Complement C5 Inhibition Reduces T Cell-Mediated Allograft Vasculopathy Caused by Both Alloantibody and Ischemia Reperfusion Injury in Humanized Mice

Abstract

Allograft vasculopathy (AV) is characterized by diffuse stenoses in the vasculature of solid organ transplants. Previously, we developed two humanized models showing that alloantibody and ischemia reperfusion injury (IRI) exacerbated T cell-mediated AV in human arterial xenografts in vivo. Here we examined a causal role for terminal complement activation in both settings. IRI, in contrast to alloantibody, elicited widespread membrane attack complex (MAC) assembly throughout the vessel wall. Both alloantibody and IRI caused early (24 h) and robust endothelial cell (EC) activation localized to regions of intimal MAC deposition, indicated by increases in NF-κB inducing kinase, a MAC-dependent activator of non-canonical NF-kB, VCAM-1 expression, and Gr-1+ neutrophil infiltration. Endothelial cell activation by alloantibody was inhibited by anti- mouse C5 mAb, but not by anti-C5a mAb or by control mAb, implicating MAC as the primary target of anti-C5 mAb. Anti-mouse C5 mAb significantly reduced alloantibody- and IRI-enhanced T cell infiltration and AV-like changes including neointimal hyperplasia as well as intraluminal thrombosis in a subset of IRI-treated arterial grafts. These results indicate that increased AV lesion formation in response to either alloantibody or IRI are dependent on complement C5 activation and accordingly inhibition of this pathway may attenuate AV.

Introduction

Longevity of cardiac,^1-4 renal,^5,6 and composite allografts^7-9 is frequently limited by allograft vasculopathy (AV), a condition characterized by irreversible stenoses and thromboses that develop throughout the graft vasculature. In the most well-appreciated form of AV, affected vessels contain a diffusely expanded neointima made up of smooth muscle-like cells along with sub-endothelial infiltrates of T cells and macrophages.¹⁰ Similar appearing neointimal lesions can be induced in human artery segments in response to human IFN-γ¹¹ or IFN-γ-producing human T cells.¹² A second, more recently recognized form of AV involves the formation of thromboses, especially in the allograft microvasculature.

Donor specific antibody (DSA) as well as delayed graft function [a manifestation of ischemia reperfusion injury (IRI)] are risk factors for AV. We propose that these factors mediate AV by increasing the capacity of EC to stimulate IFN-γ production by alloreactive T cells. Using high titer panel reactive antibody (PRA) sera pooled from allosensitized transplant candidates we developed a protocol to bind alloantibody and activate complement, resulting in increased EC immunogenicity mediated in response to deposition of membrane attack complex (MAC) and non-canonical NF-κB signaling. In vivo, treatment of human artery segments with PRA led to human IgG binding to, murine MAC deposition on, and activation of noncanonical NF-κB signaling in intimal EC.¹³ These effects resulted in exacerbated formation of T cell-mediated AV-like lesions. We have also shown that IRI-like lesions also exacerbate T cell-mediated AV in the same humanized mouse model.¹⁴ These results suggest that targeting terminal complement could be a target of therapy to reduce AV. Our more recent studies have shown that inhibition of endocytosis blocked the effect of PRA treatment,¹⁵ but this is unlikely to be tolerated clinically; prevention of complement activation, an inducible process linked to inflammation appeared a more attractive alternative. Although complement activation has been linked to IRI,^16-18 we had not determined if terminal complement activation occurs in our humanized model of IRI. Eculizumab is a monoclonal anti-human C5 antibody that blocks the generation of downstream inflammatory mediators including C5a, a fluid-phase anaphylatoxin, and C5b, a terminal complement component which, along with C6, C7, C8, and polymers of C9, assemble to form solid-phase MAC. In this report, we investigated the effects of an anti-mouse C5 mAb, whose actions may be comparable to that of eculizumab, on non-canonical NF-κB signal activation in EC and the development of AV lesions using our humanized mouse models of alloantibody- and IRI-mediated AV.

Materials and Methods

Examination of Effects of PRA on Human Vessel Grafts

All human materials were obtained under protocols approved by the Yale Human Investigations Committee or the IRB of the New England Organ Bank. All animal experiments were conducted under protocols approved by the Yale institutional Animal Care and Use Committee.

Discarded high-titer panel reactive antibody (PRA) sera were obtained from cardiac and renal transplant candidates as de-identified samples from Yale-New Haven Hospital's tissue typing laboratory. Sera from patients that had undergone panel reactive antibody (PRA) blood testing and found to have >80% reactivity to either HLA class I and/or class II antigens were pooled, tested and found to be negative for endotoxin activity (Sigma) prior to use. Human peripheral blood mononuclear cells (PBMC) were collected from healthy adult volunteer donors.

For PRA-mediated AV, adjacent 3-5 mm lengths of third or fourth order human coronary artery segments, approximating the caliber of murine aortae, were surgically implanted as end-to-end interposition grafts in the infrarenal position of descending aortae of paired SCID/bg immunodeficient mice (Taconic). The transplanted vessels were quiesced for ∼30 days prior to i.v. tail injection of 200μL neat PRA sera (in one mouse) or PRA sera depleted of IgG using a mAb Trap serum fractionation kit (GE HealthCare) into the paired mouse. Eighteen hours later, grafts were harvested and immunostained. In other experiments, each of the paired human arterial xenografts was explanted with cuffs of mouse aorta on both ends and then interpositioned into the infrarenal aortae one member of a second pair of naive SCID/bg hosts that had been inoculated with human peripheral blood mononuclear cells.100-200×10⁶ cells allogeneic to the artery donor. In these mice, the efficiency of T cell engraftment was assessed by flow cytometry of peripheral blood sampled at weekly intervals, and the percentage of CD3+ engraftment relative to total murine CD45+ cells ranged between 5-15% prior to arterial xenograft implantation. Re-implanted grafts were harvested 14 days after implantation. Harvested tissues were frozen in OCT media blocks, sectioned at 5μm thickness, and subjected to morphologic, immunohistochemical, and immunofluorescent analyses as previously described.¹³

Induction of Ischemia-Reperfusion Injury (IRI)

As above, adjacent 3-5mm lengths of third or fourth order human coronary artery segments approximating the caliber of murine aortae were surgically implanted as end-to-end interposition grafts in the infrarenal position of descending aortae of SCID/bg immunodeficient mice (Taconic). Following a ∼1 month period of quiescence, arterial xenografts along with cuffs of mouse aorta on both ends were explanted and incubated ex vivo under conditions of anoxia for 12 h prior to surgical reimplantation into a second pair of SCID/bg recipients and analyze 18 h later. Alternatively, where indicated, arterial xenografts subjected to anoxia were reimplanted into SCID/bg hosts that had been inoculated with 100-200×10⁶ human peripheral blood mononuclear cells and grafts were harvested 21 days after implantation or sooner if evidence was seen of distress or hindlimb paralysis indicative of thrombosis. Additional Supporting Information may be found in the online version of this article.

Anti-C5 and Anti-C5a Blocking Antibodies

Anti-mouse C5 blocking antibody (BB5.1), control isotype antibody (12B4), anti-mouse C5a blocking antibody (CLS026) and control murine isotype antibody (MOPC1) were provided by Alexion Pharmaceuticals. Mice were injected subcutaneously with 0.8 mg of each antibody or 0.8mg of each antibody were added to hypoxic media prior to surgical implantation as specified in the text and figure legends. Assays for neutralization of plasma C5a by CLS026 are described in the online Supporting information

Statistical Analyses

All experiments involved comparisons between pairs of animals receiving human artery segments from adjacent portions of the same donor vessel subjected to distinct manipulations. Statistical analyses were performed using computer software (Origin), and the data were analyzed by two-tailed Student t tests where P<0.05 was considered significant.

Results

Anti-C5 mAb Does Not Exert Immunomodulatory Effects on T Cell-Mediated AV in the Absence of Terminal Complement Activation

Prior to use in vivo, the efficacy of BB5.1 mAb for inhibiting terminal complement activation and its purported specificity for selectively inhibiting murine complement was confirmed in vitro. To do this we exploited the fact that the IgG+ fraction of panel reactive antibody (PRA) sera from allo-sensitized transplant candidates caused IgG binding to EC, a process that could elicit terminal activation of murine or human complement.¹³ We thus treated human umbilical vein EC (HUVEC) with the IgG+ fraction of PRA sera in the presence of exogenous human or murine complement and assessed for surface deposition of C5b-9 by flow cytometry. We found that, compared to 12B4 mAb-treated EC, BB5.1 mAb blocked the ability of the IgG+ fraction of PRA sera to elicit C5b-9 deposition in the presence of murine complement (Fig 1a, left) but not human complement (a, right). In contrast, C4d staining was unchanged by 12B4 or BB5.1 mAb (data not shown). These data show that BB5.1 mAb could effectively and selectively inhibit terminal activation of murine complement.

Fig 1. Anti-C5 mAb Does Not Exert Immunomodulatory Effects on T Cell-Mediated AV in the Absence of Terminal Complement Activation.

HUVEC were pretreated with 12B4 control mAb or BB5.1 anti-C5 mAb (25μg) for 30 min prior to the addition of the IgG+ fraction of PRA and mouse complement (C′) or human complement for 4h. Cells were then analyzed by FACS for C5b-9 (a). Quiesced human coronary artery grafts in immunodeficient mice were pre-exposed to 0.8mg 12B4 or BB5.1 antibody, and at the time of graft harvest 12B4 or BB5.1 mAbs were added into a PBS media bath and immediately retransplanted into a second host which had been pre-treated with 12B4 or BB5.1 mAbs and contained circulating human T cells from a prior adoptive transfer of PBMC allogeneic to the artery donor. The second hosts received ongoing 12B4 or BB5.1 Ab treatments at 0.8mg/injection/mouse for 14 days as shown (b). Arterial grafts were analyzed for C5b-9 (c, top row) staining and CD45RO+ T cell infiltration (c, bottom row). Neointimal lesion formation was assessed between treatment groups following EVG and Movat staining (d). n=5 treatment pairs for all experiments. No significant differences between control and specific anti-C5 antibody were seen in the absence of terminal complement activating stimuli (N.S.). Scale bars indicate 63μm.

We next assessed the effects of anti-mouse C5 mAb BB5.1 on T cell-mediated formation of AV lesions in our humanized mouse model in the absence of exacerbating factors, i.e., DSA or IRI, causing complement activation. Murine hosts engrafted with human T cells received control 12B4 or anti-C5 BB5.1 mAbs as diagrammed in Fig 1b. We did not observe complement activation in this protocol in either treatment group as indicated by a lack of staining of C5b-9, the major component of MAC (Fig 1c, top row). Additionally, there were no significant differences in CD45RO+ T cell infiltration (Fig 1c, bottom row) or neointima lesion area (Fig 1d) between hosts treated with anti-C5 mAb or isotype control mAb. Together these data indicate that anti-C5 BB5.1 mAb does not non-specifically attenuate graft inflammation or neointimal lesion formation in the absence of complement activation.

Blockade of MAC Assembly and Not C5a Generation by Anti-C5 mAb Attenuates EC Activation in Response to Alloantibody

Exposure of human arterial grafts to PRA in vivo leads to human IgG binding to EC, thus modeling the effect of DSA. IgG binding to EC is followed by murine complement activation leading to MAC assembly on the EC¹³ in SCID/bg hosts; this strain has complement activity comparable to immunocompetent mouse strains.²⁵ PRA treatment induced non-canonical NF-κB signaling as detected by NIK stabilization, a MAC-dependent effector pathway causing EC activation.¹³ We thus hypothesized that inhibition of terminal complement activation by anti-C5 mAb, sparing early complement activation, would subvert the pro-inflammatory changes in EC mediated through non-canonical NF-κB signaling. To test this, hosts bearing arterial xenografts were pre-treated with 12B4 or BB5.1 Ab prior to i.v. PRA injection. Following injection grafts were harvested 18 hours later (Fig 2a). Animals treated with BB5.1 mAb showed early (C4d) but not terminal complement activation, i.e., MAC assembly, as indicated by C9 staining which was confined to the intimal lining (Fig 2b). To confirm MAC assembly, we co-stained PRA-treated grafts with C6, which, expectedly showed a high degree of neointimal co-localization with C9 (Fig S1a). Hereafter, we will refer to the detected complexes as C5b-9. In contrast to BB5.1-treated animals, both C4d and C5b-9 deposition were detected in 12B4 control mAb-treated animals. Decreased C5b-9 staining in BB5.1-mAb-treated hosts correlated with significantly attenuated EC expression of NIK (Fig 2c,e) and VCAM-1 (Fig 2d,e), markers of non-canonical NF-κB activation and inflammatory gene expression, respectively. Real-time PCR of graft lysates showed a significant reduction in NIK-dependent inflammatory genes, CCL5 and VCAM-1 (Fig 2f). These data demonstrate the efficacy of anti-C5 mAb in blocking terminal complement activation by PRA and its attendant inflammatory effects including non-canonical NF-κB signaling and EC activation.

Fig 2. Blockade of MAC Assembly and Not C5a Generation by Anti-C5 mAb Attenuates EC Activation in Response to DSA.

To assess the effect of anti-C5 treatment on alloantibody-induced complement activation, SCID/bg immunodeficient hosts were treated with isotype control 12B4 mAb or anti-C5 BB5.1 mAb prior to i.v. tail vein injection of PRA, and grafts were harvested 18 hours later (a). Compared to 12B4 Ab-treated hosts, animals treated with BB5.1 Ab showed equivalent C4d staining but decreased C5b-9 staining (b). BB5.1 mAb treatment of coronary xenografts showed decreased intimal NIK staining (c) and VCAM-1 expression (d) compared to 12B4 Ab-treated controls. C4d, C5b-9, NIK, and VCAM-1 staining results were quantified (e). Intra-graft transcripts of CCL5 and VCAM-1 were quantified and normalized to CD31 transcript levels (f). n=5 treatment pairs for the above experiments. SCID/bg hosts were treated as shown with isotype control MOPC1 or anti-C5a CLS026 mAb prior to i.v. tail vein injection of PRA and graft harvest 18 hours later (g). Compared to hosts treated with control MOPC1 mAb, CLS026 mAb-treated hosts developed equivalent levels of C5b-9 deposition (h, top row), NIK upregulation (h, second row), VCAM-1 expression (h, third row), and neointimal recruitment of Gr-1+ cells (h, arrowheads, bottom row). n=3 treatment pairs for the above experiments. N.S. indicates no statistical difference between groups. Asterisks indicate p<0.05. Scale bars indicate 135μm.

BB5.1 prevents generation of both C5a and C5b. While our prior in vitro studies demonstrated that MAC (C5b-9), and not C5a, led to EC activation and increased immunogenicity, it is possible that C5a played some role in vivo. To determine which of these factors was the functionally relevant target of anti-C5 mAb, we treated hosts bearing human artery grafts with a mAb specifically blocking C5a (CLS026) or an isotype control mAb (MOPC1), as shown in Fig 2g, prior to injection with PRA. C5a activity was assessed in murine sera collected at the time of graft harvest. PRA treatment led to circulating mouse C5a and this was significantly inhibited in CLS026-treated hosts (Fig S1b) to levels observed in historical control mouse strains (data not shown). Importantly, inhibition of anti-C5a did not reduce the extent of intramural MAC assembly compared to controls (Fig 2h, top row) and resulted in unchanged degrees of NIK upregulation, EC activation, and EC-mediated recruitment of Gr-1+ neutrophils in hosts treated with anti-C5a mAb vs control mAb (Fig 2h). As C5a blockade did not affect parameters of EC activation, we conclude from these data that MAC is primarily responsible for PRA-induced EC activation and, accordingly, the ability of anti-C5 mAb to attenuate EC activation is due to the ability of this mAb to block the assembly of MAC rather than the generation of C5a.

Anti-C5 Antibody Reduces Neointimal Lesions Following DSA-Induced Complement Activation

Having established the complement-blocking efficacy of anti-C5 mAb in PRA-treated animals, we then assessed the effect of anti-C5 mAb on alloantibody-mediated increases in neointimal AV lesions. Grafts pre-treated with either 12B4 or BB5.1 were exposed to PRA sera and then reimplanted into a second set of immunodeficient hosts that had been previously engrafted with human T cells and pre-treated with 12B4 or BB5.1 Ab. These second set of hosts then received ongoing antibody treatments for 14 days prior to graft harvesting as depicted in Fig 3a. Compared to control hosts receiving 12B4 mAb, BB5.1 mAb-treated hosts showed significantly decreased intimal NIK staining (Fig 3b, top row) and decreased infiltration of CD45RO+ alloimmune T cells (Fig 3b, bottom row). These changes were associated with significantly decreased neointimal areas and increased luminal areas in BB5.1 mAb-treated hosts compared to 12B4 mAb-treated controls (Fig 3c). No hosts in either the BB5.1- or 12B4-treated groups developed thrombotic AV lesions. These data showed that, in a complement-dependent model of alloantibody-induced AV, anti-C5 blocking antibody inhibited terminal complement activation, EC activation, and non-canonical NF-κB and that these changes correlated with significantly reduced neointimal AV lesion formation.

Fig 3. Anti-C5 Antibody Reduces Neointimal Lesions Following Alloantibody-Induced Complement Activation.

12B4 isotype control or BB5.1 anti-C5 Ab were tested for their effects on alloantibody-induced complement activation and neointimal formation. Arterial grafts were pre-treated with 12B4 or BB5.1 Ab prior to i.v. PRA injection and reimplantation into a second immunodeficient hosts engrafted with human T cells and similarly pre-treated with 12B4 or BB5.1 Ab. These hosts additionally received ongoing antibody treatments for 14 days prior to graft harvesting (a). Compared to 12B4 Ab-treated controls, hosts treated with BB5.1 mAb showed attenuated intimal NIK staining (b, top row) and intimal infiltration of CD45RO+ alloimmune T cells (b, bottom row). Neointimal and luminal areas were quantified in 12B4- and BB5.1 mAb-treated hosts (c). n=6 treatment pairs for all experiments. IEL indicates internal elastic lamina. Asterisks indicate p<0.05. Scale bars indicate 63μm.

IRI Activates Complement and Induces MAC Formation in the Artery Wall without Overt Target Cell Loss

Activation of terminal complement has been shown in other models to contribute to IRI.^16-18 We used an anti-C5 mAb therapy to test the contribution of terminal complement activation in our model of IRI¹⁴ where a role for terminal complement and non-canonical NF-κB had not been previously established. Before embarking on the current experiments, we re-evaluated the optimal time of hypoxia using human artery segments and found that 12 h led to more pronounced staining with hypoxyprobe (Fig S2) without evidence of cell necrosis or aneurysm development upon re-transplantation. All subsequent experiments used 12 h of ex vivo hypoxia to trigger IRI. Using this adapted protocol as shown in Fig 4a, we examined whether early (C4d) or terminal (C5b-9) complement activation occurred in arterial xenografts following 12 hours of ex vivo hypoxia relative to control grafts which were immediately reimplanted into second recipient hosts. In contrast to our model of PRA-induced coronary AV where complement staining was confined exclusively to EC (Fig 2b), coronary artery xenografts exposed to hypoxia showed significantly increased C4d and C5b-9 deposition in both the intimal and medial regions of the vessel wall (Fig 4b), consistent with prior observations that IRI affected mural smooth muscle cells as well as luminal EC.¹⁴ We then assessed whether IRI had caused non-specific complement activation on murine EC outside of human arterial xenografts by assessing C5b-9 staining in serial sections contiguous to the distal surgical suture line. In IRI-treated grafts we found that C5b-9 staining, while abundant on human intima and media proximal to the suture line, was completely absent in murine EC distal to the suture line, indicating that complement activation specifically occurred on human IRI-treated tissue (Fig S3). We furthermore did not observe loss of EC (CD31) or SMC (smooth muscle α-actin) in regions of MAC assembly (Fig 4c), consistent with non-lethal injury. We also assessed downstream effects of terminal complement activation. We observed significantly increased NIK protein expression throughout the vessel wall, co-localizing with MAC and no longer restricted to EC, (Fig 4d, top row). NIK expression correlated with significantly increased VCAM-1 (Fig 4d, second row) and recruitment of mouse Gr1+ neutrophils (Fig 4d, bottom row). These data suggest that IRI activated non-canonical NF-κB signaling in both EC and SMC, a process correlating with EC activation and acute mural inflammation. Our prior study showed that in recipient animals lacking human T cells, this episode of perioperative acute inflammation resolves without long term sequelae.¹⁴

Fig 4. IRI Activates Complement and Induces MAC Formation in the Artery Wall Without Overt Target Cell Loss.

For IRI-induced complement activation, coronary artery xenografts parked in an immunodeficient host were explanted and either immediately surgically reimplanted into a second immunodeficient host [(-) hypoxia] or placed in organ culture under hypoxic conditions for 12 hours [(+) hypoxia] prior to reimplantation (a). Explanted coronary arteries subjected to 12 hours of hypoxia ex vivo were examined by I.F. for CD31 and either C4d (b, top row) or C5b-9 (b, bottom row) and staining results were quantified as indicated. CD31 and SMA staining were quantified between control grafts not subjected to hypoxia or grafts subjected to 12 hours of hypoxia (c). NIK (d, top row), VCAM-1 (d, middle row), and Gr-1+ (d, bottom row) staining were performed in coronary grafts and results were quantified as indicated. n=6 treatment pairs for all experiments. N.S. indicates no statistical difference between groups. Asterisks indicate p<0.05. Double asterisks indicate p<0.001. Scale bars indicate 63μm.

Terminal Complement Inhibition Blocks IRI-Induced MAC Formation, Non-Canonical NF-κB Signaling, and EC Activation

We investigated the effects of BB5.1 mAb in our humanized model of IRI-exacerbated AV. We first examined whether BB5.1 could block terminal complement activation in this model as shown in Fig 5a. Compared to hosts treated with 12B4 control mAb, hosts treated with BB5.1 mAb showed early (C4d, Fig 5b, top row) but not terminal (C5b-9, Fig 5b, bottom row) complement activation following induction of IRI. Blockade of terminal complement was associated with significantly decreased NIK (Fig 5c, top row), VCAM-1 expression (Fig 5c, middle row), and decreased infiltration of mouse Gr-1+ neutrophils (Fig 5c, bottom row) 18 hours after exposure to IRI.

Fig 5. Terminal Complement Inhibition Blocks IRI-Induced MAC Formation, Non-Canonical NF-κB, and EC Activation.

Coronary artery segments were exposed to 12B4 or BB5.1 mAb and subjected to ex vivo hypoxia as shown prior to graft harvesting 18 hours later (a). Explanted coronary artery xenografts were analyzed for early and terminal complement activation, with C4d (b, top row) and C5b-9 (b, bottom row), respectively. Grafts were additionally analyzed for EC activation with NIK (c, top row), VCAM-1 (c, middle row), and Gr-1 (c, bottom row). n=4 treatment pairs for all experiments. N.S. indicates no statistical difference between groups. Asterisks indicate p<0.05. Double asterisks indicate p<0.001. Scale bars indicate 63μm.

Anti-C5 mAb Attenuates T cell-mediated Neointimal and Thrombotic AV Lesions Following IRI

In a final set of experiments, we assessed the effect of anti-C5 antibody or control antibody on the development of AV lesions in arteries subjected to IRI and then retransplanted into hosts that had been previously given human PBMC allogeneic to the artery donor (Fig 6a). Antibodies were introduced in the first host just prior to graft harvest, and treatments with 12B4 or BB5.1 mAb continued for an additional two weeks in the second host prior to graft harvesting and analysis. By immunofluorescence microscopy, coronary grafts treated with BB5.1 anti-C5 mAb showed significant reduction in MAC (C5b-9) formation (Fig 6b, top row), NIK expression (Fig 6b, second row), and an ∼50% decrease in mural infiltrates of alloimmune CD45RO+ T cells (Fig 6b, third row). Interestingly, in hosts treated with either 12B4 or BB5.1 mAb, we observed significantly greater numbers of CD4+ T cells in the intima (Fig 6b, bottom row, left graph) and significantly greater numbers of CD8+ T cells in the media (right graph). While significantly reducing overall numbers of both CD4+ and CD8+ T cells compared to hosts treated with 12B4 mAb (asterisks and double asterisks), BB5.1 mAb did not significantly alter the ratio of CD4+ T cells vs CD8+ T cells in the intima (p=0.76) or media (p=0.21), indicating that anti-C5 mAb reduced infiltration by CD4+ and CD8+ T cells to an equal extent. The changes above were associated with significantly decreased neointimal area without significant change in luminal area in BB5.1 mAb-treated hosts compared to controls, suggesting early outward remodeling of the allograft vessel during AV lesion formation (Fig 6c). These data show that terminal complement blockade with anti-C5 antibody attenuated the formation of neointimal AV lesions following IRI.

Fig 6. Anti-C5 Antibody Attenuates Neointimal and Thrombotic AV Lesions Following IRI.

Anti-C5 therapy was tested for its effects on IRI-induced complement activation and development of AV lesions. Human arterial xenografts were pretreated with either 12B4 or BB5.1 Ab prior to graft explanation, exposure to hypoxia for 12 hours, and reimplantation into a second immunodeficient host that had been pre-treated with 12B4 or BB5.1 mAb and engrafted with human T cells. These hosts received mAb treatment as shown (a). Effects of anti-C5 Ab were assessed in coronary artery xenografts exposed to IR injury. Compared to 12B4 Ab-treated controls, hosts given BB5.1 Ab showed significantly decreased intimal and medial staining of C5b-9 (b, top row), NIK staining (b, second row), and a significant reduction in the number of CD45RO+ infiltrating T cells (b, third row). Intimal- and medial-infiltrating CD4+ and CD8+ T cells were quantified following 12B4 or BB5.1 mAb treatment (b, bottom row). Neointimal thickness, luminal area, and medial thickness were quantified in hosts receiving 12B4 or BB5.1 treatments (c). Intraluminal thrombosis was visualized in 3 of 8 hosts treated with 12B4 Ab (d, top row) and 0 of 8 hosts treated with BB5.1 Ab (d, bottom row). n=8 treatment pairs for all experiments. N.S. indicates no statistical difference between groups. Asterisks indicate p<0.05. Double asterisks indicate p<0.001. Scale bars indicate 63μm.

Three out of eight hosts receiving IRI-treated grafts and treated with 12B4 mAb became moribund and developed bilateral hindlimb ischemia during the experimental protocol and were thus prematurely sacrificed at days 4, 7, and 10 post-transplant. Necropsy of harvested grafts revealed the presence of flow-limiting thromboses in all three hosts (Fig 6d, top row) which were confirmed with fibrinogen staining (Fig 6d, top row). Thrombosis was not due to a secondary effect of 12B4 mAb treatment as hosts implanted with grafts not exposed to hypoxia and treated with 12B4 mAb did not develop similar lesions (Fig 1d). Remarkably, all BB5.1-treated hosts were spared from the development of thrombotic lesions during the treatment period (0/8, Fig 6d, bottom row). We conclude that anti-C5 antibody prevents both complement-mediated neointimal and thrombotic AV lesions induced by IRI.

Discussion

In this study we demonstrate a positive effect of anti-C5 therapy on experimental AV lesions which develop following exposure to alloantibody or IRI. Repeated administration of anti-C5 blocked terminal but not early complement activation in human coronary artery tissues, resulting in attenuated activation of non-canonical NF-κB signaling and decreased formation of both stenotic and thrombotic AV lesions in vivo. Importantly, our study demonstrates a causal connection between terminal complement activation with non-canonical NF-κB activation and IRI. Hosts treated with anti-C5 mAb showed attenuated NIK expression and AV lesions compared to controls (Fig 6). In comparison with DSA, IRI induced MAC assembly throughout the vessel wall in a broader area of distribution including endothelial cells (EC) in the intima and smooth muscle cells (SMC) in the media. Whether MAC and/or non-canonical NF-κB elicits differential functional effects in these cell types is unknown and is especially relevant considering mechanisms of immunoprivilege in the media relative to the intima¹⁹ and also considering the observed differential spatial localizations of infiltrating CD4+ and CD8+ T cells following IRI (Fig 6c).

In addition to the neointimal expansion, a more recently appreciated form of AV is characterized by activation of the coagulation cascade¹⁸ resulting in occlusive thromboses.^20,21 Thrombotic AV lesions affecting microvessels^21,22 and large-caliber vessels²³ are associated with worsened long-term graft outcomes.²⁴ Unexpectedly, we observed that several of our vessels subjected to IRI and then transplanted into animals with circulating human T cells developed thrombosis (Fig 6d). Remarkably, anti-C5 mAb also prevented thrombotic complications as a result of transplant-associated IRI. It is unclear why thrombosis developed in some, but not all grafts exposed to IRI and was not seen in any grafts treated with PRA. A possible explanation is that the more extensive injury produced by IRI, compared to PRA, extending into the vessel media, results in greater production of activators of coagulation and/or loss of greater inhibitors of coagulation. The identity of the molecules responsible is the subject of ongoing investigation. Although the total number of samples exhibiting this change (3/8) is relatively small, these data raise the possibility that complement-inhibiting therapies may have a role in preventing thrombotic sequelae in conditions associated with complement activation such as acute myocardial infarction (MI), severe sepsis, and disseminated intravascular coagulation accompanying trauma.

A strength of this study is that we used human arteries, alloantibodies and PBMCs to model CAV, a disorder which is poorly modeled in rodents¹⁰. However, the complement components and the coagulation factors that are activated are of mouse origin. Because some regulators of these processes are species-restricted, caution should be exercised in extrapolation of our findings to the clinical setting. Furthermore, our humanized mouse model does not fully recapitulate the human immune system in that the circulating human leukocyte populations following adoptive transfer of PBMC lack dendritic cells and monocytes which have been shown to synthesize complement components to amplify local terminal complement activation. This population also lacks NK cells, which like monocytes may potentially contribute to CAV via Fc receptor binding to alloantibody. On the other hand, the absence of these complicating factors allowed direct interrogation of the effect of MAC-mediated activation of graft EC on T cell recruitment to and activation by both IRI- and alloantibody-treated human xenografts. We also note that we have not formally established that C5a does not contribute to CAV, but only that it is not required for alloantibody and complement-mediated EC activation via non-canonical NF-κB signaling. Nevertheless, our key conclusion that MAC plays a critical role in exacerbation of CAV is buttressed by our prior studies showing inhibition of EC activation by blocking endocytosis of MAC.¹⁵

Supplementary Material

Supp Fig S1-S3

Figure S1: Naïve SCID/bg mice containing coronary artery interposition grafts implanted in the descending aorta received i.v. injection of PRA sera prior to graft harvest 18 hours later and analysis by I.F. (n=3 treatment pairs). C5a activity levels were assessed in murine plasma at the time of graft harvest as depicted in Fig 2g following treatment with blocking anti-C5a mAb CLS026 or an isotype control mAb MOPC1 (n=3 treatment pairs). PRA, panel-reactive antibody.

Figure S2: Coronary artery grafts were explanted, placed in organ culture under anoxic conditions for the times indicated and stained for hypoxyprobe (n=5 mice for each time point). Scale bars indicate 63μm.

Figure S3: Coronary artery grafts were explanted, placed in organ culture under anoxic conditions for 12 hours and then reimplanted into a second set of naïve SCID/bg hosts. Grafts were harvested 18 hours after re-implantation and serial sections were taken for I.F. analysis adjacent to the distal suture line, indicating the margin between human arterial graft and murine descending aorta tissue.

Abbreviations

AV

allograft vasculopathy

DSA

donor specific antibody

EC

endothelial cell

IRI

ischemia reperfusion injury

MAC

membrane attack complex

NIK

NF-κb inducing kinase

PRA

panel-reactive antibody