Fibrosis without fibroblast TGF-β receptors?

Abstract

A report by Neelisetty et al. suggests that TGFBR2 deletion from matrix-producing interstitial cells results in decreased transforming growth factor-β (TGF-β) signaling in the cells but does not decrease renal fibrosis after injury. Considered in the context of TGF-β signaling in different cell types involved in renal fibrosis and the existence of other ligands that may produce fibrosis, these findings are provocative, but owing to technical issues of recombination efficiency in inducible models of Cre-lox gene deletion, further studies are needed.

In this issue of Kidney International, Neelisetty et al.¹ report that conditional deletion of transforming growth factor-β (TGF-β) type II receptors (TGFBR2) from matrix-producing interstitial cells (MPICs) is insufficient to reduce the severity of fibrosis in two models of kidney injury. They used tamoxifen to induce Cre recombinase by estrogen-sensitive Col1A2 or tenascin promoters with a view to delete TGFBR2 in collagen- or tenascin-producing cells, and then produced kidney injury by unilateral ureteral obstruction or the DNA-damaging toxin aristolochic acid. Successful induction of Cre in the system was indicated by expression of green fluorescent protein (GFP) in MPICs of Col1A2-Cre/ERT and Tenascin-Cre/ERT mice that had been bred with mT/mG reporter mice. Genetic ablation of TGFBR2 suppressed TGF-β signaling and collagen expression in GFP-positive MPICs sorted and cultured from kidneys, but not fibrosis in whole kidneys.¹

That TGF-β signaling is a major driver for fibrosis is accepted dogma. Because fibroblasts are endowed with the full repertoire of molecules required for signaling by TGF-β and its downstream effects to produce fibrosis, it is assumed that inhibition of TGF-β signaling in fibroblasts would prevent fibrosis. Finding that several therapeutically conceived modalities to suppress TGF-β signaling do in fact reduce fibrosis after injury has only reinforced this view. While essentially true, these findings have given rise to oversimplified views of how TGF-β works to produce fibrosis. A number of issues related to the role played by TGF-β in fibrosis can be questioned:

1. To what extent is active TGF-β produced in one tissue location available for signaling by cells at remote sites?

2. What are the ligands that induce in a paracrine or autocrine manner the fibroblast activation, proliferation, and collagen production required for fibrosis?

3. How is TGF-β production and activation by cells other than fibroblasts related to fibrosis?

An extensive body of research indicates that active TGF-β produced at one site is available to the cell producing it and immediately adjacent cells but not to cells at more remote locations even if they are nearby (Figure 1). TGF-β is secreted to some extent as a small latent complex (SLC) together with latency-associated peptide (LAP) and mostly as a large latent complex (LLC) with LAP and latent TGF-β-binding proteins (LTBPs). LTBPs are also secreted by cells as such; in molar excess, LTBPs covalently bind SLCs and themselves bind to extracellular matrix proteins.² Active TGF-β required for signaling is produced on cell surfaces by activation of SLC bound to LTBP and release of the peptide from non-covalently bound LAP. Following receptor binding and endocytosis that give rise to signaling, excess TGF-β is rapidly inactivated through binding to LAP and stored again in the extracellular matrix in the LTBP-bound form; TGF-β is also tethered directly or indirectly to other extracellular matrix proteins such as fibrillin, beta-glycan, decorin, fucoidan, and heparin.²

Figure 1. Relationships between the activation of TGF-β through paracrine, autocrine, and endocrine signaling by other ligands, autocrine TGF-β signaling, and regulation of TGF-β bioavailability in a model applicable to all of the cell types depicted in Figure 2.

Diverse stimuli trigger the release of active TGF-β from non-covalently bound latency-associated peptide (LAP) in the small latent complex (SLC) that is covalently bound to latent TGF-β-binding protein (LTBP) to form the large latent complex (LLC). The LLC is in turn bound covalently to the extracellular matrix (ECM) and to the LTBP-related protein fibrillin. Following receptor binding and endocytosis, excess TGF-β is rapidly reconverted to the latent form, as part of the ECM-bound SLC/LLC complexes. TGF-β is also directly bound by several other molecules tethered to the ECM. Thus, active TGF-β made by cells has a short half-life, becomes locally bound in the inactive form after signaling, and is not available at remote locations. TGF-β is secreted in inactive bound form as SLCs and LLCs into the extracellular space, and is not available for signaling unless it is activated. Secretion and activation are provoked through signaling by a variety of stimuli and ligands. The autocrine TGF-β signaling thus produced results in the secretion of factors that can diffuse across tissue boundaries and induce effects at remote sites. Specificity with respect to the ligands that evoke TGF-β signaling and the secretion of factors that produce effects at other locations is context dependent.

Since LLC is matrix bound, and TGF-β is stored in an inactive complex with LLC, it follows that neither active TGF-β nor latent TGF-β can diffuse to remote sites unless they are overproduced to an extent that saturates the capacity of LTBPs. Such overproduction is unexpected even in TGF-β-mediated fibrotic disease since fibrosis is a slow and indolent process, but it can certainly occur with experimental transgenic overexpression. Thus, active TGF-β produced on the surfaces of tubule epithelial cells may be available to intimately adjacent cells, but would not be able to cross tubule basement membranes, enter the interstitium, and activate interstitial fibroblasts. These inferences can be extrapolated also to inflammatory cells such as monocytes, and microvascular cells, which produce and activate TGF-β.

Similar considerations apply to endocrine TGF-β released by platelets and other sources into blood. Secreted latent TGF-β becomes covalently bound to α₂-macroglobulin and can be released only by forces and factors operating at and on cell surfaces that dissociate active TGF-β from its inactive LAP-bound complex with α₂-macro-globulin.² Virtually all cells have the capacity for autocrine TGF-β signaling by the releasing of active peptide from its latent form on their cell surfaces. On the other hand, they cannot use ‘ready-made’ active TGF-β from other sources, as tight control of TGF-β signaling in vivo requires that nascent peptide produced in excess is rapidly reconverted to the latent, bound form. It follows that for overactive TGF-β signaling to occur in any given cell type, the relevant cell must receive appropriate stimuli from its immediate environment or elsewhere to trigger the release of active TGF-β on its surface in an autocrine fashion (Figure 1).

Responses to stimuli such as mechanical forces, radiation, proteolysis, free radicals, or activating ligands—growth factors, hormones, and autacoids that bind to cellular sites—are involved in the activation of TGF-β required for signaling.^2,3 To correctly interpret the findings reported by Neelisetty et al.,¹ we must ask how it is that renal interstitial fibroblasts are becoming activated during injury states and whether TGF-β signaling in fibroblasts is an absolute requirement for fibrogenesis. This is an important question to answer in view of considerations reviewed above, which make it unlikely that fibroblasts are being activated by TGF-β derived by diffusion from adjacent injured epithelium or from surrounding inflammatory exudate. As noted above, excess active TGF-β produced anywhere becomes rapidly inactivated and matrix bound. Ligands and factors other than TGF-β that activate and induce fibroblasts to proliferate and synthesize collagens are numerous and ubiquitous, being derived from diverse endocrine and paracrine sources, as well as from autocrine production.³ Most of these factors do not have the unique attributes of TGF-β regulation—the exquisite control of biological availability through matrix binding and regulated release. Several or many of them can freely diffuse across tissue barriers in free or loosely bound form. Although well established as being fibrogenic, the signaling mechanisms that these factors employ to produce fibrosis are mostly unexplained. Mechanisms involving co-regulation with TGF-β and cooperation with autocrine TGF-β have been reported, but it is possible that TGF-β-independent mechanisms, such as mitogen-activated protein kinase- and AP-1-mediated collagen transcription, exist.⁴ Dissection of these signaling intricacies involving AP-1 is made difficult, however, by imponderables such as codependency of AP-1 and SMADs in classical TGF-β signaling.

The ubiquity and number of candidate ligands for interstitial fibroblast activation make it nearly impossible to know for sure whether TGFBR2-dependent signaling is necessary for fibrogenesis induced by any or all of these factors. Neelisetty et al. measured the responses of TGFBR2-null fibroblasts to platelet-derived growth factor-B (PDGF-B), connective tissue growth factor (CTGF), epidermal growth factor (EGF), and TGF-β.¹ Wild-type fibroblasts responded to TGF-β with increased collagen synthesis that was abolished in TGFBR2-null cells. Tested in SV40 T-antigen-transformed fibroblasts, PDGF-B, CTGF, and EGF did not increase collagen production in cells with TGFBR2, but deletion of the receptor appeared to variably decrease the expression of collagen in the presence of the peptides. Importantly, the fibroblasts had measurable basal levels of collagen synthesis without or with TGFBR2 receptors in the presence of 1% serum.¹ Since even low levels of serum may provide factors such as lysophosphatidic acid or other growth factors, further research is clearly needed to determine how TGF-β-independent basal collagen synthesis takes place in TGFBR2-null fibroblasts. Interestingly, despite TGFBR2 deletion, the fibroblasts showed morphological signs of activation by PDGF-B and CTGF.¹ This raises the possibility that fibrosis that is independent of TGF-β receptors on fibroblasts occurs as the result of basal or even sub-basal levels of collagen synthesis by greatly increased numbers of fibroblasts proliferating in response to stimulation by fibrogenic ligands.

Finally, if fibrosis can occur in the absence of TGF-β signaling in fibroblasts, how is it that TGF-β antagonism is so potently antifibrotic? A reasonable answer could be that such interventions act at critical tissue locations where inhibition of pathologically high autocrine TGF-β signaling aborts an entire cascade of downstream fibrogenic events. In such a scenario, fibrogenic peptides—PDGF-B, CTGF, or other ligands produced in response to autocrine TGF-β signaling by cells other than fibroblasts—diffuse across tissue barriers to activate fibroblasts that proliferate and give rise to fibrosis (Figure 2). Basal levels of collagen synthesis might conceivably be sufficient for this to occur provided fibroblasts increase in number. The origin of fibroblasts responsible for tubulointerstitial fibrosis has been extensively debated, with various proposed candidates including tubule cells themselves via so called epithelial–mesenchymal transition, infiltrating bone marrow-derived cells, and resident cells of the FoxD1 lineage such as pericytes and related Gli1-expressing perivascular cells.^5,6 Most recent data suggest that the latter type of cells are the major source. Several molecules such as LRP6 ligands, TWEAK, CTGF, Hedgehog ligands, PDGF-AA, PDGF-B, hypoxia-inducible factor-1α, and factors related to innate immunity are capable of evoking fibrogenic responses from fibroblasts.^6,7 Although inflammatory cells in the interstitium can also produce a variety of ligands that activate fibroblasts, they are unlikely to be the most critical cellular sites where such ligands originate, except in primary interstitial inflammatory disease. In most renal parenchymal disease, interstitial inflammation is secondary to tubule damage. Therefore injured/stressed tubule epithelium is an attractive candidate for the site that initiates early profibrotic signals. Cultured tubule cells produce PDGF-B and CTGF in response to induced autocrine TGF-β signaling, and persuasive circumstantial evidence suggests that this occurs also in vivo after injury.⁸ Furthermore, selective inhibition of PDGF-B signaling abrogated early fibrogenic events by preventing pericyte–capillary dissociation.⁵ PDGF-B, CTGF, and TGF-β plausibly can cause pericyte–capillary dissociation,⁵ but for reasons stated above, active TGF-β cannot cross the distances required for diffusion from tubules to the interstitium, but PDGF-B and CTGF can. Production of these fibrogenic ligands is not restricted to activated epithelium; they are released also by endothelial cells and monocytes at sites of injury. However, epithelial activation likely has a dominant role. As reviewed recently, many lines of evidence suggest that selective tubule injury can produce the entire downstream pathology of early tubulointerstitial fibrosis—fibroblast proliferation, capillary rarefaction, and inflammation.⁹ Thus, a tubule–pericyte–capillary endothelium axis of signaling initiated in tubules may be the common denominator for fibrosis after diverse tubule injuries, and the efficiency of TGF-β antagonists as antifibrotic agents can be explained by their actions on the most upstream signaling perturbation—in tubules. Because of the divergence and complexity of downstream events, interventions directed at downstream sites are likely to be less effective. This line of thinking is complicated by prior observations from the Gewin laboratory that TGFBR2 deletion from collecting duct epithelium produces more, not less, fibrosis. However, it is conceivable that the consequences of TGFBR2 deletion from proximal tubules will be different, and information about this is awaited with interest.

Figure 2. Relationships between TGF-β signaling and signaling evoked by ligands other than TGF-β in the diseased tubulointerstitial microenvironment that becomes fibrotic.

Each of the participating cell types—tubule epithelial (top), microvascular (bottom right), inflammatory (bottom center), and fibroblastic (bottom left)—has the molecular apparatus required for autocrine TGF-β signaling that is elicited by extraneous non-TGF-β ligands, and each of them has the ability to secrete such factors. Moreover, such non-TGF-β stimuli—in addition to TGF-β—are potentially fibrogenic by virtue of their ability to induce activation, proliferation, and ECM synthesis by fibroblasts. TGFBR2-dependent TGF-β signaling that is elicited in fibroblasts by extraneous non-TGF-β ligands participates in the fibrotic process, but is not required for fibrosis to occur if TGFBR2 is deleted from the fibroblasts. The ligands responsible, the signaling pathways that actually operate, and the hierarchical order of participation in the process by the different types of cells are likely to be strictly context dependent. However, after primary tubule damage, endogenous signaling in tubule epithelium likely has overarching importance as the initiator of a downstream cascade of fibrogenic events that result in fibrosis. Abbreviations are as in Figure 1.

The above arguments in favor of a primary role for tubules in the causation of tubulointerstitial fibrosis must be tempered by an important caveat. Ischemia, and possibly other noxious stimuli, can damage capillary endothelium in parallel to tubule damage; therefore activated endothelium could assume an important role also, in such injuries. Interestingly, Xavier et al. from the Goligorsky laboratory reported recently that partial TGFBR2 ablation in endothelial cells increased microvascular preservation, improved renal blood flow, decreased tissue hypoxia, and reduced tubulointerstitial fibrosis following folic acid nephropathy and unilateral ureteral obstruction, indicating an important role for endothelial TGF-β signaling and microvascular pathology in renal fibrosis.¹⁰ These findings implicate TGF-β signaling in injured/activated endothelial cells as a critical factor that causes microvascular rarefaction, decreased tissue perfusion, cellular hypoxia, and fibrosis. How to reconcile such critical roles for both tubules and blood vessels in TGF-β-dependent fibrosis? Is it possible that judiciously used genetic interventions will yet reveal such a critical role for inflammatory-cell TGF-β signaling as well? Such perplexing apparently conflicting considerations can only be reconciled by accepting that the evolution of tubulointerstitial fibrosis involves a vicious cycle of interactions between tubules, microvascular cells (endothelial cells and pericytes), inflammatory cells, and fibroblasts; these interactions feed on each other to amplify and boost the fibrogenic signaling milieu. We have elaborated how such a vicious cycle might operate in a recent review.⁹

The findings of Neelisetty et al.¹ are provocative and need serious consideration. However, they also need to be tempered by technical considerations relating to recombination efficiency required for TGFBR2 deletion in fibroblasts and precursors. All of the evidence for high recombination shown in the paper was derived from interstitial cells sorted for GFP expressed by the recombined mT/mG construct. Thus, the cells that were examined were preselected for the Cre-expressing population in which TGFBR2 deletion is expected. What if there were issues related to failed recombination—for whatever reasons, including non-expression of required transcription factors, silencing of the transgene, mosaicism, and so on—that resulted in the deletion of TGFBR2 from only a sub-population of fibroblasts? In other words, fibroblasts that did not express GFP as a marker for being sorted would be missed. The in vivo fluorescence images as well as counts for α-smooth muscle actin-positive and GFP-positive cells suggest the existence of a significant number of α-smooth muscle actin-positive cells that did not express GFP. The data are less clear on whether there was significant disparity in counts between PDGFRβ⁺ cells and GFP-expressing cells in the interstitium. These considerations suggest that further work is needed to make the point with certainty. This is an important point to make, a needed step in the winding road toward understanding the vagaries of TGF-β signaling as it relates to renal fibrosis.