if the C-terminal part of the α3 chain may also participate in the interaction. Therefore, to determine if the isolated C5 domain could be a ligand for NG2, and use this receptor to convey signals into the cell, we treated in parallel NG2-expressing cells with either recombinant C5 domain or the conditioned medium from osteoblast cultures or conditioned medium from collagen VI α2 deficient osteoblasts and stained them with the C5 antibody. We found that while cells treated with the conditioned medium showed a strong signal on the cell surface, no staining was present on the cells treated with the recombinant C5 protein, showing that the C5 domain alone does not possess the ability to bind to NG2. Moreover, the conditioned medium derived from the α2 deficient osteoblast cultures that contain only single α1 and α3 chains of collagen VI, but no assembled tetramers, showed no significant binding of collagen VI molecules to the cell surface, implying that the α2 chain or a full tetramer conformation is needed to allow efficient binding to NG2. (Figures 3F and 3G).

Our results so far have failed to demonstrate the role of anthrax toxin receptors as surface receptors for collagen VI or for C5. Next, we sought to investigate if anthrax receptors may interfere with collagen VI production in a cell-autonomous manner when expressed in collagen VI-producing cells. Therefore, we generated MC3T3-E1 osteoblasts with inducible expression of His-tagged full-length Antxr1 and -2. Again, we included NG2 in our experiments as it has been previously shown that expression of NG2 in collagen VI producing cells caused retention of collagen VI on the cell surface (Nishiyama and Stallcup, 1993). Cells were grown in the presence of ascorbate for 5 days with or without doxycycline and then analyzed by immunofluorescence. Upon doxycycline treatment expression of the three receptors was strongly induced in MC3T3-E1 cells as shown by staining with an antibody against the His-tag. To study the effect on collagen VI deposition, we stained the cells with the C5 and α3N antibodies. The α3N antibody mainly stained extracellular collagen VI fibers incorporated into the ECM of the MC3T3-E1 osteoblasts, while the C5 antibody gave a more prominent intracellular signal, as previously described also in other cell types (Mayer et al., 1994). Upon expression of the anthrax toxin receptors, we could not detect any significant differences in collagen VI localization and deposition. Moreover, no colocalization between anthrax toxin receptors and C5 or α3N was evident. On the other hand, NG2 expression caused marked retention of collagen VI on the cell surface, where the C5 and α3N signals were largely overlapping with the signal for NG2 (Figures 4A and 4B).

Figure 4

Expression of NG2, but not of anthrax toxin receptors, causes retention of type VI collagen on the surface of MC3T3-E1 cells

(A) Confocal microscopy images of His-tagged Antxr1, Antxr2, or NG2 overexpressing MC3T3-E1 cells labeled with the indicated antibodies. Cells were grown for 5 days in the presence of ascorbate with or without doxycycline to induce the expression of the receptors.

(B) Quantification of the mean C5 fluorescence intensity of labeled cells as in (A). Mean fluorescence was measured in receptor-expressing cells and related to area size. One-way ANOVA, a p value of ≤0.05 was considered significant. Scale bar: 100 μM.

Discussion

C5, also named endotrophin, the C-terminal cleavage product of the collagen VI α3 chain has received a great deal of attention, as it was proposed to act as an adipokine that is involved in tumorigenesis and fibrogenesis. Moreover, endotrophin has been shown to be a valuable biomarker for fibrotic conditions (for review see (Williams et al., 2021)), in this context also named PRO-C6. Long before the C5 domain was recognized as an adipokine and biomarker, it was shown that it binds to Antxr1 (Nanda et al., 2004). To date, no other receptor for endotrophin has been identified (Williams et al., 2021) and it is remarkable that although the physiological function of endotrophin has been extensively studied the interaction between endotrophin and a receptor that transmits the signal into the cell has not been further shown. The still enigmatic function of endotrophin is in part reminiscent of that of the adipokine adiponectin, where more than 25 years after its discovery (Scherer et al., 1995) it is still not clear how and why it exerts multiple beneficial effects on various tissues and organs (Maeda et al., 2020). However, in contrast to endotrophin adiponectin-binding partners, such as AdipoR1/2 (Yamauchi et al., 2003) or T-cadherin (Hug et al., 2004), have been identified.

Therefore, after we showed that endotrophin (C5) is released by BMP1 (Heumuller et al., 2019) we carefully searched for receptor-ligand interactions to gain better insight into the mechanism behind the proposed physiological responses to endotrophin. However, despite extensive attempts, we were not able to find any evidence for receptor interaction, although we applied not only methods to detect direct binding of recombinant proteins but also performed cell culture-based interaction studies. It remains uncertain how the earlier results indicating a binding of the C5 domain to Antxr1 were obtained. The understanding and reproduction of the former experiments are hampered by the fact that the exact borders of the C5 construct were not given by Nanda et al. (2004). A 210 bp cDNA clone encoding the C-terminal end of the collagen VI α3 chain is mentioned. If this is consistent with a fragment encompassing the last 70 amino acid residues of the α3 chain, this fragment would start with E3108TDI. It would be slightly shorter than endotrophin released by BMP1 which starts with T3101EPL or endotrophin released by MMP14 which starts at A3087 RSA. Nevertheless, also the C5 fragment used by Nanda et al. most likely contained all six cysteine residues that are required for proper folding of a Kunitz domain (Zweckstetter et al., 1996). However, the yeast-two-hybrid screening that was used for the identification of the interaction may not be suited to study protein-protein interactions that naturally occur in the extracellular space. Structurally important posttranslational modifications of extracellular proteins, in particular the formation of disulphide bonds, are hindered by the reducing conditions in the nucleus where the interaction in yeast-two-hybrid assays actually takes place. The impact of reduction on the C5 structure is evident by the difference in its mobility on SDS-PAGE (Figure S1). Finally, the immunoprecipitation experiment shown to verify the yeast-two-hybrid result was performed without a positive control, using only myc-tagged proteins and an anti-TEM8 antibody, while the reverse pulldown using a C5 antibody was not performed (Nanda et al., 2004). The result is, therefore, of uncertain validity. The fact that we did not observe the binding of the C5 domain to a variety of cultured cells makes it difficult to understand the mechanism behind its proposed function as an adipokine. In particular, we could not see a binding or uptake of endotrophin by MCF-7 or MDA-MB-231 cells. This is remarkable as these cell lines have been shown to display an up-regulation of the epithelial-mesenchymal transition markers Twist and Snail, and an enhanced cell migration upon endotrophin exposure (Bu et al., 2019). In line with our negative results, there are conflicting data on if collagen VI affects MDA-MB-231 cell migration not through endotrophin, but via interaction of the triple helical part of collagen VI with NG2 (Wishart et al., 2020).

Nevertheless, when we used NG2 as a positive control we could confirm that NG2 binds to collagen VI, but also show that NG2 does not bind to endotrophin. This differs from our results on the proposed role of the Antxr2 in the uptake of collagen VI. Although it is evident that this receptor, like Antxr1, does not bind to endotrophin, it was surprising that we also could not detect the binding of Antxr2 to collagen VI while it was proposed that Antxr2 is involved in the endocytosis and lysosomal degradation of assembled collagen VI tetramers (Burgi et al., 2017). We could not provide evidence for a direct interaction between Antxr2 and collagen VI. If Antxr2, as proposed, is involved in the endocytosis of collagen VI it must contribute indirectly, perhaps by the activation of proteases located at the plasma membrane (Reeves et al., 2012).

In summary, we could not obtain evidence for the binding of endotrophin to cells, which would be a prerequisite for its manifold proposed roles. It remains difficult to explain how endotrophin influences lipogenesis, lipolysis, and inflammation (Oh et al., 2021), how it induces the transcription of pro-fibrotic and pro-inflammatory genes (Sun et al., 2014), or how it acts as a chemokine augmenting tumor growth (Park and Scherer, 2012). Our results, however, are consistent with the proposed role of endotrophin during the complex assembly of collagen VI (Aigner et al., 2002; Lamande et al., 2006). It is, therefore, likely that endotrophin as a biomarker senses newly expressed collagen VI, which is a hallmark of fibrotic conditions.

Limitations of the study

In this study we provide evidence supporting the concept that ANTXR1 does not function as a receptor for the cleaved-off collagen VI domain C5/endotrophin contrary to what was previously reported (Nanda et al., 2004). Therefore, the identity of the cellular receptor(s) that mediate the biological functions of C5/endotrophin remains unclear. Physiological ligands for ANTXR1 remain largely unknown and further studies will be required for their identification. In addition, we could not detect any interaction between collagen VI and ANTXR2, despite that earlier studies suggested a function as a collagen VI cell surface receptor (Burgi et al., 2017, 2020). Therefore, further investigation will be needed to clarify the role of ANTXRs in collagen VI function in health and disease.

STAR★Methods

Key resources table

Resource availability

Lead contact

Further information and requests for reagents should be directed to and will be fulfilled by the lead contact: Alvise Schiavinato (aschiav1@uni-koeln.de), Center for Biochemistry, Medical Faculty, University of Cologne, Joseph-Stelzmann-Str. 52, 50931 Cologne, Germany.

Materials availability

All materials generated in this study are available from the lead contact without restriction.

Experimental model and subject details

Cell culture

All cell lines used in this study were obtained from the American Type Culture Collection (ATCC). Cells were cultured in Dulbecco's Modified Eagle';s Medium (DMEM) or Ham's F-12 nutritional mix. MC3T3-E1 cells were cultured with Minimum Essential Medium (MEM) growth medium. All media were supplemented with 10% FBS and 1% penicillin/streptomycin. MC3T3-E1 cells lacking the ⍺2 chain of collagen VI were previously described (Schiavinato et al., 2021).

To generate stable cell lines, HEK293T and CHO-K1 cells were transfected with the FuGENE6 transfection reagent (Promega) and selected by the addition of puromycin (1 μg/mL and 10 μg/mL respectively) for full-length anthrax toxin receptors 1 and 2, or blasticidin for NG2. Collagen VI-rich conditioned medium was obtained by letting MC3T3-E1 cells grow to confluence in 100 mm culture dishes and afterwards shifting them to serum-free medium. Ascorbate (0.625 M L-ascorbate/1.125 M L-ascorbate-2-phosphate) was supplied every second day for a total of 7 days. Harvested conditioned media were briefly centrifuged at 13,000 x g to pellet dead cells and cell debris.

Method details

Plasmids and recombinant proteins

Human (Thr3101–Thr3177) and murine C5 (Thr3217-Val3284) domains were cloned as previously described (Heumuller et al., 2019) using the following primers: hC5: AAAGCTAGCGACAGAACCATTGGCTCTC(forward) and TTTGGATCCGGTTCCCATCACACTGAT (reverse); mC5: AAAGCTAGCAACAGAACCATTGTTTCT (forward) and TTTGGATCCAACTGTTAACTCAGGACTACACA (reverse). Both human and murine C5 sequences were chosen to start at the BMP1 cleavage site (Heumuller et al., 2019). Murine Antxr1-VWA (Gly34-Ile221) and Antxr2-VWA (Glu35-Ile221) were cloned using the following primers. Antxr1-VWA: AAAGCTAGCAGCTTGCTACGGAGGATTC (forward) and TTTGGATCCTTCGATGCAGGATTTCTT (reverse); Antxr2-VWA: AAAGCTAGCATCTTGCAAAAAAGCCTTC (forward) and TTTGGATCC TTCAGTACATGATTGAGC (reverse). Amplified PCR products were inserted into a modified pCEP-pu vector containing an N-terminal BM40 signal peptide as well as a C-terminal Twin-Strep-tag (Maertens et al., 2007). Full-length murine Antxr1 and 2 were cloned using the following primers. Antxr1: CAATGCTAGCGGGCCGCCGCGAGGATG (forward) and CAATGGATCCGACAGAAGGCCTTGGAGGAGG (reverse); Antxr2: CAATGCTAGCCCAGGAGCAGCCCTC (forward) and CAATCTCGAGTTGATGTGGAACCCGGGAG (reverse). Amplified PCR products were inserted into a modified sleeping beauty vector (Moya-Torres et al., 2021). A construct coding for the full-length human NG2 cloned into the pEF6/MycHis-B vector was previously described (Yuan et al., 2015). The NG2-ectodomain (Tillet et al., 1997) was subcloned into the same modified sleeping beauty vector from the full-length construct using the following primers: GATGCTAGCGGCTTCCTTCTTCGGTGAGAAC (forward) and GAATGCGGCCGCGAACATGTTGGCCTCAAGG (reverse).

Human and murine collagen VI C5 domains, anthrax toxin receptor VWA domains and the NG2 ectodomain were expressed and purified as previously described (Heumuller et al., 2019). Briefly, plasmids were introduced into HEK293T cells using the FuGENE 6 transfection reagent (Promega). Cells were selected with puromycin (1 μg/mL), and the recombinant proteins were purified directly from serum-free culture medium. After filtration and centrifugation (1 h, 10,000 × g), the supernatants were applied to a Strep-Tactin column (1.5 mL; IBA GmbH) and eluted with 2.5 mm desthiobiotin, 10 mm Tris-HCl, pH 8.0. Protein purity and concentration was estimated via SDS-polyacrylamide gel electrophoresis followed by Coomassie blue staining.

Deglycosylation experiments

Deglycosylation was performed using commercially available PNGase F (NEB #P0708), O-Glycosidase (NEB #P0733) or the Protein Deglycosylation Mix II Kit (NEB #P6044) according to manufacturer’s protocols. Briefly, 5 μg of recombinant C5 was denatured and after addition of the enzymes and buffers incubated for at least 1 h at 37°C. Samples used for deglycosylation of O-glycans were beforehand treated with 0.1 M trifluoric acid at 75°C for 30 minutes to remove sialic acid residues.

Coomassie staining

Proteins separated by SDS-polyacrylamide gel electrophoresis were stained overnight using a solution of 0.02% Coomassie brilliant blue g-250, 5% aluminium sulfate-(14–18)-hydrate, 10% ethanol and 2% orthophosphoric acid. A solution of 10% ethanol and 2% orthophosphoric acid was used for destaining.

Size exclusion chromatography

Size exclusion chromatography of recombinant C5 was performed using an ÄktaTM FPLC together with a Superose 6 Increase 10/300 GL column (Cytivia) equilibrated in TBS. Proteins were injected with a flow rate of 0.3 mL/min and measured at a wavelength of 215 nm. A gel filtration standard (Biorad #1511901) was used to calibrate the column.

Circular dichroism (CD)-spectroscopy

Recombinant C5 dissolved in sodium phosphate buffer was subjected to measurement in a JASCO-715 CD-spectrometer using a 0.1-cm-path length quartz cell. Spectra were recorded between 170 nm and 260 nm at 20 nm/min and at 20°C.

Surface plasmon resonance (SPR) interaction assays

SPR experiments were performed as described using a BIAcore 2000 system (BIAcore AB, Uppsala, Sweden) (Sengle et al., 2008). Recombinant VWA domains of Antxr1 and-2 were immobilized at 1000 reference units (RUs) to a CM5 sensor chip using the amine coupling kit following the manufacturer’s instructions (Cytiva). Interaction studies were performed by injecting 0–320 nM analyte in HBS-P buffer (0.01 M HEPES, pH 7.4, 0.15 M NaCl, 0.005% (v/v) surfactant P20) supplemented with 10 mM CaCl2 and HBS-EP (0.01 M HEPES, pH 7.4, 0.15 M NaCl, 3 mM EDTA, 0.005% (v/v) surfactant P20) (Cytiva). Kinetic constants were calculated by nonlinear fitting (1:1 interaction model with mass transfer) to the association and dissociation curves (BIAevaluation version 3.0 software). Apparent equilibrium dissociation constants (KD values) were then calculated as the ratio of the dissociation rate constant (kd) and the association rate constant (ka). Experiments in the calcium-containing buffer were performed as triplicates.

Binding assays with C5 domain, PA and collagen VI-rich conditioned medium

For immunofluorescence staining 6 × 104 cells were plated on 12 mm glass cover slips in 24-well plates; for western blot analysis 1.5 × 105 cells were plated in 12-well plates. Cells were left to grow for 24 hours before shifting them to serum-free conditions for another 24 hours. Cells were then incubated with recombinant proteins in serum-free medium (5 μg/mL for immunofluorescence or 1 μg/mL for immunoblotting) or with conditioned media for 1 h at 37°C. Afterwards, cells were washed twice with serum-free medium and fixed with 4% paraformaldehyde or lysed with RIPA buffer.

Immunofluorescence microscopy

Fixed cells were permeabilized with 0.5% NP-40 for 10 min and blocked with 1% FBS in PBS for 30 min. Cells were then incubated with primary antibodies diluted in 1% FBS in PBS for 1 hour at RT, washed 3 x with PBS for 10 minutes, and incubated with appropriate highly cross-adsorbed secondary antibodies conjugated to Alexa Fluor 488, 555 or 647 (Thermo Fisher Scientific) and DAPI (0.1 μg/mL; Sigma-Aldrich) for 1 hour at RT. Cover slips were mounted with Fluorescence Mounting Medium (DAKO) on glass slides and images taken with a Leica TCS SP5 confocal microscope. For statistical evaluation, at least 10 images containing a single cell were randomly acquired and fluorescence intensities determined with the Leica LAS AF Lite software.

Immunoblotting and pull-down assays

Cell lysates and cell culture supernatants were separated by 10% or 15% SDS-polyacrylamide gel electrophoresis under reducing conditions (non-reducing when using the antibody against C5), transferred to a nitrocellulose membrane and blocked with 5% non-fat dry milk in 0.1% Tween 20/TBS solution. The membranes were probed with primary goat antibodies against PA (1:5000, List Biological Laboratories), rabbit antibodies against C5 (1:1000, in-house generated), mouse antibodies against the His-tag (1:1000, Qiagen), rabbit antibodies against the C-terminal region of the collagen VI α1 chain (1:1000, in-house generated), guinea pig antibodies against the N-terminal region of the collagen VI α3 chain (1:500, in-house generated), or mouse antibodies against actin (1:5000, Millipore) and species corresponding horseradish peroxidase conjugated polyclonal secondary antibodies (DAKO) in blocking solution. Signals were detected by chemiluminescence (Cytivia, AmershamTM ECLTM prime western blot detection reagent). For immunoblotting of collagen VI tetramers, samples were subjected to electrophoresis on 0.5% (w/v) agarose/2.4% (w/v) polyacrylamide composite gels under non reducing conditions (Schiavinato et al., 2021). For anthrax receptor pulldown experiments, Indigo-Ni Agarose beads were used (Cube Biotech GmbH). Beads were washed and equilibrated in 0.15 M NaCl, 0.025 M Tris-HCl pH 7.5 prior to incubation with cell lysates from anthrax toxin receptor overexpressing cells (0.15 M NaCl, 0.025 M TrisHCl pH 7.5, 10% Glycerol, 1% NP-40, 1mM EDTA) for 3–6 hours. Afterwards the beads were washed three times (equilibration buffer with 0.1% NP-40 and 0.02 M imidazol) and incubated with MC3T3-E1 conditioned media or conditioned media spiked with 1 μg/mL PA overnight at 4°C. Elution of the bound material was performed after extensive washing by boiling the samples at 95°C with 2X Laemmli buffer. NG2 pulldown experiments were performed by binding recombinant NG2-ectodomain to Strep-Tactin coupled Sepharose beads (IBA GmbH) for 3–6 hours. Afterwards the beads were incubated with collagen VI rich conditioned medium overnight. After extensive washing, bead bound material was eluted by boiling the samples at 95°C with 2x Laemmli buffer.

Quantification and statistical analysis

GraphPad was used for generation of graphs and statistical analysis. For data comparison between multiple groups, one way ANOVA with Dunnett multiple testing correction was used, where a p value of ≤0.05 was considered as significant.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Rabbit polyclonal anti-Col6a3 C5	(Heumuller et al., 2019)	N/A
Rabbit polyclonal anti-Col6a1 C-term	(Maaß et al., 2016)	N/A
Rabbit polyclonal anti-Col6a3 N-term	(Maaß et al., 2016)	N/A
Guinea pig polyclonal anti-Col6a3 N-term	(Maaß et al., 2016)	N/A
Goat polyclonal anti PA	list labs	#771B
Mouse monoclonal anti-Tetra-His	Qiagen	#34650
Mouse monoclonal anti-Actin	Millipore	#1501R
Swine anti-rabbit IgG HRP-conjugated	DAKO	#P0399
Rabbit anti-mouse IgG HRP-conjugated	DAKO	#P0260
Rabbit anti-goat IgG HRP-conjugated	DAKO	#P0449
Rabbit anti-guinea pig IgG HRP-conjugated	Sigma Aldrich	#A5545
Goat anti-guinea pig IgG Alexa fluor 555	Thermo Fisher Scientific	#A21435
Goat anti-rabbit Alexa IgG Alexa fluor 488	Thermo Fisher Scientific	#A11034
Chicken anti goat IgG Alexa fluor 488	Thermo Fisher Scientific	#A21467
Donkey anti-mouse IgG Cy5	Jackson ImmunoResearch	#715-175-151
Chemicals, peptides, and recombinant proteins
FuGENE 6	Promega	#E2691
Strep-Tactin XT superflow	IBA Lifesciences	#2-5030-002
Desthiobiotin	IBA Lifesciences	#2-1000-001
PNGase F	NEB	#P0708
O-Glycosidase	NEB	#P0733
Superose 6 Increase 10/300 GL	Cytivia	#29091596
Gel filtration standard	Biorad	#1511901
PureCube 100 INDIGO Ni-Agarose	Cube Biotech	#75103
HBS-P Buffer	Cytivia	#BR100368
HBS-EP Buffer	Cytivia	#BR100188
Murine Antrx1 VWA domain	This work	N/A
Murine Antrx2 VWA domain	This work	N/A
Murine C5	This work	N/A
Human C5	This work	N/A
Experimental models: Cell lines
MC3T3-E1	ATCC	#CRL-2593
HEK293T	ATCC	#CRL-11268
CHO-K1	ATCC	CCL-61
Critical commercial assays
Protein Deglycosylation Mix II Kit	NEB	#P6044
Amine coupling kit	Cytivia	#BR100050
Amersham ECL Prime Western Blotting Detection Reagent	Cytivia	#RPN2232
Oligonucleotides
See detailed methods part
Recombinant DNA
Plasmid: murine Strep-C5/endotrophin	(Heumuller et al., 2019)	N/A
Plasmid: human Strep-C5/endotrophin	(Heumuller et al., 2019)	N/A
Plasmid: murine Strep-ANTXR1-VWA	This work	N/A
Plasmid: murine Strep-ANTXR2-VWA	This work	N/A
Plasmid: full-length murine His-ANTXR1	This work	N/A
Plasmid: full-length murine His-ANTXR2	This work	N/A
Plasmid: Strep-NG2-ectodomain	This work	N/A
Software and algorithms
GraphPad Prism 8.02	GraphPad	https://www.graphpad.com/
Leica LAS AF Lite 4.0	Leica	https://leica-las-af-lite.software.informer.com/
BIAevaluation version 3.0 software	Biacore	https://biaevaluation.software.informer.com/3.0/