Directed evolution of broadly crossreactive chemokine-blocking antibodies efficacious in arthritis

Chemokine receptors typically have multiple ligands. Consequently, treatment with a blocking antibody against a single chemokine is expected to be insufficient for efficacy. Here we show single-chain antibodies can be engineered for broad crossreactivity toward multiple human and mouse proinflammatory ELR+ CXC chemokines. The engineered molecules recognize functional epitopes of ELR+ CXC chemokines and inhibit neutrophil activation ex vivo. Furthermore, an albumin fusion of the most crossreactive single-chain antibody prevents and reverses inflammation in the K/BxN mouse model of arthritis. Thus, we report an approach for the molecular evolution and selection of broadly crossreactive antibodies towards a family of structurally related, yet sequence-diverse protein targets, with general implications for the development of novel therapeutics.

The paper describes the results of a colossal effort, in trying to generate broadly-crossreactive antibodies, capable of simultaneous recognition of multiple (ELR+)CXC chemokines. The goal to isolate one antibody clone, which could recognize with high affinity ALL (ELR+)CXC chemokines was not reached. However, two attractive clones (SA129 and SA138), which exhibited cross-reactivity with certain chemokine targets, were isolated and characterized in detail. One of these antibodies (SA138), used as scFv-albumin fusion, inhibited the activity of certain chemokines and provided an inhibition of arthritis progression in a murine model of the disease, which features a transient elevation of clinical scores.
The paper is exemplary in the thoroughness of the approach that was taken, as well as in the quality and completeness of documentation. Even though "total" cross-reactivity was not achieved, one of the clones exhibited encouraging biological activity. Most importantly, the paper highlights the potential and the limitations associated with the desire to generate cross-reacting antibodies against homologous protein targets, using a combination of protein evolution and yeast display techniques.
I recommend that this article is accepted for publication. The only request that I have (and that I strongly believe should be implemented in the article) is that the authors "reduce" the tone of some statements and provide a more balanced evaluation of the results that they have achieved, as well as of the challenges that still remain. The therapeutic results, while promising, are not spectacular. In an ideal world, I would have liked to see SA138 tested on a second model of arthritis (e.g., collagen-induced arthritis), but I understand that already a lot of work had gone into this article. Moreover, while the wish to engineer cross-reactive antibodies remains important, there are clear limitations (which are also evident from the results of this very extensive discovery program). In other words, the methodology and the results obtained (both the positive and the negative ones) are the most valuable aspects of the paper, not necessarily individual "wow factors".
Reviewer #3 (Remarks to the Author): There is also unmet need for the treatment of arthritides including rheumatoid arthritis (RA). The role of chemokines and chemokine receptors have been widely characterized in the pathogenesis of RA. Yet, almost all studies using antibodies or small molecule inhibitors of chemokines or chemokine receptors failed. Reasons for failure may include cross-species incompatibility, processing of the targeting molecule, wrong target and timing, receptor occupancy, etc. Until know, CC chemokines were mostly chosen as targets, several CCR1, CCR2 and CCR5 antagonists were developed but all failed in phase I-II trials due to reasons described above. Therefore one could come to the conlcusion that anti-chemokine/receptor targeting using conventional antibodies or small molecule inhibitors is a "no-go" in RA. The present study utilizes a significantly different approach: using single-chain antibodies with multiple chemokine specificities. This is an absolutely original approach, no published data have become available using this approach in the treatment of arthritis. Authors picked two albuminbound antibodies, SA129 with specificity for CXCL1 and SA138 with multiple specificities for CXCL1,2,3,5. All these chemokines bind to the CXCR2 receptor and CXCR2 has been characterized as the most important CXC receptor in RA binding gro-alpha, ENA-78, IL-18. In addition, these 3 chemokines, also identified in the present study, seem to be the most relevant in RA (maybe authors should cite the 3 original papers that defined the role of gro-alpha, ENA-78 and IL- 3) The multi-specific SA138 is a lot more effective in all aspects than the mono-specific SA129. These are my remarks: 1. There has been only one study in adjuvant-induced arthritis using CXCR2 blockade. (Barsante et al, Br J Pharmacol 2008, needs to be cited). This study was very successful in animals, yet CXCR2 blockade did not evolve as a target in human RA. What could be the reason that chemokine targeting successful in animals would not work in humans? Why would the approach of the present study be any different? Do authors propose that their approach would be more successful in human RA? 2. In general, targeted therapies with broader specificity may have significantly more side effects. Did the authors observe any complications in the animals? Would they propose more side effects when using SA138 compared to SA129 due to the fact that CXCL2, CXCL5 blocked by the latter but not the former may have important role in host defense? 3. The ELR motif has been implicated in angiogenesis. Authors target ELR-positive CXC chemokines. So when they scored the grade of inflammation, could they also score vascularity of the synovial tissue? 4. What was the rationale of measuring ankle thickness rather than paw volume or ankle circumference? 5. Authors write "We next quantified the number of synovial fluid neutrophils isolated from the arthritic joints of mice treated with SA129, SA138, and SACTR antibody fusions. Synovial tissues were harvested at the peak of the disease (day 8 after disease initiation). We discovered that..." I am a bit confused. Did they actually assess synovial fluid neutrophis or synovial tissue cells? Certainly, in a chemokine study, the quantification of synovial tissue cells would be more relevant. 6. Finally, taking together their data and previous failures, would the authors summarize why their approach would be more successful and would lack all issues that led to failure of previous studies?
We thank the reviewers for their positive comments regarding our manuscript, and for pointing out that our study is "really thorough and comprehensive" (rev #1), a "colossal effort" that is "exemplary in the thoroughness of the approach that was taken, as well as in the quality and completeness of documentation" (rev #2), and represent "an absolutely original approach" (rev #3).
At the same time, the reviewers make important points, which we addressed by addition of a number of experiments to the revised version (major changes in the main text are marked in blue font).

Referee #1 (Remarks to the Author):
This is a really thorough and comprehensive study by Angelini and colleagues developing and testing anti-CXC chemokine antibodies. The redundancy of the chemokine system is difficult to deal with from the therapeutic point of view and the others study a system to generate diversity in antibody repertoire in an attempt to generate broad-acting antibodies. They eventually manage to develop a few antibodies with broad specificity for the CXC chemokine of choice, study antibody-epitope binding and test these in vivo. Antibodies raised worked against both the human and murine proteins.

Authors' reply:
We thank the reviewer for the positive comments and for his/her appreciation of our approach.
I do not really like the title. Perhaps "Directed evolution of broadly crossreactive chemokine-blocking antibodies efficacious in murine model of arthritis." Rheumatoid arthritis (RA) is a very chronic human disease, whose aspects are only partially modeled in mice. The model chosen is reasonable it is far-fetched to say this is RA.

Authors' reply:
We agree with the reviewer's comment that rheumatoid arthritis is only partially modeled by existing mouse models, including the one used in this work. We changed the title as suggested, and also edited the abstract.
To me there are two major issues. First, I take the point of the authors that " this is the first systematic study reporting the selection strategy for the in vitro directed evolution of binders with such extensive promiscuity towards a panel of structurally related, yet sequence-diverse, protein targets." This is the major message and it is a clear methodological advance with good results. I am less convinced this is much better than available tools for ligands for CXCR1/2 receptors. Yes, I agree that the chemokine system is probably an excellent system to test the proof of concept of developing broad-acting antibodies. And choosing neutrophil-active CXC chemokines is also reasonable. The authors acknowledge this and mention that current available "antagonists have shown only limited therapeutic effects. Failures of such receptor-based therapies have often been attributed to (i) differences between the orthologous rodent (pre-clinical) and human (clinical) systems and (ii) the extremely high doses of antagonist required to guarantee continuous receptor occupancy, such that all receptors in the body are antagonized". Point (ii) may be a problem for the developed antibodies as acknowledged by the authors ("To ensure complete inhibition of all ELR+ CXC chemokines present in circulation, we used fairly high doses of our engineered antibody fusions (50 mg/kg)" I must be said there are excellent CXCR1/2 antagonists that function against the human, murine and rat receptors. These molecules have been found to be excellent at blocking neutrophil influx in vivo. Authors cite three of such references (7)(8)(9). Some of the available antagonists are allosteric inhibitors and do not necessarily suffer from the need to be used at very high doses.

Authors' reply:
We thank the reviewer for identifying this imprecise statement. We agree with the reviewer on the existence of non-competitive allosteric modulators of CXCR1 and CXCR2 receptors showing advantages over conventional molecules and therapeutic efficacy in both preclinical and clinical settings. We modified the introduction accordingly and added the following references: Athough the CXCR1 and CXCR2 allosteric inhibitors appear to provide numerous advantages over conventional drug formats, we believe that our study presents an alternative approach to the existing CXCR1 and CXCR2 receptor-based therapies.
In the summary, authors conclude that "The present study demonstrates that simultaneous blocking of multiple CXCR2 ligands is a promising strategy to treat rheumatoid arthritis, with general implications for the development of novel therapeutics for multifactorial diseases." Blocking CXCR2 is already known to be promising for rheumatoid arthritis (not fulfilled). It may be that broad-acting antibodies will be better. I find the abstract does little to help to "sell" the major point of the paper (mentioned at the beginning of the paragraph). The discussion is way better.

Authors' reply:
We agree with the reviewer that the therapeutic potential of CXCR2 blockade is not the major conclusion of the data presented in our manuscript. We modified the abstract to emphasize as the major finding the development of a novel approach capable of targeting multiple soluble factors with a single broadly crossreactive molecule.
Another major point I had refers to the testing of the best final antibodies against other chemokines. At least to evaluate whether enhancing effects against other chemokines would not be accompanied by non-specific binding to most chemokines -specially because they could map some antibody binding to carbohydrate binding sites. It is mentioned that " Importantly, this was not merely due to non-specific polyreactivity of our engineered antibodies, as no binding was detected toward a panel of unrelated proteins ( Supplementary Fig. 7e)." I cannot see why other chemokines and even other non-chemokine chemoattractants (e.g.. C5a) have not been used as irrelevant protein.

Authors' reply:
We are grateful to the reviewer for raising this important point, which prompted us to carry out an experiment that was missing from our original submission.
We have now measured the binding of our engineered CK129, CK138 and CK157 singlechain variable antibody fragments toward twenty soluble CXCL-SA fusions, five structurally related CCL chemokines (CCL2, CCL5, CCL20, CCL22, CCL28), and eleven structurally unrelated proteins, including the murine complement component C5a. The new data are shown below and have been also included in Supplementary Fig. 7e of the revised manuscript and commented on page 9.
Columns graph reporting the fluorescence binding intensities of yeast-displayed CK138 (blue), CK157 (dark gray) and CK129 (red) against twenty soluble CXCL-SA fusions, five structurally related CCL chemokines (white; CCL2, CCL5, CCL20, CCL22, CCL28; Supplementary The five CCL chemokines where chosen based on i) their ability to bind different CCR receptors (CCL2 binds CCR2; CCL5 binds CCR1, CCR3, and CCR5; CCL20 binds CCR6; CCL22 binds CCR4; CCL28 binds CCR10) and ii) their diverse amino acid sequences (selected CCLs share 20-35% sequence identity among them and 12-26% sequence identity with ELR+ CXC chemokines). The experiment showed that the binding of yeastdisplayed CK129 and CK138 towards CCLs and C5a is weak and comparable to that observed with structurally unrelated proteins. Contrariwise, yeast-displayed CK157 exhibits crossreactivity toward three CCL chemokines (CCL20, CCL22, CCL28) and C5a. This last result is not surprising because, although chemokines can be very different in linear sequence, they share a conserved three-dimensional fold. Moreover, all CC and CXC chemokines as well as C5a (pI = 9.7) are highly positively charged proteins that could potentially interact with the negatively charged patches present on the surface of CK157 (see Supplementary Fig. 13). This hypothesis appears to be supported by our epitope mapping experiments and tridimensional structure models that seem to confirm that the binding mode of CK157 could be similar to that of naturally evolved broadly crossreactive viral chemokine binding proteins (vCKBPs) 1-6 , where a large negatively charged area engages conserved positively charged glycosaminoglycans (GAG)-binding regions located on the surface of CC and CXC chemokines, thus explaining their extensive promiscuity. Nonetheless, it is also relevant to underscore that some chemokines (CXCL7, CXCL8, CXCL9, CCL2 and CCL5) and ten structurally unrelated proteins are not recognized by CK157, implying a unique mode of binding and still restrained crossreactivity. Further studies will be required to fully understand the underlying molecular basis of CK157 binding promiscuity, an effort that is currently ongoing in the laboratory, but which we would reason is beyond the scope of a first description of these molecules.
We again thank the reviewer for the positive comments and for his/her appreciation of our work.

Referee #2 (Remarks to the Author):
The paper describes the results of a colossal effort, in trying to generate broadlycrossreactive antibodies, capable of simultaneous recognition of multiple (ELR+) CXC chemokines.
The goal to isolate one antibody clone, which could recognize with high affinity ALL (ELR+)CXC chemokines was not reached. However, two attractive clones (SA129 and SA138), which exhibited cross-reactivity with certain chemokine targets, were isolated and characterized in detail.
One of these antibodies (SA138), used as scFv-albumin fusion, inhibited the activity of certain chemokines and provided an inhibition of arthritis progression in a murine model of the disease, which features a transient elevation of clinical scores.
The paper is exemplary in the thoroughness of the approach that was taken, as well as in the quality and completeness of documentation. Even though "total" cross-reactivity was not achieved, one of the clones exhibited encouraging biological activity. Most importantly, the paper highlights the potential and the limitations associated with the desire to generate cross-reacting antibodies against homologous protein targets, using a combination of protein evolution and yeast display techniques.
I recommend that this article is accepted for publication. The only request that I have (and that I strongly believe should be implemented in the article) is that the authors "reduce" the tone of some statements and provide a more balanced evaluation of the results that they have achieved, as well as of the challenges that still remain.
The therapeutic results, while promising, are not spectacular. In an ideal world, I would have liked to see SA138 tested on a second model of arthritis (e.g., collagen-induced arthritis), but I understand that already a lot of work had gone into this article.
Moreover, while the wish to engineer cross-reactive antibodies remains important, there are clear limitations (which are also evident from the results of this very extensive discovery program).
In other words, the methodology and the results obtained (both the positive and the negative ones) are the most valuable aspects of the paper, not necessarily individual "wow factors".

Authors' reply:
We thank the reviewer for the positive comments and for his/her appreciation of our approach. We agree with the Reviewer's suggestion to reduce the tone of some statements and tried to provide a more balanced evaluation of our results. We added a paragraph in the discussion section highlighting the limitations of our approach, and we state clearly the need of additional testing in other animal models of arthritis to better evaluate the therapeutic potential of our approach (page [15][16]. Moreover, we edited several sentences throughout the text.
We again thank the reviewer for the positive comments and for his/her appreciation of our work.

Referee #3 (Remarks to the Author):
There is also unmet need for the treatment of arthritides including rheumatoid arthritis (RA). The role of chemokines and chemokine receptors have been widely characterized in the pathogenesis of RA. Yet, almost all studies using antibodies or small molecule inhibitors of chemokines or chemokine receptors failed. Reasons for failure may include cross-species incompatibility, processing of the targeting molecule, wrong target and timing, receptor occupancy, etc. Until know, CC chemokines were mostly chosen as targets, several CCR1, CCR2 and CCR5 antagonists were developed but all failed in phase I-II trials due to reasons described above. Therefore one could come to the conlcusion that anti-chemokine/receptor targeting using conventional antibodies or small molecule inhibitors is a "no-go" in RA.
The present study utilizes a significantly different approach: using single-chain antibodies with multiple chemokine specificities. This is an absolutely original approach, no published data have become available using this approach in the treatment of arthritis. Authors picked two albumin-bound antibodies, SA129 with specificity for CXCL1 and SA138 with multiple specificities for CXCL1,2,3,5. All these chemokines bind to the CXCR2 receptor and CXCR2 has been characterized as the most important CXC receptor in RA binding gro-alpha, ENA-78, IL-18. In addition, these 3 chemokines, also identified in the present study, seem to be the most relevant in RA (maybe authors should cite the 3 original papers that defined the role of gro-alpha, ENA-78 and IL-8 The study is focused, aims are clear. Methodology is extensive, using relevant modern techniques. The K/BxN model is relevant for human RA. The most important results (in addition to development and molecular characterization of the antibodies) are: 1. The developed antibodies successfully inhibit neutrophil ingress into the synovium. 2) These molecules can both prevent and treat murine arthritis.
3) The multi-specific SA138 is a lot more effective in all aspects than the mono-specific SA129.
I am happy with the modifications made.
Reviewer #3 (Remarks to the Author): I am happy with the changes. I have no further comments.