Can CD47 blockade help bridge innate and adaptive immunity?

The recent Nature Medicine paper, "CD47 blockade triggers T cell-mediated destruction of immunogenic tumors," suggests that dendritic cells might have a larger than previously appreciated role in the anti-tumor efficacy of CD47 antibodies. Previous work, mostly from the Weissman lab at Stanford, had suggested that macrophages were the predominant effector cell for CD47 antibodies. Macrophages, in general, are the major phagocytic cell in the body, eating pathogens and dead or dying cells. CD47 is thought to serve as a "don't eat me" or anti-phagocytic signal on the surface of cells. CD47 is expressed on normal cells, and its expression is upregulated on blood stem & progenitor cells when they are mobilized to enter the peripheral blood. Specifically, CD47 expression on red blood cells appears to limit their destruction, and a potential toxicity that has emerged from preclinical testing of CD47 antibodies is the depletion of red blood cells, and subsequent anemia.

CD47 has emerged as a potential target in cancer, as it is frequently overexpressed on a variety of tumor types. CD47 is potentially a way cancer cells protect themselves from phagocytosis, since they can have elevated levels of pro-phagocytic signals. Initially found to be upregulated in hematologic malignancies such as AML by the Weissman group (papers here & here), CD47 was subsequently found to also be overexpressed in a wide variety of solid tumors as well. Furthermore, in each of these papers, the Weissman lab demonstrated that an antibody that blocks the interaction between CD47 on tumors & SIRPalpha on macrophages (and other immune cells) has substantial anti-tumor activity against human tumors xenografted into immunodeficient mice. In some additional experiments by the Weissman lab, they found that macrophages, after having phagocytosed cancer cells, could present antigens and prime a T cell response. However, these experiments were primarily done in vitro, not in live animals with an intact immune system.

Here are the things I took away from the paper:

Intratumoral anti-CD47 can cause rejection of immunogenic mouse tumors:
To better understand the effect of CD47 antibodies in the context of a complete immune system, the authors of this paper tested CD47 antibodies in syngeneic mouse models, where mouse tumors could be transplanted into mice with fully intact immune systems. They used two mouse cancer cell lines that were previously characterized as being immunogenic, the A20 (B-cell) & M38 (colon) cancer cell lines. They used a mouse CD47 antibody (MIAP301) to treat these tumors. They started out treating the mice with a systemic dose of 400 ug, which is a similar dose to those used by previous groups. They saw responses, but interestingly then decided "to rule out any effect on peripheral tissues" by injecting a lower dose (50 ug) directly into tumors, which is the regimen they used for the remainder of the paper. With this dosing method, they were able to get substantial anti-tumor activity against established tumors.

Anti-tumor activity of CD47 antibodies is dependent on CD8+ T cells & dendritic cells:
Testing the tumors & treatment regimen in different mouse backgrounds that lack different components of the immune system, the authors found that CD8+ T cells were required for activity, as well as dendritic cells. Specifically, the inflammatory STING pathway in dendritic cells, which senses cytosolic DNA, was necessary for CD47 antibody activity. CD47 antibodies were found to stimulate the dendritic cells to allow them to prime an immune response in endogenous T cells against tumor antigens. Interestingly, depletion of tumor associated macrophages with a CSF1R CSF1-blocking antibody had no effect on response to CD47 antibodies. Additionally, in vitro, CD47 antibodies were able to enhance the cross-priming ability of dendritic cells more significantly than macrophages, suggesting macrophages may not play as large of a role in this ability.

Further, after intratumoral administration of CD47 antibodies, these mice were subsequently able to reject tumors upon rechallenge with tumor cells, suggesting a systemic immunological memory against the tumor cells was established. In the supplemental figures, if they injected the drug at the higher dose & systemically, there was still some efficacy in the absence of T cells. Perhaps the low dose & local administration accounts for some of the discrepancy between the results here versus the human xenograft immunodeficient experiments.

Some additional reasons for the discrepancies between the relative importance of macrophages (previous work) versus dendritic cells (this paper) that were offered:

1. The immunodeficient mice used previously, NSG mice, have a SIRPalpha that can bind human CD47, and thus they might be more susceptible to antibody blockade.
- Not sure what the reasoning is here for why this would be more significant for human CD47 bound to NSG SIRPalpha versus mouse CD47 bound to mouse (BALB/c) SIRPalpha.

2. In xenograft models, the only targets for the human CD47 antibody is CD47 on the human tumor cells, whereas in the syngeneic mouse models, the mouse CD47 antibody can bind to either CD47 on the tumor or on normal mouse tissue that also expresses CD47.
- Trillium actually has data that the "antigen-sink" effect of having target present on both normal and tumor cells doesn't mitigate the efficacy in AML lines xenografted into the NSG mouse, shown below:

Trillium Therapeutics, ASH 2013

3. The cross-priming ability of dendritic cells may have been enhanced in this paper's in vitro assays due to the fact they used culture conditions the authors say allow for better dendritic cell growth. Thus, previous work may have underestimated the ability of dendritic cells in cross-priming of T cells in the presence of CD47 antibodies.

Additionally, depending on the levels of CD47 expression and the immunogenicity of the tumor, it's possible that the relative contributions to anti-tumor activity of macrophages, dendritic cells or NK cells may be different between different tumors.

Some experiments I would have liked to see them do within the context of this paper:
1. They don't appear to quantify changes in T cell infiltration or PD-L1 expression changes on tumor from before to after treatment, which may have shown the expected change to a more "T cell-inflamed" microenvironment.

2. I also would have liked to see the induction of a systemic immune response by implanting 2 tumors on either mouse flank and only injecting one, and seeing if the contralateral tumor shrunk in response to CD47 antibody injection. A necessary control would be injecting low dose anti-CD47 into a non-tumor flank with presumably little systemic effects on the contralateral tumor.

Potential combinations:
The necessity of STING for the anti-tumor activity of anti-CD47 antibodies suggests that an intratumoral injection of CD47 antibodies can activate STING. This pathway has garnered some interest by biotech, with Aduro seemingly the furthest advanced in developing STING agonists. They fairly recently partnered this program with Novartis. The hope is that activation of this pathway in innate immune cells in the tumor microenvironment can stimulate the infiltration and activity of T cells. The hypothesis is that PD-1/PD-L1 antibodies work best when there is already an active immune response at the tumor, that is being suppressed by those pathways. There are other tumors that lack this immunophenotype and are characterized as non-T cell-inflamed. This is illustrated in the figures from a review by Thomas Gajewski and colleagues:

T cell-inflamed microenvironment

Non-T cell-inflamed microenvironment

Gajewski's lab has recently published a paper showing that intratumoral injection of STING agonists, developed with Aduro, could increase T cell activation, leading to a strong anti-tumor immune response. Improving the T cell infiltration of a tumor may cause tumor PD-L1 levels to increase in response. Thus, it is logical to combine STING agonists with PD-1/PD-L1 blockade, which was also recently published in preclinical mouse models with even greater anti-tumor activity than either agent alone.

Now, there is a difference between STING being required for the activity of CD47 antibodies versus something being a direct STING agonist. For instance, the CD47 antibodies are probably much more dependent on tumor CD47 levels for their ability to cause all their potential downstream effects. We don't know the amount of STING activation with anti-CD47 antibodies versus direct STING activation, or how differences inherent to a specific tumor or its microenvironment would affect the indirect apparent STING activation after CD47 antibody treatment. However, I would imagine that this suggests CD47 antibodies could rationally be combined with PD-1/PD-L1 blockade. As mentioned earlier, it would be interesting to see how intratumoral anti-CD47 affects T cell infiltration. The Weissman lab did use one syngeneic model in Figure 6 of this paper from Willingham et al., and found that there was increased T cell infiltration after treatment, although this was not quantified.

Conclusions:
This paper demonstrated that CD47 antibodies can potentially stimulate a CD8+ T cell adaptive immune response, and this is largely dependent on dendritic cells. While this is somewhat at odds over which innate immune cell type is critical for the cross-priming activity, both this group and the Weissman group have shown these antibodies have the potential to induce an adaptive immune response, and this is in addition to any direct effects through phagocytosis or ADCC. The dependence of the anti-tumor activity on T cells and the STING pathway in dendritic cells is of interest, as it suggests this therapeutic approach may make for a good combination with immune checkpoint blockade. One final point is that intratumoral administration of anti-CD47 could induce these effects. This could potentially be a different route of administration, and while the previously mentioned paper from Willingham et al. found that intratumoral administration still led to systemic exposure, the lower doses used in this paper maybe could limit this. Perhaps this could somewhat mitigate the potential for systemic toxicity (against red blood cells, for instance), that might be dose-limiting, as CD47 antibodies are being developed in the clinic. While this paper presents preclinical data (like all the other current data on CD47-targeted therapeutics), it does add useful information about how this approach may also stimulate an adaptive immune response against tumors and could inform their clinical development.


Edit: DrX published a comment below that brings up some points about the methodology of the paper for distinguishing the effect of macrophages and dendritic cells that is worth checking out.





Disclosure: I own shares of Trillium Therapeutics and Aduro Biotech