To isolate recombinant antibodies with specificity for human arthritic synovium and to develop targeting reagents with joint-specific delivery capacity for therapeutic and/or diagnostic applications.
To isolate recombinant antibodies with specificity for human arthritic synovium and to develop targeting reagents with joint-specific delivery capacity for therapeutic and/or diagnostic applications.
In vivo single-chain Fv (scFv) antibody phage display screening using a human synovial xenograft model was used to isolate antibodies specific to the microvasculature of human arthritic synovium. Single-chain Fv antibody tissue-specific reactivity was assessed by immunostaining of synovial tissues from normal controls and from patients with rheumatoid arthritis and osteoarthritis, normal human tissue arrays, and tissues from other patients with inflammatory diseases displaying neovasculogenesis. In vivo scFv antibody tissue-specific targeting capacity was examined in the human synovial xenograft model using both 125I-labeled and biotinylated antibody.
We isolated a novel recombinant human antibody, scFv A7, with specificity for the microvasculature of human arthritic synovium. We showed that in vivo, this antibody could efficiently target human synovial microvasculature in SCID mice transplanted with human arthritic synovial xenografts. Our results demonstrated that scFv A7 antibody had no reactivity with the microvasculature or with other cellular components found in a comprehensive range of normal human tissues including normal human synovium. Further, we showed that the reactivity of the scFv A7 antibody was not a common feature of neovasculogenesis associated with chronic inflammatory conditions.
Here we report for the first time the identification of an scFv antibody, A7, that specifically recognizes an epitope expressed in the microvasculature of human arthritic synovium and that has the potential to be developed as a joint-specific pharmaceutical.
Rheumatoid arthritis (RA) is a chronic inflammatory disease that principally affects synovial joints, causing disability with significant associated morbidity and mortality (1, 2). In RA, the synovium becomes hyperplastic and locally invasive at the interface between the cartilage and bone, resulting in the destruction of articular cartilage and subchondral bone, leading to joint damage and disability. Notably, in RA, the synovium becomes heavily infiltrated by T and B cells, plasma cells, and monocytes through the development of new blood vessels (angiogenesis) (3, 4). It is now clear that synovial angiogenesis contributes significantly to disease pathogenesis and progression (5, 6) and may precede other pathologic features of RA, since synovial hypercellularity is sustained by an increase in the number and density of synovial blood vessels (6–8). Further, not only angiogenesis but also vasculogenesis may contribute to the increased vascularity observed in the RA synovium (9), making the newly formed blood vessels a particularly attractive therapeutic target for the management of inflammatory arthritis (10). In support of this, blockade of inflammatory neovascularization has been shown to lead to the suppression of synovial inflammation and proliferation and to an attenuation of synovitis in RA (11).
The treatment of this condition has been transformed in the last decade by the use of recombinant antibodies targeting proinflammatory cytokines such as tumor necrosis factor α (TNFα) (12). However, despite the obvious impact of such therapies, 20–40% of patients do not respond (13), sustained and high-magnitude clinical response is achieved only in a minority of cases (14), and prolonged treatment-free remission has not been obtained. Additionally, these therapies exhibit several adverse side effects that make persistent administration undesirable (14–16). Therefore, the development of new agents that offer greater efficacy and improved safety profiles remains an important goal for the treatment of RA. In this context, tissue-specific drug delivery systems for targeting and improving the retention of bioactive agents are particularly important, as they could be used to achieve higher levels of pharmaceuticals at the site of therapeutic interference and thus prolong local activity within the joint, thereby reducing systemic exposure and toxicity. Recombinant, single-chain antibodies lend themselves well to the development of such targeted pharmacodelivery strategies.
To date, joint-specific targeting for the treatment of arthritic disease remains an unmet clinical goal. In order to address this, we have developed a synovial xenograft model in SCID mice where functional vascular anastomoses allow the delivery of agents targeting human synovial tissue to be assessed in vivo (17). Previously, we successfully used this model for in vivo peptide phage display and imaging of synovial tissue (18, 19). Here, we have extended the use of this model system to carry out in vivo single-chain Fv (scFv) antibody phage display screening, in order to identify scFv antibody clones with specificity for the human synovial vasculature of the xenografts. We report for the first time the isolation and characterization of an scFv antibody (A7) that exhibits specificity for the microvasculature of human arthritic synovial tissue, and we discuss its potential application as an innovative recombinant pharmaceutical agent for the treatment of arthritic diseases.
Beige SCID CB17 mice ages 4–10 weeks were used in this study. Human tissues (synovium and skin) were transplanted subcutaneously in a dorsal position distal to the shoulder joints (2 transplants per animal) as previously described (20). Mice were inspected daily, and animal work was performed under a Project License (PPL 70-6109). Human synovial tissue was obtained from RA patients or osteoarthritis (OA) patients undergoing joint replacement. Human skin tissue was obtained from patients undergoing cosmetic surgery. Informed consent was obtained from all patients. Additionally, ethical approval to use human synovial and skin tissue for research purposes was obtained from the Ethics Committee (Local Research Ethics Committee no. 05/Q0703/198).
The human scFv libraries I + J (Tomlinson I + J) (21) were kindly provided by Dr. Greg Winter (Medical Research Council Centre for Protein Engineering, Cambridge, UK). Synovium-specific phage was isolated following 4 rounds of enrichment in SCID mice carrying human arthritic synovial tissue and skin tissue xenografts (18, 19). Selection and enrichment were monitored by phage titration, and the integrity of scFv coding regions from phage from the final round of selection was assessed by polymerase chain reaction using LMB3 and pHEN seq primers (see below). Clones that showed expression of full-length scFv fragments were used to infect the nonsuppressor strain Escherichia coli HB2151 for the production of soluble scFv protein. One hundred colonies were selected and analyzed for scFv expression by enzyme-linked immunosorbent assay for protein A and protein L binding (22).
The DNA sequences encoding the scFv inserts of phage clones from the final round of selection were determined using the vector primers LMB3 (CAGGAAACAGCTATGAC) and pHEN seq (CTATGCGGCCCCATTCA) (21, 23). Sequencing was performed using the Big Dye Terminator v3.1 Cycle Sequencing kit (Applied Biosystems) on an ABI PRISM 3130 Genetic analyzer.
Single-chain Fv fragments encoded by phage from the final round of selection were expressed in E coli HB2151 as soluble, secreted proteins and purified from bacterial culture supernatants by affinity chromatography using protein A–Sepharose Fast Flow Resin (GE Healthcare), as described previously (24). Purified antibodies were analyzed by sodium dodecyl sulfate–polyacrylamide gel electrophoresis and size-exclusion chromatography on Superdex 75 HR10/30 columns (Amersham Biosciences).
Single-chain Fv antibodies were biotinylated using the EZ-Link Sulfo-NHS-SS-Biotinylation kit (Perbio Science). Briefly, purified scFv protein was diluted in 0.5–2 ml phosphate buffered saline (PBS), added to a 20-fold molar excess of 10 mM Sulfo-NHS-SS-Biotin, and incubated on ice for 1 hour. Biotinylated proteins were subsequently purified using spin-column chromatography (Perbio Science) according to the manufacturer's instructions.
Single-chain Fv antibody fragments were radiolabeled with Na125I using the iodogen method (25). Iodination reaction tubes precoated with iodogen were used according to the manufacturer's instructions (Perbio Science). Typically, 25 μg of purified scFv in 150 μl of PBS was radiolabeled to specific activities of 0.15–0.2 MBq/μg. The efficiency of iodination was evaluated by instant thin-layer chromatography and typically found to be >90%. The purity of the labeled scFv was determined by size-exclusion high-performance liquid chromatography.
Two SCID mice bearing double xenografts of human arthritic synovial and skin tissues (2 arthritic synovium and 2 human skin grafts per animal) were injected with 6 μg of biotinylated scFv A7 four weeks after transplantation. Biotinylated anti–hen egg lysozyme antibody, scFv HEL (22), was used as a negative control. The biotinylated antibody fragments were administered via the tail vein in a total volume of 200 μl and were allowed to circulate for 15 minutes, after which time the mice were perfused under terminal anesthesia. The human grafts along with murine tissues were excised and immediately snap-frozen in liquid nitrogen for histologic examination. The tissue-specific localization of soluble scFv A7 was examined by immunohistochemical detection of biotinylated antibody using avidin–biotin–horseradish peroxidase complex (Dako).
Five double-transplanted SCID mice (2 arthritic synovium and 2 human skin grafts per animal) were injected with iodinated scFv antibody 4 weeks following transplantation. Each animal was administered an injection, via the tail vein, of 200 μl sterile saline containing 1.25 μg labeled scFv with a specific activity of 0.16 MBq/μg. Mice were killed 4 hours or 24 hours after injection, and grafts as well as mouse organs were collected for gamma counting. The results were corrected for tissue weight and background radioactivity in the blood pool and expressed as a percentage of the total injected dose. Iodinated scFv HEL was used as an untargeted scFv control.
Slide-mounted frozen tissue sections were fixed in ice-cold acetone. Paraffin-embedded tissues were dewaxed and subsequently treated with proteinase K (Dako) for 4 minutes at room temperature for antigen retrieval. Slides were stained with 1 μg of biotinylated scFv A7 and visualized with avidin–biotin–horseradish peroxidase complex using 3,3′-diaminobenzidine chromogen. The presence of human blood vessels in tissue sections was visualized using mouse anti-human von Willebrand factor (vWF) (Dako), followed by a horseradish peroxidase–conjugated anti-mouse antibody (Dako). Rabbit anti-mouse CD31 (BD Biosciences) and rabbit anti-mouse CD34 (Cambridge Bioscience) were used to detect mouse endothelial cells in murine tissues. An anti–α-smooth muscle actin antibody (Sigma) was used to visualize the stromal component of the microvasculature. Sections were counterstained with hematoxylin, mounted with Depex mounting medium (Dako), and analyzed using a light microscope (Olympus). Images were acquired with CellP Soft Imaging System version 1.2 (Olympus).
Slide-mounted frozen tissue sections were fixed in ice-cold acetone prior to antibody staining. Biotinylated scFv A7 reactivity was detected with Texas Red–conjugated NeutrAvidin (Invitrogen). Mouse anti-human vWF, mouse anti-human CD31 (Sigma), and rabbit anti-NG2 antibody (Millipore) reactivity was detected using goat anti-mouse and goat anti-rabbit antibodies conjugated to Alexa Fluor 488 or Alexa Fluor 594 (Invitrogen). Sections were subsequently mounted in fluorescent mounting media (Vectashield) with DAPI nuclear counterstain (Vector) and examined using an Axioskop 2 microscope (Carl Zeiss). Images were captured by an AxioCam digital color camera using KS300 image analysis software (Carl Zeiss).
Velocity 5.5 imaging software (PerkinElmer) was used to perform thresholded Pearson's correlation coefficient analysis of images in order to accurately quantify and correlate overlap of image pixels from 2 different channels (26). A value of +1 indicates complete pixel-to-pixel overlap of the pixels from the 2 chosen channels. A value of 0 indicates no overlap or correlation of pixels from 2 different channels, and a value of −1 indicates complete disparity/exclusion of pixels from the 2 channels that have been compared.
Results are expressed as the mean ± SEM. Parametric analyses were performed by unpaired 2-tailed t-test using GraphPad Prism software.
In order to select scFv fragments targeting the human synovial microvasculature, 4 cycles of in vivo selection using the human scFv libraries I + J (Tomlinson I + J) (21) were conducted using mice with dual xenografts of synovium (target) and skin (control) tissues. The composition of recovered scFv fragments in the final round of in vivo selection was assessed by examining the ability of 100 phage-encoding full-length inserts to transduce bacteria for secreted antibody protein expression. The functionality of these clones was further confirmed by demonstrating secreted scFv binding to protein A and protein L. Of these 100 clones, 24 expressed secreted scFv antibody at high levels and were subsequently shown to encode the same scFv sequence. Notably, this scFv sequence had already been identified in a previous, independent, in vivo screen of the Tomlinson library, using the same synovial xenograft model system (results not shown). Thus, clone A7, an scFv from the group of 24 identical clones from the final screen, which showed robust soluble antibody expression (>1 mg/100 ml bacterial culture), was chosen for further studies.
Soluble scFv A7 protein was purified as a monomeric protein and used in immunohistochemical analysis to assess binding specificity in human RA and OA synovial tissues in comparison to that of skin, the control tissue used in the in vivo selection process. We examined 15 OA and 8 RA synovial tissue samples and 5 skin samples. As shown in Figure 1, scFv A7 exhibited specific and strong reactivity with the microvasculature of OA and RA synovial tissue. Importantly, scFv A7 exhibited no detectable reactivity with control human skin tissue. The control, nontargeted antibody scFv HEL did not exhibit binding to either synovium or control skin tissue samples.
Following the demonstration that scFv A7 exhibits strong reactivity with the microvasculature of arthritic synovial tissue, we sought to determine which cell types within the microvasculature were recognized by this antibody. To do this, costaining of RA synovial tissues was performed using scFv A7 and the 2 endothelial markers vWF and CD31, and the pericyte-specific marker NG2. As shown in Figure 2, there was no overlap in the pattern of staining observed in RA tissue stained with scFv A7 and vWF or CD31. However, costaining with scFv A7 and the pericyte marker NG2 showed complete overlap in the pattern of cellular staining observed. This demonstrates that scFv A7 recognizes an epitope localized to pericytes and the stromal component of the microvasculature of RA synovium.
In order to increase the accuracy and level of confidence about the degree of colocalization, we used Pearson's correlation coefficient to provide a numeric and nonsubjective analysis (27). The Pearson correlation ranges from +1 to −1, whereby a correlation of +1 indicates complete overlap of pixels from 2 different channels. A value of 0 indicates no overlap, and a correlation of −1 indicates complete pixel disparity/exclusion between the 2 channels being compared. Our Pearson's colocalization analysis of scFv A7 reactivity (red pixels) and CD31 reactivity (green pixels) resulted in a Pearson's correlation of 0.07, demonstrating no colocalization of scFv A7 and CD31 reactivity. Similarly, Pearson's colocalization analysis of NG2 reactivity (green pixels) and vWF reactivity (red pixels) resulted in a Pearson's correlation of 0.01, demonstrating no colocalization of NG2 and vWF reactivity. However, Pearson's colocalization analysis of scFv A7 reactivity (red pixels) and NG2 reactivity (green pixels) resulted in a Pearson's correlation of 0.6, demonstrating significant colocalization of scFv A7 and NG2 reactivity.
In order to examine the specificity of scFv A7 targeting in vivo, SCID mice bearing arthritic synovium (test tissue) and human skin (control tissue) xenografts were injected intravenously with biotinylated scFv A7 or biotinylated scFv HEL (as a negative control). As shown in Figure 3, after in vivo circulation, biotinylated scFv A7 could be detected in the human synovial microvasculature by simply adding avidin–biotin–horseradish peroxidase complex to xenograft sections. In contrast, scFv A7 reactivity was not observed with control human skin tissue xenografts. Additionally, no detectable reactivity was observed with the control scFv HEL antibody (Figure 3). To confirm the vascular reactivity of scFv A7 in the positive grafts and to exclude the possibility that the negative staining observed in human skin grafts and mouse tongue are not due to an absence of vasculature, all tissues were stained for human vWF and/or mouse CD31 (Figure 3). No cross-reactivity with host mouse tissues was observed, as exemplified by examination of mouse tongue tissue, where microvasculature was clearly visible (Figure 3). Taken together, these data further confirm the synovium-specific reactivity of the scFv A7 antibody and its capacity to reach its target in vivo.
To quantitatively assess antibody tissue specificity in vivo, we examined the ability of iodinated scFv A7 to target human synovial tissue xenografts. Iodinated scFv HEL was used as a nontargeted control antibody, while human skin was used as control xenograft tissue. Figure 4 shows the tissue-to-blood ratio of the percentage of the injected dose localizing in each tissue at 4 hours and 24 hours. These data demonstrate that 4 hours following injection, 3-fold more radiolabeled scFv A7 was localized to the human arthritic synovium xenografts than to the human skin xenografts. Further, despite an apparent fall in overall activity of scFv A7 in the synovium at 24 hours, significant differential reactivity was still observed when compared to skin at this time point. Overall, these data further confirm that the scFv A7 antibody retains the synovial targeting specificity of the parental phage clone in vivo.
In order to further investigate scFv A7 binding specificity over and above the original targeted tissues, we examined the reactivity of this antibody using an array for a comprehensive range of normal human tissues. As shown in Figure 5A, scFv A7 did not exhibit reactivity with the various cellular components of the tissues represented on this array. In particular, no detectable reactivity was seen with the microvasculature of organs such as the adrenal gland, ovary, heart, ileum, and esophagus, all of which were shown to be positive for vWF staining with evident microvasculature.
Next we examined scFv A7 reactivity with the microvasculature of normal human synovial tissue obtained from subjects undergoing joint arthroscopy for prolonged, unexplained knee pain that did not develop into arthritic conditions during a 5-year followup survey (28). The results presented in Figure 5B are representative of 11 samples and demonstrate that the microvasculature found in normal human synovium, as detected by vWF reactivity, contains a stromal vascular component as detected by α-smooth muscle actin reactivity. In contrast, scFv A7 showed no reactivity with the microvasculature found in these synovium samples.
Finally, in order to establish whether the reactivity of scFv A7 is specific to the microvasculature of arthritic synovium or a common feature of neovasculogenesis related to the presence of inflammation, we examined scFv A7 staining in tissue samples from patients with Crohn's disease (n = 7) and psoriasis (n = 5), where the presence of microvasculature was detected using anti-human vWF. The results presented in Figure 6 demonstrate that scFv A7 exhibits no detectable reactivity with the microvasculature found in tissues from patients with either Crohn's disease or psoriasis. Thus, these results demonstrate that the target epitope for scFv A7 is absent from normal human tissues and microvasculature and is not expressed in the neovasculogenesis seen in inflammatory conditions. Taken together, these results further support the conclusion that scFv A7 is specific for the microvasculature found in arthritic synovium.
Over the past decade, the therapy of RA has been transformed through the application of recombinant antibodies targeting inflammatory cytokines (24). However, despite the obvious impact of these agents, high-magnitude responses and treatment-free remission remain elusive goals (29, 30). Further, many patients remain nonresponders or partial responders (14), and within the responder cohort a loss of efficacy can be seen over time, as can specific adverse effects (31, 32).
The development of new therapeutics for RA, with the ability to elicit greater clinical responses and acceptable safety profiles, remains an unmet need. Toward this aim, we have used a human synovial xenograft model established in our laboratory (19, 20) to carry out in vivo phage display selection of scFv antibodies with specificity for the human synovial microvasculature. In this model, synovial grafts implanted subcutaneously into SCID mice remain viable and continue to express human tissue–specific markers (17, 20). Using this approach we have isolated and characterized scFv A7, a novel human scFv antibody that efficiently and preferentially targets the synovial microvasculature in RA, and we have demonstrated that this antibody specifically recognizes perivascular cells in this tissue. Abnormalities of vascular morphology and angiogenesis in arthritic synovium have been previously described at the macroscopic, histologic, and molecular levels (11, 33). It is well established that blood vessels of inflammatory tissue lack the tight endothelial monolayer essential for normal barrier function, resulting in increased endothelial permeability (leakiness) and extravasation of immune cells to the extracellular space (34, 35). In this context, given that our in vivo phage screening strategy is against human synovial grafts vascularized by permeable vessels, the selection of an antibody that recognizes a stromal vascular antigen is not surprising.
Our results demonstrate that the reactivity of scFv A7 is specific to the microvasculature of arthritic synovium, since the antibody does not exhibit reactivity with the microvasculature or other cellular components of normal human tissue from a spectrum of organs. Further, expression of the scFv A7 epitope is not a general feature of neovasculogenesis, since we detected no binding to the microvasculature of the tissue from patients with Crohn's disease or psoriasis. These results indicate that the expression of the epitope for scFv A7 is likely to be tissue specific and restricted to the microvasculature found in arthritic synovium rather than a feature of the microvasculature seen in neoangiogenesis or vasculogenesis in inflammatory diseases. The specific reactivity of scFv A7 suggests that the target molecule for scFv A7 may have potential as a biomarker in arthritis and may also have applications as an immunotherapeutic target in the development of new strategies for therapy of this condition.
Angiogenesis is an important and possibly a primary event in the pathogenesis of the chronic inflammatory process of RA (36). Hence, targeting angiogenesis could play a part in a polypharmacy intervention strategy for the treatment of arthritic disease (9, 37–39). In the course of angiogenesis, the associated tissue remodeling leads to the expression and/or exposure of molecules on endothelial and perivascular cells, which are inaccessible, much lower in abundance, or undetectable in healthy adult tissues. For example, the oncofetal extra domain B (ED-B) of fibronectin represents one of the best characterized markers of angiogenesis (37) and is abundantly expressed in many diseases including RA (38, 40). Moreover, antibody-mediated targeted delivery of proinflammatory cytokines using L19, the human antibody specific for ED-B (41), can result in a significant increase in the therapeutic index of biopharmaceuticals in animal models of cancer (42, 43) and has recently been evaluated for interleukin-2 (IL-2) and TNFα delivery in phase I and II clinical trials (39). Further, L19 together with F8, an antibody specific for the ED-A of fibronectin, have been used to deliver IL-10 to inhibit the progression of collagen-induced arthritis (38, 44). Most recently, 3 recombinant human antibodies specific for matrix metalloproteinases 1, 2, and 3 have been developed and are currently being evaluated for antibody-based pharmacodelivery applications in arthritis (45).
Taken together, these findings clearly demonstrate the utility of perivascular and stromal targeting for the development of ligand-based strategies for treatment of arthritic disease. In this context, we have shown that in vivo, scFv A7 can target the microvasculature of arthritic synovium efficiently and is preferentially retained on the target tissue for at least 24 hours after systemic administration. These data provide functional evidence for the potential use of scFv A7 as an agent to target therapeutics to the arthritic joint.
Although vascular targeting research has mainly focused on tumor angiogenesis, the development of nononcologic applications has recently gained momentum and is likely to become an important area of pharmaceutical intervention. Over the last decade, a spectrum of innovative bispecific antibody formats has been described, with scFv fragments being extensively used as fundamental building blocks to develop 13 novel antibodies that are currently in clinical phase I and II trials (46), as well as a spectrum of new candidate antibodies with biologic potency (47, 48). As described here, scFv A7 represents a new building block for the development of vascular targeting of biopharmaceuticals, capable of selective accumulation at neovascular sites in RA.
All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Pitzalis had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study conception and design. Kamperidis, Kamalati, Ferrari, Jones, Garrood, Smith, Diez-Posada, Hughes, Finucane, Mather, Nissim, George, Pitzalis.
Acquisition of data. Kamperidis, Kamalati, Ferrari, Jones, Garrood, Smith, Diez-Posada, Hughes, Finucane, Mather, Nissim.
Analysis and interpretation of data. Kamperidis, Kamalati, Ferrari, Jones, Garrood, Smith, Diez-Posada, Hughes, Finucane, Mather, Nissim, George, Pitzalis.
We are grateful to Professor Thomas MacDonald for providing tissue samples from patients with Crohn's disease and Professor Rino Cerio for providing tissue samples of psoriatic skin. We thank Dr. Vineeth Rajkumar for advice and assistance with image analysis.