Category Archives: Uncategorized

What aren’t we better as weighing results and making sound conclusions?

Yesterday on Biotech Clubhouse the conversation veered from the anti-vaccine movement to recent FDA actions. It was an interesting discussion, and got me wondering:

What aren’t we better as weighing results and making sound conclusions?

Whether its vaccines preventing SARS-Cov2 infection, or greenhouse gases causing Climate Change, we find it hard to convince people that our hypotheses are supported by the available evidence.

First, it’s not new. We had the “HIV doesn’t cause AIDS” debacle, aka HIV Denialism, spearheaded most famously by the now-disgraced molecular biologist Peter Duesberg and most damagingly by the former South African President Thabo Mbeki.  Mbeki in particular railed against testing, and delayed the deployment of retroviral therapies in South Africa for years, policies that led to the unnecessary death of hundreds of thousands of citizens. President Ronald Reagan remained stone-cold silent on the subject of HIV and AIDs for more than 4 years while the epidemic grew ever larger in the US.

You think we’d learn.

Scientific denialism is the extreme result of disputing scientific findings are disputed on irrational grounds. As we see in the anti-vaxxer community, the arguments are not scientifically grounded but rely instead on a bevy of illogical arguments, including rhetorical fallacies, eg. appeals to false authority, appeals to “fairness”, and the right to disagree.  But at heart, ignoring and discounting observations that clearly support critical scientific hypotheses is a unifying theme of denialism.

You might wonder: why focus on these hypotheses?  Why not simply state the facts. The answer is that we can find gradations of scientific denialism that extend from crazy anti-vaxxers to the recent aducanumab approval, with a lot of gray area in between that involves hypothesis testing.

This is a hypothesis that is also considered to be a true statement:


This is a true statement as far as we have tested it – no spiders have been found that have more or fewer than 8 legs, and those with less have accidently lost a leg that was originally there. This exercise, of viewing (collectively) many spiders and deriving the statement is an example of inductive reasoning, meaning we have taken individual observations that THIS SPIDER HAS 8 LEGS, which are singular statements, and have formulated the hypothesis that ALL SPIDERS HAVE 8 LEGS, and having (collectively) failed to disprove the hypothesis it has become a universal statement, considered true.

Formally, this is an exercise in evidence-based epistemology. We can ignore the philosophy though, and focus on what “evidence-based” means here:

1) we started with observations: THESE SPIDERS HAVE 8 LEGS.

2) we (perhaps subconsciously) formulated a hypothesis – NO SPIDERS EXIST THAT DO NOT HAVE 8 LEGS (unless they lose 1 or 2 by accident)

3) we collectively have tested the hypotheses and drawn the appropriate evidence-based conclusion that, indeed, ALL SPIDERS HAVE 8 LEGS.

So far, we’re good. But there is an asymmetric element to our hypotheses, first articulated by Karl Popper. Note that the universal statement ALL SPIDERS HAVE 8 LEGS can never be verified by any number of singular statements because there is in theory always another spider to examine. However, they can be refuted by a singular statement, eg.


(Popper used ALL SWANS ARE WHITE as his example of a universal statement, readily disproved by the appearance of a BLACK SWAN, an image repurposed to describe a rare and unexpected “black swan event”. To be fair, black swans are rather common which is why Popper used them in his example.)

Popper used the asymmetry inherent in the hypothesis (impossible to prove, easy to disprove) to formulate new criteria for statements about the world, specifically, that scientific statements are fundamentally different from other classes of statements such as metaphysical statements (eg. “man is good”). In Popper’s view, scientific statements can provide evidence-based truth because they can be falsified. His “Criteria of Falsifiability” states that “only those hypotheses which can potentially be contradicted by singular statements qualify as scientific”.[1]

It’s never so cut and dry as spider’s legs, and the acceptance or rejection of a hypothesis can be less straightforward than the weight of any single observation (thus we use statistics to bracket the uncertainty of our measurements). Regardless we can make the statement that the “empirical content” of a hypothesis is a useful measure of its inherent value.

Simply put, there is more empirical content in the statement ALL SPIDERS HAVE 8 LEGS than in the statement ALL SPIDERS HAVE LEGS.

Empirical content reflects the testability and falsifiability of a hypothesis, and an empirically rich hypothesis yields many predictions and therefore many potentially falsifying statements. We can see the richness of complex hypotheses in their ability to generate testable predictions, and the most powerful scientific hypotheses can survive making incorrect predictions. Hypotheses having truly robust empirical power become Theories, as in the Theory of Evolution or the Theory of Relativity.

So, another hypothesis is:


This hypothesis has a lot of empirical content and predicts, at a minimum, that vaccinated individuals should get sick less often than unvaccinated individuals, and that vaccinated individuals should spread the virus to other people less frequently than non-vaccinated individuals. We can make these predictions based by our experience with other vaccines: flu vaccines for example.

When we review the data we find our predictions are robustly supported: in every instance in which it has been studied BNT162b2 prevents SARS-COV2 virus infection and reduces viral spreading. Note that the foundational hypothesis is eminently falsifiable – just ask GSK or Merck or Sanofi or Curevac, all of whose vaccines failed – but with respect to BNT162b2, and Janssen’s JNJ-78436735 vaccine, and Moderna’s mRNA-1273 vaccine – the hypotheses have withstood rigorous testing. So that’s good!

In contrast, a weak hypothesis generates predictions that fail more often than not.  

One example is the hypothesis that hydroxychloroquine can be used to treat SARS-COV2 patients. A study, published in 2005, suggested that chloroquine could prevent SARS (the older virus) from infecting cells kept in culture.[2] This was a simple study published in an obscure journal, but it did make predictions: that chloroquine acted by interfering with the “terminal glycosylation of the cellular receptor, angiotensin-converting enzyme 2” thereby blocking virus/cell interaction, and that chloroquine might be useful in treating SARS patients. This study, accessed over 1M times online and widely shared on social media, was offered as evidence that a form of chloroquine called hydroxychloroquine could treat SARS-Cov2 patients. Thus, the observation (chloroquine blocked SARS infection of cells in a cell culture dish) generated a hypothesis:


This hypothesis was tested clinically, many times over, and failed each time. Note that the empirical power of the hypothesis is not zero: angiotensin-converting enzyme 2 (ACE-2) plays a critical role in mediating SARS-COV2 infectivity, but hydroxychloroquine doesn’t block this pathway in patients. Therefore, the hypothesis failed a critical prediction and must be rejected.

Anti-vaccine sentiment offers an extreme example of Science Denialism, and its adherents have many and complicated ways of rationalizing their views built for example on religious, political and cultural mores. Regardless, the active disregard or denial of results that contradict their viewpoint is a central component of their stance.

In science we expect a more clear-headed evaluation of results in the context of hypothesis testing. And yet, much of the published scientific research literature cannot be reproduced. Venture capital firms and pharmaceutic companies that perform ‘wet diligence’ have consistently concluded that 2/3rds of the results in the literature are not reproducible (wet diligence refers to the hiring of an independent lab to repeat published experiments). Note here that we are not referring to fraud, nor are we saying that 2/3rds of published results are wrong. Science is hard!  Indeed, the reproduction of results is a core element of the scientific literature; failure to reproduce results is a part of the process that drives science forward, because this process allows us to reject a hypothesis and move on.

Except when we don’t: the counter-weight to the process of rejecting hypotheses in the face of data is the use of ad hoc explanations for results that run counter to the predictions made. This practice is remarkably common in academic labs, in biotech and pharma labs, and in investor communities. “Oh the samples must be mixed up, let’s run the analysis again” is one simple example. But the slope is slippery: tossing out inconvenient results as “outliers”, declaring a specific assay or model irrelevant because it didn’t show you what you wanted to see, the mining of statistics in search of significance, aka ‘p-hacking’, selective display of data and so on. Nonetheless, most ad hoc responses to inconvenient results are pretty harmless, since they fail to survive scrutiny. Of course, a lot of people’s time and energy is wasted trying to reproduce bad experiments.

A rule of thumb, articulated by James Farris[3] among many others, is that each use of an ad hoc explanation to dismiss a result predicted by a hypothesis weakens that hypothesis. Farris made this argument in defense of the use of parsimony in evolutionary systematics, where the minimum number of explanations is used mathematically to derive relationships from a data set. This works as well with fossils as it does with genomic sequences, so we see parsimony can be a powerful tool because it specifically reduces the number of ad hoc explanations.

That ad hoc justifications for inconvenient results should be avoided may seem obvious, but again, the slope is slippery, and made more so when pressure is brought to bear on a hypothesis that is not scientific. This is true of our anti-vaxxers as discussed. It appears to also be true of the approval of aducanumab for the treatment of Alzheimer’s Disease (AD).  This issue has been covered in detail, see for example the in-depth analyses by Derek Lowe and by STAT News.[4],[5]

The drug was approved despite inconsistent Phase 3 clinical trial results, and despite the use of post-hoc analysis to identify potentially responding patients. The use of post hoc analysis to generate an ad hoc explanation for inconvenient results sounds confusing, but basically you could say the after the fact (post hoc) the company looked at just some patients to conclude the clinical trial worked, for this population (an ad hoc explanation). As stressed earlier, ad hoc explanations weaken the original hypothesis, in this case that aducanumab can treat AD. If we applied the parsimony principle to the available clinical results, we would reject this hypothesis.

So, what should have happened, of we were to apply scientific rigor here? The hypothesis being tested had changed (which is fine, happens all the time, and should). Specifically, the hypothesis:


was altered to:


That hypothesis could, and should have been, the foundation for a new clinical trial. The approval means the hypothesis will be tested in the public domain, without benefit of the clinical trial design that could tell us if its working, or not. This is an excellent example of ad hoc justification undermining the principle of evidence-based truth.

Back in July the Biotech Clubhouse panel worried aloud that the FDA approval of aducanumab and the resulting outcry and controversy would further confuse, and embolden, a public already resistant to the scientific advice for mask and vaccine use to prevent SARS-Cov2. And here we find the real danger in the slippery slope: if we won’t hold the line on scientific integrity, how can we reason with a skeptical general public?

Next time, counterpoint: how successful hypothesis testing led to novel and effective treatments for a nasty autoimmune disease.

Stay tuned.

[1]  –  from Conjectures and Refutations by K. Popper (1963)

[2] – Virology Journal

[3] – Farris on explanation and parsimony

[4] – Derek Lowe’s blog

[5] – STAT

Advantage Albumin, Immunovant in Limbo

Pathway covered: FcRN

Companies mentioned: Immunovant, Momenta/JNJ, Alexion, Argenix

Immunovant stock (NASDAQ: IMVT) got hammered earlier this month on news of a clinical hold in its phase 2b trial for IMVT-1401, a treatment for thyroid eye disease (Graves Disease). The company said it voluntarily decided to pause dosing “out of an abundance of caution,” due to elevated total cholesterol.  Apparently, cholesterol levels were not measured in the earlier Phase 1/2 trials. The company was hammered again 2 days ago when they released earnings, noting that a second trial in Warm AutoImmune Hemolytic Anemia (WAIHA) was also stopped. This is a bundle of bad news at several levels.

Level 1 – The drug.

 IMVT-1401 is an anti-FcRN antibody, designed to prevent IgG from recycling through FcRN-mediated cell internalization and re-release. This is a heavily prosecuted target in disease indications, like Graves, that are thought to be driven by pathogenic autoantibodies. The thinking is pretty straightforward – preventing recycling of IgG will cause a drop in the entire antibody pool, including the pathogenic ones.

Here is some of the activity ongoing in this space:

Screen Shot 2021-02-18 at 5.53.25 PM

One can appreciate pretty quickly that these programs are tightly focused on just a handful of rare diseases.  IMVT-1401 is Immunovant’s only drug, and so a delay here could become a critical issue in the face of robust competition.

Level 2 – The target.

As noted, FcRN recycles IgG by binding and shunting the bound IgG through the endosomal pathway, thereby avoiding the lysosomal pathway that would break down the IgG rather than release it intact.  A second protein uses the same trick – albumin. Notably, the IgG and albumin binding sites on FcRN are distinct, and anti-FcRN antibodies that block IgG uptake do not necessarily block albumin binding.

But IMVT-1401 from Immunovant appears to block both IgG and albumin, as does Nipocalimab from Momenta/JNJ.  In their 2020 10-K, Immunovant revealed drops in serum albumin to the lower range of normal in their Phase 1 studies: “Dose-dependent and reversible albumin reductions were observed in the single-ascending and multiple-ascending dose cohorts … Mean reduction in albumin levels at day 28 were 20% in the 340 mg multiple-dose cohort, and 31% in the 680 mg multiple-dose cohort. For subjects in the 340 mg and 680 mg cohorts, the mean albumin levels at day 28 were 37.5 g/L and 32.4 g/L, respectively (normal range 36-51 g/L). These reductions were not associated with any AEs or clinical symptoms and did not lead to any study discontinuations.”

Why is this important?

Albumin is a globular protein with many functions – one function is to bind (non-specifically) to lipids in blood, including cholesterol and its fractions (HDL and LDL). Therefore, reduction in the total albumin concentration in blood can result in a change in the amount of cholesterol held within cells versus released into circulation (cholesterol efflux). This causes circulating cholesterol to rise. Indeed, low serum albumin is itself considered a risk factor for cardiovascular disease and stroke. Notably, albumin reduction to even lower levels (below 25 g/L) were reported by Momenta in their Nipocalimab clinical trials (corporate deck dated November 2019, Appendix) but we have not had reports of an impact on total cholesterol with this drug.

One thing to consider in passing is that Momenta claims their anti-FcRN specifically binds the IgG binding site and does not bind the albumin. This suggests that specificity may not be enough here to prevent an impact on albumin levels and, potentially, increase cholesterol. That would be a worst-case scenario- a drug class effect. As Jacob Pleith wrote a few weeks ago, “Immunovant’s decision yesterday to pause clinical trials of IMVT-1401 … sent ripples through the anti-FcRn space..”. In response to the Immunovant hold, Argenix quickly noted that they measured total cholesterol, HDL and LDL in two separate trials of Efgartigimod and saw no impact. So yes, the news caused ripples.

Level 3 – The indication.

A reasonable conclusion is that the problem is indication-specific, and indeed this was the assumption made until the announcement this week that the WAIHA trial was also paused along with the Graves Disease trial.

Graves is a thyroid condition and is associated with elevated cholesterol levels itself. Therefore, the combination of a drug effect on top of the disease effect causes concern. The measured increases were modest – LDL-C increased by ~ 65% in the 680mg dose, ~40% at the 340mg dose, and ~25% in the 255mg dose – but this was after only 12 weeks of treatment, in an indication that would likely require life-long chronic dosing.  One might just shrug and say “pravastatin is free” so just give them statins and get on with it. That’s not a bad proposition for these patients.

All of this would make sense: the drug decreased circulating albumin, causing cholesterol to rise, in patients already at cardiovascular risk … but why stop the WAIHA trial?  We have to await more data from the competing programs in other diseases to fully understand what is happening,

In the meantime, a prediction: the hold is in place while the company gets its regulatory ducks in a row, which is a smart de-risking move, then that hold will be lifted.

Stay tuned.

Fidgeting with TIGIT – Part 2 – pathway complexity

Part 2 of 2

Pathways and targets covered: TIGIT, PVR, PVRL2, PVRIG, DNAM-1

Companies mentioned: Compugen, Surface Oncology, Eli Lilly, Merck, Roche

In Part 1 ( I did a drive-by on TIGIT targeting, more or less in isolation.  But TIGIT exists in a complex web of ligands and receptors, expressed on diverse cell types.  Here is a simplified view:


In the context of anti-tumor immunity we want to know how ligands are expressed on myeloid cells, dendritic cells and the tumor cells in the tumor microenvironment (TME) and how the receptors are expressed on infiltrating T and NK cells.  Expression of both ligands, PVRL2 and PVR, is upregulated in many cancers.  Also, shed PVR is thought to prevent productive signaling through DNAM-1.  TIGIT and PVRIG expression is bright on activated T cells, including PD-1-positive T cells. DNAM-1 expression is variable, in part due to internalization and degradation induced by PVR binding.  DNAM-1 activity is also negatively regulated in cis by TIGIT and is reportedly downregulated in the TME of many cancers.

So, it’s complicated.

As noted above we can break this into two main pathways. Compugen has argued that these pathways represent parallel means of regulating DNAM-1 interactions by coopting ligand binding, ie. that the negative signaling receptors TIGIT and PVRIG both compete with DNAM-1 for ligand engagement. These data below are from a Compugen paper (DOI: 10.1158/2326-6066.CIR-18-0442).  The cytokine data are from a 2-week antigen-dependent activation assay that yields CD8+ T cell responses that can be measured in the presence of blocking antibodies.


The take home message is that one only sees synergistic activation of IFNy when the parallel pathways are both blocked, thus blocking both TIGIT/PVRL2 or TIGIT/PVRIG or PVR/PVRl2 or PVR/PVRIG, but not both sides of the same pathway, eg. TIGIT/PVR.  Here again are the two pathways:


The effect can be duplicated by blocking either side plus blocking PD-1.  Compugen also showed that blocking PVRIG induced upregulation of TIGIT, suggesting a compensatory effect that may be overcome by blocking both sides..

These data are the basis for Compugen’s “triple” combination study in collaboration with BMS (NCT04570839).  The trial will evaluate the simultaneous blockade of three immune checkpoint pathways, PVRIG (COM701), TIGIT (BMS-986207) and PD-1 (nivolumab), in patients with advanced solid tumors including those that have failed anti-PD-1 or anti-PD-L1 as prior therapy.  A quick note here: Compugen’s anti-PVRIG is an IgG4 isotype antibody (like nivolumab) and the BMS anti-TIGIT is a FcγR-null IgG1. The idea here (as with anti-PD-1) is that you don’t want to deplete the T cells, just block the pathways. The role of FcR-engagement in this space is controversial as was discussed in Part 1 (

Compared to the wealth of programs targeting TIGIT only a few efforts to tackle the other pathways have been disclosed.  Surface Oncology presented anti-CD112R (PVRIG) data in an AACR 2020 poster.  This included in vivo data using syngeneic mouse models, including in the rechallenge setting.


Note the focus on comparing antibody isotypes (mouse IgG2a > mouse IgG1), which is reminiscent of the findings surrounding anti-TIGIT antibody isotypes, where Fc-effector function may be important for efficacy (note that mouse IgG2a is equivalent to human IgG1).  We’ve not touched on the complexity of target expression on both T cells and NK cells, but here Surface shows a role for each, as depletion of either cell type prevented activity in their in vivo model:


Similar to Merck’s publication, Surface identified a key role for FcgR-engagement in mediating activity.  Finally, they demonstrated synergy with anti-PD-1 treatment:


Compugen’s anti-PVRIG data were generated with an IgG4, showing additive activity with anti-PD-L1.  These data are compared to gene KO results across three different models (from SITC 2019 poster):


The poster also showed robust biophysical characterization of the anti-PVRIG antibodies where the isotype perhaps mattered less.  The in vivo modeling data are somewhat modest – note the short duration of modeling. Regardless, the combination KO data suggest that blocking both arms (TIGIT and PVRIG) is more efficacious than blocking either single arm, which is in line with their in vitro analyses.

A proposed mechanism of action for blocking two pathways

An interesting model is that blocking either pathway (PVRL2/PVRIG or PVR/TIGIT) will “free” DNAM-1 to engage ligands to transduce a productive immune signal.  Of note TIGIT blockade can also release DNAM-1 (from disruptive interaction in cis) as shown by the Roche group:


Therefore, anti-TIGIT and anti-PVRIG (CD112R) blockade may both increase surface DNAM-1 expression or availability. This is important since several papers have directly examined the role of DNAM-1 expression in immune responses. DNAM-1 is degraded upon activation, in a phosphorylation followed by ubiquitination-dependent manner. Mark Smythe’s lab has examined DNAM-1 degradation in the context of anti-tumor immunity. In vivo modeling data showed an improvement in tumor control when DNAM-1 degradation is blocked, and synergy with immune checkpoint blockade is also demonstrated –  nb. this paper is full of very untraditional gating of flow cytometry data (  A follow-on paper from the same lab uses DNAM-1 expression to examine the functional state of TIL, suggesting that down-regulation of DNAM-1 in the tumor microenvironment contributes to T cell dysfunction – this echoes Compugen’s analyses.  There is little work directed to DNAM-1 itself although apparently Eli Lilly has a agonist antibody in the clinic (LY3435151).

TIGIT-related pathways in IO resistance

TIGIT is reproducibly identified on PD-1-positive T cells and appears as a signal of resistance to anti-PD-1 therapy. DNAM-1 downregulation is consistently seen in TIL subsets linked to exhaustion.   Of course, whether these two observations are linked is not known.  In contrast, PVR upregulation has not been identified in any of the many unbiased profiling studies on mechanisms of resistance to or relapse from anti-PD-1 or anti-PD-L1 therapy.  We should recognize that these relationships are complicated – recall that CGEN showed that PVRIG blockade increased TIGIT expression and Roche showed that TIGIT binding in cis can disrupt DNAM-1 activity.   Some of these features are likely to be seen in the context of anti-PVR blockade – we do not have enough data to know.

Of note we do have enough (clinical) data that shows that single agent anti-TIGIT antibody treatment is ineffective, and co-administration of anti-PD-L1 or anti-PD-1 is needed.  It is reasonable to further hypothesize that antagonism of both sides of this complex inhibitory network – eg. anti-TIGIT plus anti-PVRIG – may produce optimal synergy with anti-PD-1 or anti-PD-L1 therapies.

Compugen has presented anti-PVRIG monotherapy clinical data that suggests some activity:


Several partial responses were reported.  With respect to their conclusions we will want to assess the biomarker data alluded to below:


Compugen, in collaboration with BMS, is running a “triple” study: anti-TIGIT, anti-PVRIG, anti-PD-1 (nivolumab).  This sounds promising and will yield useful information one way or the other, since, as noted earlier, the anti-TIGIT antibody being used is an IgG4, as is the anti-PVRIG antibody – activity with this combination would further complicate our understanding of the mechanisms of action.

Updates will post as we get more clinical data from these interesting targets.

Stay tuned.



Fidgeting about TIGIT

Part 1 of 2

Pathways and targets covered: TGF-beta, PD-L1, PD-1, TIGIT

Companies mentioned: Merck KgaA, GSK, Roche, Merck, Mereo, iTeos, BMS, Arcus/Gilead, Compugen, Seagen, Beigene, Innovent, Agenus

Last week we had the bad news that Merck KGaA and GSK had thrown in the towel on bintrafusp alfa therapy for first-line advanced NSCLC.  Bintrafusp alfa is an anti-PD-L1/TGFbR2 TRAP therapeutic designed to selectively antagonize TGF-beta isoforms 1 and 3 while also blocking PD-L1, thereby delivering two-for-one anti-immunosuppression.  Bintrafusp alfa was being tested in a head-to-head trial vs. pembrolizumab and showed no added benefit in a patient population selected for PD-L1-high tumor expression (50%+ of cells in the tumor biopsy sample positive for expression).

This stirred up a fair amount of discussion, as TGF-beta blocking therapies are in vogue for immuno-oncology (IO), with small molecules, biologics, RNA-antagonists and genetic knockouts (in CAR T cells) all in the pipeline. I have high hopes for this space, despite the news out of Darmstadt. And to be fair, the press release stressed the ongoing bintrafusp alfa trials in bladder cancer, cervical cancer, and NSCLC using various drug combinations, and noted new trials in urothelial cancer and TNBC (  Still, the failure stung, due mainly to the promise of the early (open label) Phase 1 expansion cohort data that had suggested significant benefit from the therapy.

This got me thinking about TIGIT, another hot IO target.  The last time I wrote about TIGIT I ended with this question: “How to select patients who should respond to anti-TIGIT co-therapy (or anti-TIM-3 or anti-LAG-3)…?” ( This is a question we should ask about any pathway – including TGF-beta of course – particularly as we are now in the post-immune-checkpoint era, that is, in a setting where many patients in the most IO-responsive indications like melanoma and NSCLC will have already been treated with an anti-PD-1 or anti-PD-L1.  So, is there anything known about TIGIT expression that can guide us in patient (or indication) selection?

Roche leads the field with tiragolumab an anti-TIGIT Fc-competent IgG1 that has shown activity in combination with the anti-PD-L1 antibody atezolizumab in first-line NSCLC, and only in patients with PD-L1- expressing tumors (> 1% of cells in the tumor biopsy sample positive for expression).  We can pause here to recall that this is about where we started the discussion above regarding the TGF-beta TRAP/anti-PD-L1 asset from Merck KGaA, being trialed in the PD-L1-high (>50%) setting in NSCLC.

In front-line NSCLC (EGFR and ALK wildtype), Roche reported responses higher than with atezolizumab alone. Data were shown at AACR and then updated at ASCO.  Here are some of the ASCO data:


The response rate with dual therapy looks rather better than atezo alone, especially in the PD-L1 high cohort (middle panel).  Atezo alone appears to have underperformed, with an ORR = 21% (left panel, all patient data (ITT)).  In the comparable phase 3 trial of atezo vs chemotherapy in front-line NSCLC (also EGFR and ALK wildtype) the ORR = 38.3% in the atezo arm (n=285) and 28.6% in the chemotherapy arm (n=287), see Regardless the 66% response rate in the PD-L1-high cohort (middle panel) attracted attention.

The PFS data were also striking when compared to the prior trial.  This is tiragolumab plus atezolizumab / PD-L1 high cohort:


We can go back and compare this to the atezo alone Phase 3 interim data shown at ESMO in 2019 (I was stuck in the overflow “room” which was a curtained space on the floor of the Barcelona convention center).  This is the PD-L1-high cohort:


Here the median PFS is 8 months, certainly shorter than what is shown for tiragolumab plus atezolizumab, but again, note the disparity with the atezo alone arm of the study (medPFS for = 4 months).

Just to be clear, here are the PD-L1-high patient data compared:


We’re left with the always troubling question of variability between trials and the possibility that the tiragolumab plus atezolizumab results are a fluke.  Unfortunately, we will have to wait and see.

There are two features here worth noting.  One is that TIGIT, the target, is expressed on T cells, along with PD-1.  So far this makes sense – they might very well synergize, particularly given the function of DNAM-1 in the context of T cell signaling (see part 2).  But the anti-TIGIT antibody is an IgG1 isotype, thought to trigger ADCC and CDC-mediated target cell (ie. the T cell) death.  But we want the T cells, that’s the whole point of blocking PD-L1 with atezo.  So what the heck is going on here?

Merck seems to have an answer, but first, some more data.  Merck’s anti-TIGIT antibody, vibostolimab, like Roche’s tiragolumab, is a wildtype IgG1.  Early data on the combination of vibostolimab and pembrolizumab (anti-PD-1), presented at ESMO2020, looked promising in immune checkpoint naïve patients (75% had prior chemotherapy, the rest were treatment naïve):



We can benchmark these results to monotherapy, just as we did with the Roche data, focusing on the PD-L1-positive subset (here we can see data using a cutoff of >1% or >50% of cells positive in the tumor biopsy):



The results compare favorably with pembro-alone using the >1% PD-L1 cutoff and are similar to pembro-alone using the >50% PD-L1 cutoff.  As usual it is difficult to compare between trials, but the signal is encouraging.

Preclinically, Merck has addressed the MOA, stressing the requirement for the intact Fc functionality imparted by the IgG1 antibody isotype.  As mentioned earlier, the mechanistic puzzle is that canonical IgG1 activity includes the triggering of target cell killing via ADCC and CDC mediated cytotoxicity.  Of course, TIGIT is expressed on the very T cells we want to preserve and activate, not kill.  Given this reality we need alternate hypotheses for the action of the IgG1 antibodies.  The predominant hypothesis is that anti-TIGIT antibodies are selectively depleting T-regulatory cells that are TIGIT-bright and immunosuppressive.  This is reminiscent of the now-T-regulatory cells that are TIGIT-bright and immunosuppressive.  It’s an easy hypothesis to advance, similar to the now-debunked arguments made on behalf of anti-CTLA4 and anti-GITR antibodies, and very likely incorrect.

Merck has demonstrated in preclinical models that antagonistic anti-TIGIT antibodies having a  FcgR-engaging isotype induce strong anti-tumor efficacy whereas anti-tumor activity is drastically reduced when using the same anti-TIGIT antibodies that are null for FcgR-engagement (doi: 10.3389/fimmu.2020.573405). These results are consistent with data presented by multiple groups, eg. Mereo and iTeos.  The Merck team further showed shown that FcgR engagement persistently activated myeloid lineage antigen-representing cells APCs, including the induction of proinflammatory cytokines and chemokines while TIGIT blockade simultaneously enhanced T cell activation including elevated secretion of granzyme B and perforin, which synergizes with anti-PD-1 antagonism.  I favor this hypothesis.  Nb. This suggests we’ve a lot to learn still about the best way to engage Fcg receptors, a theme I introduced in the last post (link).

Where does this hypothesis leave everyone else in the TIGIT space?  Let’s line them up:


A few quick notes: EMD Serono/Merck KGaA and Innovent have anti-TIGIT programs without disclosed isotype information; Arcus has disclosed a second, Fc-competent, anti-TIGIT program (AB308); Agenus is developing both IgG1 and IgG4 anti-TIGIT antibodies.

A question: is Seagen’s hyper-killing IgG1 a step too far?

In summary, we have preliminary data in NSCLC that suggest that anti-TIGIT may synergize with anti-PD-1 or anti-PD-L1 therapies, consistent with the expression of TIGIT on PD-1 positive (ie. activated) T cells.  We have several hypotheses addressing the Fc-end of the therapeutics, and some information on why blocking TIGIT may enhance T cell responses.

Other than selecting patients with PD-L1-positive tumors, can we gate on TIGIT expression?  Apparently not, at least not in NSCLC, as just reported at the World Conference on Lung Cancer (abstract P77.02 – Efficacy of Tiragolumab + Atezolizumab in PD-L1 IHC and TIGIT Subgroups in the Phase II CITYSCAPE Study in First-Line NSCLC).

Here’s their text:

“Among the 135 enrolled patients with PD-L1-positive NSCLC (intent-to-treat [ITT] population), 113 had results from the SP263 assay and 105 had results from the TIGIT assay. The biomarker-evaluable populations (BEP) for both of these assays were similar to the ITT population. Comparable PFS improvement with tira + atezo relative to atezo monotherapy was seen in PD-L1–high (≥50% TC) subgroups defined by SP263 (PFS HR 0.23, 95% CI: 0.10–0.53) when compared with PD-L1-high subgroups defined by 22C3. However, for patients whose tumors were defined as TIGIT-high (≥5% IC), no strong association with PFS improvement was observed.

Biomarker subgroup Subgroup, n (BEP, N) PFS HR (CI) relative to atezo monotherapy arm
ITT (PD-L1 IHC 22C3 >1% TPS) 135 (135) 0.58 (0.39–0.88)
PD-L1 IHC 22C3 (≥50% TPS) 58 (135) 0.30* (0.15–0.61)
PD-L1 IHC SP263 (≥50% TC) 45 (113) 0.23* (0.10–0.53)
TIGIT IHC (≥5% IC) 49 (105) 0.62* (0.30–1.32)
*Unstratified HR

Prevalence of PD-L1 subgroups in the BEP was comparable with previous reports for both IHC assays. The PFS benefit observed with tira + atezo in patients with tumors defined as PD-L1-high by 22C3 was also observed using the SP263 IHC assay, but not in tumors classified as TIGIT-high using an exploratory TIGIT IHC assay. Our results suggest that PD-L1 expression, assessed by 22C3 or SP263, may be a biomarker for tira + atezo combination therapy in metastatic PD-L1-positive untreated NSCLC.”

So that the answer to the question we started with, can we pick patients, is ‘no’ for TIGIT expression, at least in this indication.

Regardless, to actually understand what blocking TIGIT does, we need to better understand the pathway.

That will be discussed in Part 2, coming soon.

Stay tuned.

The next great drug hunt, part 2: WHAT RECENT PAPERS TELLS US ABOUT TGF-beta-MEDIATED IMMUNOSUPPRESSION (and what they don’t)

Back in 2017 I put together a presentation I informally called “The Big Stick Talk” for a series of local immuno-oncology and investor conferences, including at one of my favorite venues, the Jefferies IO conference hosted by Birin Amin which is always a great event. You can find the slides here (IO combos). The premise of the presentation was that resistance to immunotherapeutics (anti-PD-1, anti-PD-L1, anti-CTLA4) was driven by pathways that controlled complex biologies – TGF-β, beta-catenin and VEGF – and that other targets (eg. TIM-3, LAG-3, IDO, etc) were secondary features. It followed that drug development targeting the secondary phenomena were likely doomed to failure – and here we are now, 4 years later, with our focus back on those big biological pathways.

Today: the TGF-β story revisited.

As noted in the prior post (link), there are three distinct isoforms of TGF-β, and all three signal through the TGF receptor complex. All isoforms are expressed in an inactive form, bound as prodomains to the latency-associated peptide (LAP). As discussed last time, one way that active TGF-β isoforms 1 and 3 can be released or exposed to the receptor complex is via the action of specific alpha-v integrins that bind to an amino acid consensus sequence (RGD) on LAP and induce a conformational change when the integrin is activated.

The latent TGF-β complex (TGF-β/LAP) becomes more complicated with the addition of proteins that can covalently bind to the latent complex. This even larger latent complex (LLC) comes in a number of different forms depending on the identity of the covalently bound protein. Two proteins (LTBP1 and LTBP3) are used when the LLC is bound to the extracellular matrix (ECM): imagine a cell secreting and placing a protein complex, just so, onto ECM (collagen, fibrinogen, GAGs, elastin et al). Of course the ECM is also created by cells. Two different proteins, GARP (LRRC32) and LRRC33, hold the LLC on the surface of specific immune cells: regulatory T cells and differentiated myeloid cells.

That’s just by way of introducing this complex biology.

Of the three isoforms, isoform 2 is most often cited as the cause of the toxicity seen when pan-TGF-β inhibitors were used clinically, and drug development efforts avoid this isoform. The new paper from the team at Scholar Rock (link to paper) focuses on a specific isoform, TGF-β1, thereby avoiding β-2 altogether but also not targeting β-3. The rationale is that β-1 is the critical isoform in most cancers and is the specific cause of TGF-β-mediated immunosuppression in those cancers. In an analysis of mRNA expression data from the TCGA database, the authors found that TGF-β1 was the predominant isoform in most human cancers and further that β1 expression correlated with a gene signature of resistance to checkpoint inhibition (called IPRES) in 7 different cancer types for which immune checkpoint inhibitors have been approved.

The authors present an antibody, SRK-181, that binds to the TGF-β/LAP complex and prevents TGF-β from becoming activated. Remarkably, the antibody was screened so that it would block alpha-v integrin mediated activation of the latent complex in all 4 LLC contexts (ie. complexed with LTBP1, LTBP3, GARP or LLRC33). To be clear, the antibody only binds the complex, not to free TGF-β1. The binding potency is very good, in the low pM range. The activity of the antibody was assessed in cellular assays. The cell-based assay is very clever and deserves some explanation. Using a LN229 glioblastoma cell line that endogenously express LTBP1 and LTBP3 they transfected in TGF-β1 or TGF-β3-encoding plasmids. This allows these cells to now make LLCs that can be deposited on ECM. In order to also make LLCs that can stay on the cell surface they then co-transfected in either GARP or LLRC33 expression plasmids. Now the cells could be used to make all 4 of the LLCs.

Now, here it gets fun. As detailed in the last post, αvβ8 is an integrin that can regulate the release of TGF-β from the latent state. In the prior paper (found here), it was demonstrated that when “sprung” from a GARP-containing LLC, the TGF-β was not released from the cell surface but remained tethered to the complex in such a way as to be able to activate TGF-β-receptors on a second cell (remember, that was a two-cell assay). So, LN229 cells express the αvβ8 integrin and can activate latent TGF-β1 complexes. Therefore once the LLCs are expressed, either into the extracellular matrix or on the cell surface, they can be activated by endogenous αvβ8 integrins expressed in cis (ie. on the same cell).

To measure the activity of released TGF-β1 the transfected LN229 cells were co-cultured with a reporter cell that expresses luciferase then the TGF-β-receptor complex is activated. In this assay the antibody inhibited TGF-β1 release from all four LLCs and blocked activation of the reporter cells that express TGF-β-receptors. An interesting question to ask here is whether the TGF-β1 was actually released and was in solution or remained tethered, and whether this was different for each of the LLCs made. A simple bilayer culture system in which the reporter cells are physically separated from the LN229 cells would answer this question.

Ok but that’s enough on the assay – the biology is also very interesting. I want to focus just on the tumor microenvironment findings, although the in vitro assays and the in vivo tumor growth data are also of interest. The key finding is that the combination of an anti-PD-1 antibody and the anti-TGF-β/LAP complex antibody improved tumor control by a mechanism that includes an increase of the tumor by CD8-positive T cells. In contrast to the data reported by the Genentech group ( this was not obviously due to redistribution of the T cells from an “immune exclusion” zone, set up by cancer-associated fibroblasts and attendant collagen matrix. Rather the effect was associated with a decrease in myeloid lineage cells and an influx of T cells from the vasculature. So, not surprisingly, there may be two different TGF-β-dependent biologies at work in the two tumor models used (MBT-2 tumors in this study and EMT-6 tumors in the Genentech study). One reasonable explanation is the difference in TGF-β isoform expression between the two models. MBT-2 tumors only express isoform β-1 whereas EMT-6 tumor express both isoforms 1 and 3. Somewhat confusing though is the finding reported here that in their hands, the Scholar Rock group found a similar result (decrease of myeloid suppression and influx of T cell across the vasculature) using EMT-6 tumors as they did using MBT-2 tumors (shown in the supplemental data). In the Genentech paper an impact on a myeloid cell signature was not detected while fibroblast genes associated with matrix remodeling were significantly reduced. Importantly, in the Genentech study a pan-anti-TGF-β antibody was used, that would be able to block both isoforms. Whether there is differential contribution of isoforms 1 and 3 to the immunosuppressive biology seen in these syngeneic models is not known. The authors do note that the MBT-2 and EMT-6 tumors used expressed specific LLCs, with LRRC33 and LTBP1 being the most highly expressed. However, whether these different LLCs contain different TGF-β isoforms in various cells types within EMT-6 tumors (eg. fibroblasts, myeloid cells and endothelium) is not described. Since GARP does not appear to be present in these tumor types, a role for TGF-β-1 expressed by Tregs is ruled out. Clearly there is more to learn from these models and their translatability to the human tumor setting.

There is emerging clinical evidence of the relevance of the TGF-β pathway to cancer therapy. Preliminary clinical data from studies using M7824, an anti-PD-1-TGF-βR2-TRAP protein, reported responses across several indications including a complete response in cervical cancer and partial responses in pancreatic and anal cancers (link) An expansion cohort study of patients with advanced NSCLC (n = 80) treated with M7824 in the second-line setting showed an objective response rate of 86% in the subgroup with high PD-L1 tumor expression (link 2). In an expansion cohort of 30 patients with pretreated advanced biliary tract carcinomas, M7824 monotherapy demonstrated a 23.3% response rate, with some durable responses (link 3). This TRAP protein is stated to be selective for the blockade of TGF-β1 and TGF-β3 isoforms and of course also blocks the PD-1 pathway. In the cell therapy space, encoded inhibition of TGF-β (eg. by using a dominant-negative TGF-β-R on the CAR T cell) is one of a variety of methods being explored to increase the efficacy of CAR T cells in solid tumor therapy. Our publication (here) provides much more discussion regarding strategies for successful solid tumor cell therapy.

In summary the regulation of TGF-β activity remains a fascinating area for drug development. If we see benefits in line with those achieved in some cancers by combining anti-PD-1 therapy with anti-VEGF therapy, we will indeed have picked up another big stick, and learned how to use it.

Stay tuned.

New Horizons Across the Immunotherapy Landscape – Lymphoid Structures Drive Immune Checkpoint Therapy and the Efficacy of Cellular Therapeutics

We’d been hearing the rumors for months. But the simultaneous publication in Nature of three papers describing a critical role for lymphoid structures and B cells in supporting T cell anti-tumor immunity was a remarkable milestone in our evolving understanding of immuno-oncology. Really stunning work. Importantly, these papers fit into a new contextual framework and cap a series of studies that have come out over the last year or so that have enriched our understanding of how the immune system and tumor cell populations interact. This broader and still evolving contextual framework will impact immunotherapy drug development across the immune checkpoint field, the tumor vaccine space, innate immune approaches, the T-cell-directed biologics, and cellular therapies.

But first, these new papers are gorgeous:

The study presented by Petitprez et al. is focused on the response of sarcomas to immunotherapy ( The soft tissue sarcomas (STS) have mixed clinical responses to immune checkpoint blockade (ICB) treatment, and it is not clear what drives the variable response. In general, STS have been classified as having a low tumor mutational burden (TMB) and are considered non-immunogenic, or ‘cold’, and have little expression of PD-L1. A few STS subtypes are characterized by more complex genetic abnormalities and could potentially have more actionable mutations for the immune system to recognize. Regardless, two of most widely used biomarkers of ICB response (TMB-high or PD-L1-positive) are not generally relevant in STS. In this study, gene expression profiling was used to examine patterns of ICB response in patients across a wide variety of STS subtypes and pathologies. Three distinct genetic classes were identified that match known tumor microenvironments (TME) – immune desert (A), highly vascular (C), and inflamed (E) with two intermediates: B and D. These are well understood classifications and mirror many prior studies of the TME and ICB response and resistance. However, several details that emerged are critical – 1) the inflamed TME (E) and the intermediate form (D) were not associated with any particular STS classification, but were distributed across STS histologies, and 2) the E/D inflamed signature was characterized by a pronounced B cell signature, and by expression of the chemokine CXCL13. These results suggest secondary lymphoid organ development and organization.

A sidebar here: secondary lymphoid organs include spleen, lymph nodes, and Peyers patches and are characterized by critical structural features that include a T cell zone and adjacent B cell follicles that orchestrate coordinated immune responses, for example, to pathogens. Localized lymphoid organs can form in chronically inflamed tissues – these are the tertiary lymphoid structures (TLS) and are classically associated with prolonged inflammation and autoimmunity. The organization of cell types in these structures is controlled in part by chemokines, including CXCL13 and CXCR5.

Using immunohistochemical staining, the authors went searching for TLS in tumor sections, and, as suggested by the gene transcript data, TLS were found in groups E and a bit in D. The presence of TLS in tumors has been noted before, but here the presence of TLS, and of B cells, was associated with patient overall survival. Further in a small cohort of patients (n=47), classes E and D, those most likely to have TLS, responded best to ICB therapy. These observations suggest that B cells and TLS are important for successful ICB therapy.

OK, that’s one study, in an indication in which ICB therapy in general has not worked well. So, are these observations generalizable?

The next paper looked at these features in an indication that is very different from STS. The paper by Cabrita et al. focused on ICB response in melanoma ( Melanoma is notable for several features, having a very high TMB and being among the most ICB responsive indications. Indeed, in terms of immune-responsive tumor types, melanoma and STS are on the opposite ends of the spectrum. Nonetheless, the analysis of melanoma responsiveness to ICB yielded results strikingly similar to the results of the STS analysis.

Immunofluorescent staining was used to identify T and B cell clusters, and these were associated with the chemokines CXCL13 and CXCR5. The co-occurrence of T and B cells and the identification of a TLS gene signature predicted clinical response to ICB. Mechanistically, tumors that featured TLS and were rich in B cells also had an increased population of naive and/or memory T cells while those tumors without these features had an increased population of exhausted T cells. Notably, the T cell population enriched in the presence of TLS was CD8-positive – the subset associated with cytotoxic anti-tumor immunity. Whether the B cells themselves were also contributing directly to the anti-tumor response via production of anti-tumor antibodies appears less clear. As in the STS study, TMB was not correlated with TLS formation in melanoma.

The final study by Helmink et al. analyzed patients who had enrolled in a phase 2 clinical trial of neoadjuvant ICB therapy for high-risk resectable melanoma ( In the neoadjuvant setting, ICB therapy is given prior to the surgery that is performed to remove tumors. Gene expression and immune-staining analyses were used to assess TLS and B cell presence in tumors, similar to the prior studies. Of note the results were compared to an analysis of gene signatures in a neoadjuvant trial of ICB treatment of metastatic renal cell carcinoma (RCC). As in the other two studies the presence of B cells and TLS in the tumor prior to therapy was predictive of response to ICB. This was found in patients with metastatic melanoma and in patients with metastatic RCC. In addition, the authors showed that the B cell pool was differentiated, into memory B cells and plasma cells, suggesting that the TLS was productively driving both T and B cell differentiation and perhaps indicating a role for B cell adaptive memory in supporting the anti-tumor immune response.

In these three papers covering three indications (STS, melanoma, RCC), a signature of TLS enriched in B cells was associated with ICB responses and T cell activity, notably of CD8-positive T cells. A useful model for these findings is that organized lymphoid structures like TLS provide an environment in which tumor antigen can be productively displayed to the adaptive immune system, ie. T and B cell mediated immunity. In this setting B cells may have such a strong predictive signature for several reasons: B cells are antigen-presenting cells capable of supporting a persistent T cell response by restimulating memory T cells and by triggering de novo naive T cell responses to tumor antigens. Further, B cells engage in the productive costimulation of T cells via B7/CD28, CD40L/CD40 and adhesion molecule interactions and also produce cytokines that provide activation, proliferation and survival signals to T cells (eg. TNF, IL-2, IL-6, IFNy).

Additional productive areas of investigation would include analyses of dendritic cell populations and, of critical importance I think, the status of tumor-draining lymph nodes                       (see

These studies raise some interesting questions. For example, in the context of anti-PD-(L)-1 therapeutics, what is the pattern of PD-L1 expression that accounts for the anti-tumor response? In the current paradigm, PD-L1 expression on the tumor itself is considered the critical target. This has recently been complicated by the finding that PD-L1 expression on myeloid cells in the TME may also be relevant (eg. And we must recall that PD-1/PD-L1 interaction in lymphoid organ germinal centers regulates T cell / B cell interactions (
It may be that ICB treatment is influencing the anti-tumor immune response in multiple ways.

We’ve seen many novel immunotherapy agents falter and some of those results may reflect this more complex immune-tumor landscape. This leads one to wonder where and how novel agents might function and if they are actually beneficial to anti-tumor immunity. As one example, many PD-L1-based bispecific antibodies are under development. We might hypothesize that an anti-PD-L1/anti-4-1BB bispecific antibody would actually have multiple sites of action, not only in the tumor TME, but also in the TLS and draining lymph nodes. As another example, T cell engagers (BiTEs, DARTs et al), a design that works well in the hematologic cancers, may not work as well in solid tumors unless those tumors have TLS already. Finally, it’s a bit baffling that TMB does not correlate with TLS formation, and we might wonder how this result reflects upon efforts in the neoantigen space. And so on, as we think about tumor vaccines, and innate immune stimulation, and novel checkpoints, etc.

This brings us also to cell therapy and potential lessons for that field, specifically in the solid tumor setting. We have long recognized that CAR T cells that target CD19 (CD19-CARs, eg. Kymriah from Novartis and Yescarta from Kite/Gilead) have dramatically better persistence properties than CARs that target solid tumor antigens. We have hypothesized that the self-renewing nature of the antigen itself is an important aspect of persistence – CD19 is expressed on B cells that are continuously produced by the bone marrow.

In light of these new findings on ICB responses I think we can consider a second feature – the immunological relevance of antigen presentation to the CAR T cell. We have recently seen a rash of efforts to provide artificial and immunologically favorable antigen presentation to CAR T cells. The goal here is to improve CAR T cell fitness by providing the proper immunological niche (eg. a lymph node) and driving persistence. In one example an artificial CAR T ligand was developed that could be injected into a patient who is receiving a CAR T therapy ( This artificial CAR-T ligand binds to serum albumin, traffics into lymph nodes and is taken up and displayed by antigen-presenting cells. CAR-T cells trafficking through the lymph node are stimulated both by the displayed antigen and by costimulatory receptors and cytokines, much like the system envisioned in the ICB response/TLS papers described above. Similarly, a nanoparticulate RNA vaccine was used to deliver and express an artificial CAR antigen into a tumor-draining lymph node ( Again, this is designed to promote immunologically productive presentation of the target antigen to activate and expand the injected CAR-T cells. Notably, engagement of relevant stimulatory, chemokine and adhesion signals are known to favor the development of T cell memory, a critical element in long term immune protection.

Our in-house technology ( uses CD19-targeting by CAR-T cells to leverage persistence and also to take advantage of immunologically relevant antigen presentation. We do this by building CAR T cells to CD19 that simultaneously can target any antigen of interest. We enabled this ‘repurposing’ of CD19-CAR T cells by creating small, highly potent proteins that bridge the CD19-CAR T cell to the tumor antigens we choose, triggering T cell cytotoxicity and killing the tumor cell ( Every bridging protein contains the CD19 protein, so is the target of any CD19-CAR T cell. This bridging protein strategy enables facile multi-antigen targeting because the design is highly modular. We have built CD19-CAR T cells that secrete bridging proteins that recognize CD20, Clec12a, Her2, EGFR and many other different antigens and in some instances multiple antigens simultaneously. Because the CAR T cell is a CAR to CD19, this is a simple, pragmatic and universal solution. And because the bridging protein is secreted by the CD19-CAR T cells themselves, this becomes a simple matter of encoding everything into a lentiviral vector.

CD19-CAR T cell interaction with normal B cells will support production of immunologically relevant stimulatory signals including adhesion interactions, chemokine and cytokine signals, and costimulatory signals, even if the consequence for the B cell is cytotoxic. This organic presentation of immunologically relevant antigen does not require administration of additional agents or exogenous antigen, since B cells are themselves antigen presenting cells, are present in lymphoid organs and in circulation, and represent a self-renewing source of CD19 due to production by the bone marrow.

In the context of solid tumor treatment creating an expanded and persisting CAR T cell pool using CD19-positive B cells as a non-tumor dependent source of antigen becomes very attractive. Our lead solid tumor program uses a CD19-anti-Her2 bridging protein, where the CD19 portion is the CD19 extracellular domain, and the anti-Her2 portion is an anti-Her2 scFv. This bridging protein is encoded into a lentiviral vector downstream of a CD19-CAR sequence, separated by a P2A cleavage site. Thus we have made a ‘Her2-bridging CD19-CAR’ that can bind directly to CD19 on B cells via the CAR domain, and to CD19 painted onto a Her2-positive solid tumor cell, via the bridging protein.

Our lead indication is Her2-positive metastatic breast cancer and metastatic lung cancer, specifically in patients who have relapsed from standard of care therapy by developing CNS metastases. The reasoning is simple: the Her2-bridging CD19-CAR T cells can be injected systemically to become activated by CD19 expressed on normal B cells (and B cell aplasia is a manageable toxicity). The activated CARs will traffic systemically, all the while secreting the small (and short half-life) bridging proteins. Patients can remain on standard of care, including with anti-Her2 antibodies, because we will not need to ‘see’ Her2 in the periphery in order to trigger CAR activation and expansion. And of course, anti-Her2 antibodies like Herceptin can’t cross the blood brain barrier and get into the CNS at all. Once activated CD19-CAR T cells will enter the CNS, as activated T cells are known to do, and the secreted bridging proteins will coat Her2-positive tumor cells with CD19, allowing the CD19-CAR T cells to the attack and kill those cells. Those CAR T cells can also leave, recirculate, become stimulated again by encounter with B cells, and return to the CNS – a trick no direct CAR to Her2 can hope to duplicate. Not to oversell it, but I love this program, and it should work.

The idea that we can repurpose and send off a CD19-CAR T cell to attack any antigen and indication is compelling in its simplicity and modularity. We have already built programs for treating AML and B cell tumors. Our next wave of programs takes advantage of our ability to weave together bridging proteins with two and three antigen-binding domains – and this allows us to contemplate attacking very heterogeneous tumor types with a simple CAR – the CD19-CAR – that has exemplary persistence and fitness characteristics.

Stay tuned.

Radical optimism: considering the future of immunotherapy

I wrote recently about the sense of angst taking hold in the next-generation class of immuno-therapeutics – those targets that have come after the anti-CTLA4 and anti-PD-(L)-1 classes, and raised the hope that combination immunotherapy would broadly raise response rates and durability of response across cancer indications.

There are diverse next-generation immuno-therapeutics including those that target T cells, myeloid cells, the tumor stromal cells, innate immune cells and so on. A few examples are given here (and note that only a few programs are listed for each target):

Screen Shot 2018-11-05 at 8.31.33 PM

There are of course many other therapeutic targets – OX40/CD134, Glutaminase, ICOS, TIM-3, LAG-3, TIGIT, RIG-1, the TLRs, various cytokines, NK cell targets, etc.

In the last year – since SITC 2017 – there has been a constant stream of negative results in the next generation immuno-therapy space, with few exceptions. Indeed, each program listed in the table has stumbled in the clinic, with either limited efficacy or no efficacy in the monotherapy setting or the combination therapy setting, typically with an anti-PD-(L)-1 (ie. an anti-PD-1 or an anti-PD-L-1  antibody). This is puzzling since preclinical modeling data (in mouse models and with human cell assays) and in some cases, translation medicine data (eg. target association with incidence, mortality, or clinical response to therapy), suggest that all of these targets should add value to cancer treatment, especially in the combination setting. I’ve discussed the limitations of these types of data sets here, nonetheless the lack of success to date has been startling.

With SITC 2018 coming up in a few days (link) I think it is a good time to step back and ask: “what are we missing?”

One interesting answer comes from the rapidly emerging and evolving view of tumor microenvironments (TME), and the complexity of those microenvironments across cancer indications, within cancer indications and even within individual patient tumors. TME complexity has many layers, starting with the underlying oncogenic drivers of specific tumor types, and the impact of those drivers on tumor immunosuppression. Examples include activation of the Wnt-beta catenin pathway and MYC gain of function mutations, which mediate one form of immune exclusion from the tumor (see below), and T cell immunosuppression, respectively (review). In indications where both pathways can be operative (either together or independently, eg. colorectal cancer, melanoma and many others) it is reasonable to hypothesize that different strategies would be needed for combination immuno-therapy to succeed, thereby producing clinical responses above anti-CTLA4 or anti-PD-(L)-1 antibody treatment alone.

A second and perhaps independent layer of complexity is TME geography, which has been roughly captured by the terms immune infiltrated, immune excluded, and immune desert (review). These TME types are illustrated simply here:

Screen Shot 2018-11-05 at 8.45.19 PM

The different states would appear to be distinct and self-explanatory: there are immune cells in the tumor (infiltrated), or they are pushed to the periphery (excluded), or they are absent (desert). The latter two states are often referred to as “cold” as opposed to the “hot” infiltrated state. It is common now to propose as a therapeutic strategy “turning cold tumors hot”. The problem is that these illustrated states are necessary over-simplifications. Thus, immune infiltration might suggest responsiveness to immune checkpoint therapy with anti-PD-(L)-1 antibodies, and indeed, one biomarker of tumor responsiveness is the presence of CD8+ T cells in the tumor. But in reality, many tumors are infiltrated with T cells that fail to respond to immune checkpoint therapy at all. The immune excluded phenotype, alluded to above with reference to the Wnt-beta catenin pathway, can be driven instead by TGF beta signaling, or other pathways. The immune desert may exist because of active immune exclusion, lack of immune stimulation (eg. MHC-negative tumors) or because of physical barriers to immune infiltration. Therefore, all three states represent diverse biologies within and across tumor types. Further, individual tumors have different immune states in different parts of the tumor, and different tumors within the patient can also have diverse phenotypes.

There are yet other layers of complexity: in the way tumors respond to immune checkpoint therapy (the “resistance” pathways, see below), the degree to which immune cells responding to the tumor cells are “hardwired” (via epigenetic modification), the metabolic composition of the TME, and so on. Simply put, our understanding remains limited. The effect of this limited understanding is evident: if we challenge tumors with a large enough immune attack we can measure a clinical impact – this is what has been achieved, for example, with the anti-PD-(L)-1 class of therapeutics. With a lesser immune attack we can see immune correlates of response (so something happened in the patient that we can measure as a biomarker) but the clinical impact is less. This is what has happened with nearly all next-generation immuno-therapeutics. As a side note, unless biomarker driven strategies are wedded to a deep understanding of specific tumor responsiveness to the therapeutic they can be red herrings – one example may be ICOS expression, although more work is needed there. Understanding specific tumor responsiveness is critical regardless of biomarker use, due to the layered complexity of each indication, and even each patient’s tumors within a given indication.

So why should we be optimistic?

I propose that some of the next generation immuno-therapeutics will have their day, and soon, due to several key drivers: first, for some of these classes, improved drugs are moving through preclinical and early clinical pipelines (eg. A2AR, STING). Second, the massive amount of effort being directed toward understanding the immune status of diverse tumors ought to allow more specific targeting of next generation immuno-therapeutics to more responsive tumor types. The TGF beta signature presents a particularly interesting example. Genentech researchers recently published signatures of response and resistance to atezolizumab (anti-PD-L1) in bladder cancer (link). In bladder cancer about 50% of tumors have an excluded phenotype, and about 25% each have an immune infiltrated or immune desert phenotype. The response rate to treatment with atezolizumab was 23% with a complete response rate of 9% (note that responses did not correlate with PD-L1 expression but did correlate with both tumor mutational burden and a CD8+ T cell signature). Non-responding patients were analyzed for putative resistance pathways. One clear signature of resistance emerged – the TGF beta pathway, but only in those patients whose tumor showed the immune excluded phenotype. The pathway signature was associated with fibroblasts, but not myeloid cells, in multiple tumor types. The T cells were trapped by collagen fibrils produced by the fibroblasts:

Screen Shot 2018-11-05 at 9.04.16 PM

(The image is a screenshot from Dr Turley’s talk at CICON18 last month).

It follows that a combination of a TGF beta inhibitor and a PD-(L)-1 inhibitor for the treatment of bladder and perhaps other cancers should be used in patients whose tumors show the immune excluded TME phenotype, and perhaps also show a fibroblast signature in that exclusion zone. Indeed, in a recent paper, gene expression profiling of melanoma patients was used to demonstrate that a CD8-related gene signature could predict response to immuno-therapy – but only if the TGF beta signature was low (link).

There are other immunotherapy resistance pathways – some we know and some are yet to be discovered. We should eventually be able in future to pair specific pathway targeting drugs to tumors whose profile includes that pathway’s signature – this has been done, retrospectively, with VEGF inhibitors and anti-PD-(L)-1 therapeutics. This will require a more comprehensive analysis of biopsy tissue beyond CD8+ T cell count and PD-1 or PD-L1 expression – perhaps immunohistochemistry and gene transcript profiling – but these are relatively simple technologies to develop, and adaptable for a hospital clinical lab settings. Not every next generation immuno-therapeutic will succeed as the clinical prosecution becomes more targeted, but some certainly will (we might remain hopeful about adenosine pathway inhibitors, STING agonists, and oncolytic virus therapeutics, to name a few examples).

Another driver of success will be cross-talk with other technologies within immuno-oncology – notably cell therapy (eg. CAR-T) and oncolytic virus technologies. We have already seen the successful adaptation of cytokines, 4-1BB signaling, OX40 signaling and other T cell stimulation pathways into CAR T cell designs, and the nascent use of PD-1 and TGF beta signaling domains in cell therapy strategies designed to thwart immuno-suppression (we should note here that CAR T cells, like tumor infiltrating T cells, will face  barriers to activity in different tumor indications). The example of local (and potentially safer) cytokine secretion by engineered CAR-T cells has helped drive the enormous interest in localized cytokine technologies. Most recently, the combination of CAR-T, oncolytic virus and immune checkpoint therapy has shown remarkable preclinical activity.

SITC 2018 – #SITC18 on Twitter – will feature sessions on  immunotherapy resistance and response, the tumor microenvironment, novel cytokines and other therapeutics, cell-based therapies, and lessons from immuno-oncology trials (often, what went wrong). We can expect lots of new information, much of it now focused on understanding how better to deploy the many next generation immuno-therapeutics that have been developed.

So, I would argue that “radical optimism” for next generation immunotherapy and immunotherapy combinations is warranted, despite a year or more of clinical setbacks. Much of the underlying science is sound and it is targeted clinical translation that is often lagging behind. Progress will have to come from sophisticated exploratory endpoint analysis (who responded, and why), sophisticated clinical trial inclusion criteria (who to enroll, and why) and eventually, personalized therapeutic application at the level of the indication and eventually the patient.

In the meantime, stay tuned.

Next gen IO: what we thought we knew

It’s mid-July and blazing hot here in Massachusetts. Luckily, we’re entering the summer vacation season and a break from the seemingly endless stream of biotech and oncology conferences that began in earnest last September, culminating in June with ASCO and EHA. We may also see a pause in the rush of biotech IPOs. There is always an interesting dynamic at play between progress as reported at medical conferences and the attractiveness of stock offerings – and this year the energy firing that dynamic is a bit unusual, especially across the immuno-oncology (IO) landscape.

As has been widely reported, ASCO was disappointing for next generation IO players. ‘Next generation’ refers to those companies developing assets that target diverse and novel immune regulatory targets, beyond anti-CTLA4 and anti-PD-(L)-1 antibodies. Essentially all companies bringing forward IPOs in the IO space have next-gen aspirations.

Why was ASCO disappointing? In part the answer is obvious: numerous expert reviews (eg. ours, from 2015: had promoted the compelling story that IO combinations would further improve the treatment of ever more patients in even more cancer indications. Such hypotheses drove intense investment in biotech companies, and the development of novel drugs targeting the diverse pathways of interest. During 2017 and 2018, IO combo hypotheses began reading out in clinical trial data, and the early results were underwhelming. This data wave culminated at ASCO in June.

The examples have by now been widely discussed: the collapse of the IDO inhibitor class with failures in late stage studies, the early defeat of an agonist anti-ICOS antibody, the realization that none of the many agonist antibodies to TNF superfamily receptors (4-1BB, OX40, GITR, CD27, etc.) were going to be quick wins, the modest activity of anti-CSF1R antibody, a miss from the VISTA program, and so on.

But the data were disappointing in a distinctly different way that impacts all programs, whether in trials or not. Simply put, we are due for a reality check regarding the quality of evidence used to support novel IO combination hypotheses. There are three (at least) aspects to consider:

1) IO animal models have limited utility. I think we all want to see better in vivo models of IO efficacy, and not just the same old PD-1 or PD-L1 combos in the usual mouse models. The usual mouse models are those that are partially responsive to anti-CTLA4, anti-PD-1 or anti-PD-L1, eg. tumors that grow out from the cell lines MC38, CT26, B16 and a few others when they are implanted into a mouse subcutaneously. Figure 1 is an example of the type of results we often see.

Figure 1. Activity of an IDO inhibitor in combination with anti-PD-1.


A few things to note are the modest IDO monotherapy activity (diamonds), and the modest improvement over anti-PD-L1 alone when the combination is given (circles), the fact that this is only a delay in tumor growth kinetics (by a few days at best) and a small survival advantage (2/15 mice in the combo arm survived). This study was conducted by top tier investigators with superb reputations (Spranger, Gajewsky and colleagues, the paper is at doi: 10.1186/2051-1426-2-3) – this was the best result they could get, and it is considered a positive outcome. This is the reality of traditional IO combo studies in mice: the fact is that these models have very limited value for predicting success in the clinical trial setting. The alternatives – PDX models and perhaps 3D-culture models – can be difficult to source and expensive to run.

By the way, the IDO inhibitor story may not be dead yet, as post-hoc analyses suggest that the leading programs (from NewLink and Incyte) had sub-optimal exposure and saturation of target. Also, novel compounds with a different mode of action (heme-displacement) are in development. So, we’ll see.

2) Correlation of target expression with survival data has limited value. This is an interesting area for dissection. The general argument put forward is that expression of some target of interest is associated with better or worse outcome in one or more cancer indications. The word ‘associated’ gives away the issue, as we are talking here about the correlation of expression data and patient outcome. The CSF1R story was born in such data – first the observation that myeloid cells, notable TAMS and MDSCs, were associated with poor prognosis in a large number of cancers, then the observation that CSF1R expression was also correlated with outcome, as was expression of the ligand for CSF1R, M-CSF. Here is an example of these kind of data, in this case from lung cancer patients, published earlier this year (doi:10.1038/s41598-017-18796-8):

Figure 2. CSF1R levels and overall survival in lung cancer


If you read the methods for this paper – and I’m just using this as an example, not to pick on this particular study – what you’ll see right away is that the measurement of analytes (M-CSF and CSF1R) was performed using a single biopsy or surgical resection sample. There are no longitudinal data (successive samples taken during treatment). As we learned in the old days of transcript profiling of autoimmune disease, longitudinal data are essential – the Type 1 interferon story in lupus being a poster child of the pitfalls encountered in non-longitudinal analysis.

The myeloid cell data are even more interesting – here is an example looking at myeloid clusters in NSCLC patients (smokers) and correlation with survival (from doi: 10.1371/journal.pone.0065121):

Figure 3. Myeloid cell clusters and survival

myeloid cell

There are a lot of these types of studies in many different cancer indications. So now you have a pretty clear narrative: CSF1R+ myeloid cells present in tumors are associated with poor outcome for those patients. That’s the observation. The hypotheses that flow from the observation are various: anti-tumor efficacy will result from depleting tumor-resident myeloid cells, or from interfering with this myeloid cell accumulation, or from changing the phenotype of the myeloid cell population in the tumor. The first hypothesis was tested with the anti-CSF1R antibody cabiralizumab which potently depletes all myeloid cells across a range of tumor indications but has no to limited efficacy either as monotherapy or in IO combination therapy. The second and third hypotheses were tested in glioma using a CSF1R inhibitor, PLX3397, thought to prevent myeloid cell accumulation and polarize myeloid cells to a pro-inflammatory phenotype. This drug did not show efficacy (doi: 10.1093/neuonc/nov245).

As with the IDO story, the CSF1R story is not yet fully told, and the early data in at least one tumor type (pancreatic cancer) was interesting. But the near-term conclusion is that the foundational observation (myeloid cells in tumors are associated with poor prognosis) did not produce successful hypotheses regarding CSF1R. Why not? The answer again lies in the way these analyses are done – using single point surgical resection or biopsy tissue to set the level of the analyte. This is not a criticism: there are no other readily available samples, unless the disease continues to disseminate (ie. there are metastases). But here’s the rub: metastases are themselves an indicator of advancing disease and cannot be used as an independent measure. And this raises the larger point: when the variable is survival, many parameters will correlate with outcome, and it’s very hard indeed to distinguish cause from effect – which bring us to biomarkers.

3) Good IO biomarkers are elusive. Anyone following IO has seen the confusion regarding PD-L1 expression as a biomarker of response to anti-PD-1 and anti-PD-L1 antibody therapies. Enough said. But is PD-L1 expression a uniquely complex example of biomarker use? Or is it representative of the level of background noise seen in the complex immune response to cancer that IO therapies attempt to exploit? I favor the latter assessment. A quick definition of ‘biomarker’ as used here: measurement of an analyte that is hypothesized to be associated with response to therapy. A biomarker can be used as inclusion criteria in clinical trial enrollment, or as an exploratory endpoint to assess clinical responses. The depletion of myeloid cells in the anti-CSF1R trial mentioned above is one example. A textbook attempt at the use of a biomarker in a novel IO combo trial is seen in the early development of JTX-2011, the agonist anti-ICOS antibody. The indications examined in the ICONIC clinical trial were chosen based on ICOS mRNA expression. Expression on infiltrating T cells in tumors was analyzed across ~30 cancer indications and confirmed using immunohistochemistry. Based on the frequency of high ICOS expression, ICONIC enrolled patients with lung cancer (NSCLC), Head & Neck cancer, triple negative breast cancer, gastric cancer, and melanoma. Of note, most of these indications are known to be responsive to immune checkpoint therapy with the anti-PD-(L)-1 class of antibodies, and some respond to the anti-CTLA4 classes of antibodies. In “ICOS-high” indications the rationale for increasing the overall response rate to therapy by adding an agonist anti-ICOS antibody to an immune checkpoint blocking antibody (anti-PD-1 or anti-CTLA4) looked pretty straightforward. The ICONIC trial was presented at ASCO to near universal disappointment as the response data were not higher than immunotherapy alone. As to the biomarker data, that becomes hard to judge in the absence of a clear efficacy signal, but Jounce made the best of it, suggesting that ICOS expression “appears to associate with anti-tumor activity” which is a pretty soft statement. As with IDO and CSF1R, the anti-ICOS story is not done, and more clinical data will follow. However, the clinical hypothesis that anti-ICOS combination immunotherapy with anti-PD(L)-1 or anti-CTLA4 therapies would increase patient response rates was not supported in this early study.

So what to make of all this? The response to the negative news flow has been predictable, with some of these companies losing significant share value. Several large pharmaceutical companies have taken the opportunity to declare a sharpened focus: Novartis’ Jay Bradner, for example, stated that the company will only develop new therapeutics that demonstrate monotherapy efficacy preclinically (back to those mouse models, but it’s a step), and Astra Zeneca’s Pascal Soriot has declared the company will target patients earlier in disease.

It is interesting to consider the sheer number of ongoing combination trials, a number that is tracked in real time by Beacon Intelligence ( As of last week the numbers looked like this:

Figure 4. IO combination clinical trials

beacon IO

It’s a huge number of ongoing trials. There are lots of interesting therapeutics to read out in the near term, including such classes as STING agonists, the adenosine pathway inhibitors (A2AR, A2AR/A2BR, CD39 and CD73), anti-CD47 antibodies, arginase inhibitors and on and on. Given the sheer number of trials it is timely to wonder if we are just blindly pushing ahead. It is certain that we are not learning all of the available lessons – one only has to look at the (investor) excitement generated by data from very early trials of cytokines (IL-2, IL-10) and TLR agonists. Some of these agents will certainly fail in later stage clinical trials.

The brings us back to where we started and the notion that there is an odd energy dynamic between the ongoing reporting of clinical trial results and the march of immuno-oncology IPOs to market. Given the recent spate of negative news, why do IO IPOs continue to draw investor interest?

Let’s take two examples from the recent IPO pool.

Scholar Rock is a company with a completely novel asset platform targeting the TGF beta family of proteins – this IPO was very well received even though the company is early in clinical execution, especially in IO (the lead program is for muscle wasting diseases). TGF beta is emerging as a critical biologic signal of resistance to immunotherapy. Right upfront it is important to note that the caveats in assigning causality to resistance pathways are similar to assigning causality to association with survival – these are findings based on single source pretreatment cancer biopsy or resection samples, and the patient outcome is followed over time (but the analytes are not typically followed longitudinally). But TGF beta is a different kind of beast, notable for its’ roles in immunosuppression and wound healing, ie. resolving inflammation in support of tissue remodeling. The basic biology lends itself to exploitation by cancer cells as they build an immunosuppressive microenvironment and remodel that environment, often leading to aggressive growth and metastasis. TGF beta activity also triggers TGF beta expression in a feed-forward loop. Surface Oncology, who came to market with an anti-CD47 antibody in Phase 1 (neither novel nor first) and an anti-CD73 antibody program (again neither novel nor first) was also well received – in large measure because their anti-IL-27 program is highly novel and associated with signaling through the PBAF complex – which has in turn been linked to the efficacy of T cell mediated anti-tumor immunity in studies from diverse labs.

Companies like these bring forward a new wave of therapeutic hypotheses to test, and that is attractive in an environment where the old wave of therapeutic hypotheses appear to be stumbling. Of course, the investment hypothesis, at least in the long term, is that they will succeed.

Two more comments: 1) re-examining and critiquing lines of evidence in IO drug development should be an iterative process, and 2) this is a process that can be applied to other therapeutic modalities – oncolytic viral therapy development and cellular therapy for solid tumors, to name just two.

stay tuned.

Novel fibrosis therapeutics: walking down the TGF-beta pathway

Recent weeks brought startling news of clinical trial successes in the treatment of Idiopathic Pulmonary Fibrosis (IPF). The clinical and commercial consequences have been heavily reviewed elsewhere (eg. GLPG and FGEN).

This short commentary will focus on the underpinning science, with particular reference to the TGF-beta (TGFb) pathway and the role of that pathway in fibrotic disease.

First a quick primer on the recent advances in IPF drug development. Fibrogen pulled off a successful and surprising 48 week Phase 2 trial of pamrevlumab, an old antibody targeting CTGF, while Galapogos followed with very compelling early Phase 2a data of GLPG1690, an autotaxin inhibitor, including the apparent reversal in decline of lung function that is the hallmark of IPF (nb. small sample size, short analysis period (12 weeks). These two new drugs are poised to contribute to the next wave of IPF therapeutics, joining pirfenidone (Esbriettm, Roche) and nintedanib (Ofevtm, Boehringer Ingelheim), approved for the treatment of IPF in 2014. Of note, pirfenidone and nintedanib are considered moderately efficacious, slowing but not reversing the rate of decline in lung function, and only modestly improving life expectancy, if at all. Therefore, if pamrevlumab or GLPG1690 can differentiate by either reversing lung damage or increasing life expectancy, they would be expected to overtake the earlier drugs.

What’s interesting is that the mechanism of each of these drugs intersects with the activity of TGFb, a dominant cytokine that normally controls wound healing and other tissue repair activities. When dysregulated, TGFb becomes pathogenic, supporting disease processes spanning oncology and fibrosis. We can visualize the initiation and progression of fibrosis as a series of steps controlled, at least in part, by the continuous activity of TGFb signaling through the TGFb-receptors. Indeed, pirfenidone’s mechanism of action includes inhibition of TGFb-receptor signaling, among other activities (both pirfenidone and nintedanib are tyrosine kinase receptor inhibitors, and neither is particularly selective, hitting multiple receptors simultaneously).

Here is a cartoon with different stages of fibrosis induction and progression, simplified:

 cascade 1

Without dwelling overly on the various pathologies at work, we can point to several critical steps underpinning the cascade:

1) influx of inflammatory cells upon injury

2) increase in autotaxin and therefore LPA (autotaxin cleaves the abundant lipid moiety LPC, to release LPA)

3) triggering of the LPA receptors, that, among other activities, are responsible for the activation of beta integrins, leading to the release of TGFb from sequestration

4) activation of TGFb receptors, that, among other activities, induce the secretion of growth factors including CTGF

5) induced production of TGFb by the action of CTGF

6) myofibroblast activation and ECM deposition leading to fibrosis

Now we can overlay some of the therapeutics developed for IPF on the cascade (simplified even further here):

 cascade 2

Note that Fibrogen’s anti-CTGF antibody pamrevlumab sits along side the approved drugs pirfenidone and nintedanib, a bit downstream of the initiation of the fibrotic cascade. The clinical data to date suggest that pamrevlumab has activity similar to pirfenidone and nintedanib. It is critical to stress here that CTGF secretion is tightly controlled by, among other things, TGFb. Further, in the fibrotic setting, TGFb appears to be a master regulator of CTGF expression, and, as noted above, the regulation is feed forward, since CTGF signaling through it’s receptor induces expression of more TGFb. It makes sense then that an anti-CTGF antibody could derail this chronic signaling loop, and thereby provide therapeutic benefit by reducing TGFb levels as well as tempering the pathologic activity of CTGF.

Further downstream is the anti-LOXL2 antibody, simtuzumab, formally under development at Gilead for IPF and NASH (a form of liver fibrosis). LOXL2 enzyme is responsible for crosslinking collagen, thus contributing to ECM deposition. Perhaps being too far down the fibrotic cascade is ineffective, as simtuzumab was dropped based on poor efficacy in Phase 2 trials.

Turning upstream we find autotaxin inhibition, integrin inhibition and LPA inhibition. The mechanism of action of all three targets is related, since, as noted above, autotaxin cleaves LPC to yield LPA; LPA in turns binds the LPA receptors. LPA receptors are G-protein coupled receptors with diverse functions, prominent among them being the activation of integrin beta strands. Among the integrins then, integrin avb6 has been tied directly to TGFb release in the setting of lung fibrosis. This complex system is tightly regulated, with the relationship of autotaxin/LPA/LPA receptor/integrin/TGFb preserved in different organ systems. The relative contribution of the different LPA receptors and different integrins does vary however.

The three upstream therapeutics illustrated have presented a mixed clinical picture. The anti-LPA antibody Lpathomab was a clinical flop, and the company developing it, Lpath Inc., reversed merged into oblivion several years ago. Anti-LPA might have worked if the antibody were optimal, however the suspicion at the time was that the antibody bound to only some of the myriad isoforms of LPA, and was therefore ineffective. Biogen’s anti-alpha-v beta-6 antibody STX-100 has finished a Phase 2 trial in IPF but has not reported data. This may be due to issues of efficacy and/or safety or may be due to Biogen’s frank disinterest in fibrosis (or anything outside of neurology). However, we simply don’t know. And finally, we have the new data on the Galapogos autotaxin inhibitor, GLPG1690.

The results obtained in the small Phase 2 trial of GLPG1690 in IPF were very good, and if they hold up the drug could certainly be best-in-class for IPF. Further, as a therapeutic sitting at the top of the fibrotic cascade, there is every reason to believe that this drug will show promising efficacy in other fibrotic conditions.

We expect updates from these programs and others at the fall academic meetings, starting with the European Respiratory Society Congress on 9-13 September. Importantly, Fibrogen will report data from combination studies of pamrevlumab plus pirfenidone or nintedanib in IPF.

 stay tuned.

A quick look at the axitinib data and IO/VEGF inhibitor combos

Yes I know, there has been a lot of talk of immunotherapy combination trials (and tribulations). But in reviewing #ASCO17 slides I stumbled on some interesting results. (These are screen grabs actually and I don’t have a source for all the photos – my apologies to the folks that posted these pics!)

What looked interesting at ASCO? Lots of cool stories emerged, but I think this one was overlooked:

This slide grabbed my attention- this is a summary of data on PD-(L)-1 inhibitors combined with targeted inhibitors in advanced renal cell carcinoma (RCC). I’ve tagged some of the drugs just below the slide:

Screen Shot 2017-06-09 at 10.02.56 AM

The studies shown in the graph can be a little misleading, as the monotherapy studies are in the second-line or later setting, while the combination data are in the first line (treatment-naive) setting. For example the nivolumab result is from the CheckMate 025 trial vs. everolimus (an mTOR inhibitor from Novartis) in patients with advanced RCC for who had relapsed previous treatment with one or two regimens of antiangiogenic (i.e. anti-VEGF) therapy (Motzer et al 2013: Just to note in passing, nivolumab therapy in this setting triggered an overall response rate (ORR) of 25% of patients, and had a modest but significant impact on progression free survival (PFS) and median overall survival (mOS). Indeed, 30% of the nivolumab patients that responded were alive 5 years later, as reported at ASCO last year (

The atezolizumab plus bevacizumab data come from the IMmotion 150 trial was presented at the ASCO GU meeting in February 2017 ( In that study the combination was compared to atezolizumab alone or sunitinib alone, in treatment-naive (frontline) patients. The combination produced an ORR = 32%, just a bit better that atezolizumab alone (26%) or sunitinib alone (29%) but there was a significant impact on PFS at 12 months particularly in patients having PD-L1+ tumors. Overall, patients achieved a median PFS of 11.7 months with the combination, 8.4 months with sunitinib, and 6.1 months with single-agent atezolizumab. PD-L1-positive tumors yielded a median PFS of 14.7 months with the combination, 7.8 months with sunitinib, and 5.5 months with atezolizumab monotherapy.  There was no improvement in the hazard ratio (a statistic that measures the chance of patient death in different cohorts in the trial, as compared to each other – arm vs arm).

Against that promising, if early, backdrop, the very interesting results here are shown wide right on the graph above – the combination of axitinib, a pan-VEGF-receptor inhibitor, with avelumab (anti-PD-L1, EMD Serona/Pfizer) and pembrolizumab (anti-PD-1, Merck). These studies are ongoing in the front-line setting. The overall response rate of axitinib with pembrolizumab of 67% is startlingly high, although these data are from a small study (I can’t find pembrolizumab or avelumab monotherapy data as a comparison, but the atezolizumab  monotherapy responses at 26% in the IMmotion 150 trial gives us a baseline for anti-PD-(L)-1 monotherapy in the frontline setting).

Axitinib is a second generation VEGF-R inhibitor with improved selectivity over earlier compounds, and also over some competing compounds (e.g. pazopanib from Novartis). Axitinib monotherapy in second line RCC produces an ORR around 40%, perhaps a little higher, with a modest impact on PFS when compared to sorafenib (~ 5 months v 3 months ). In the first line setting the monotherapy data are a bit better, with ORR reported at 50%+ and PFS of about a year, with a median overall survival (mOS) of about 2 years (similar to other VEGF inhibitors).

In that context an ORR of 55% (axitinib plus avelumab) or 67% (axitinib plus penbrolizumab) might be just additive (approximately 50% (axitinib) + 25% (anti-PD-(L)-1)). But then I came across this screen grab from the avelumab combo study:

Screen Shot 2017-06-09 at 12.16.02 PM

Notice here that the X-axis is in weeks, and so we have ongoing responses of greater than 1 year (52 weeks) in 14/32 patients (44%) and, as noted on the slide, a total of 24/32 patients (75%) with an ongoing response with a minimum of 24 weeks. Of interest, partial responses (PR, red triangles) evolved into complete responses (CR, blue triangles) over time, suggesting an ongoing immune response. Durability of response is critical here, but it certainly looks like this cohort will handsomely beat the 2 year mOS mark, which will best axitinib alone. The avelumab plus axitinib study ( and the pembrolizumab plus axitinib study ( certainly make the case for this particular combination (with axitinib); in contrast the combination of pembrolizumab with another kinase inhibitor pazopanib (that blocks VEGFR-2, KIT and PDGFR-β)  resulted in intolerable liver toxicity (

Ok so why is this important? I think there are a few interesting themes here. 1st, the combination of anti-PD-(L)-1 antibodies with standard of care (SOC) treatment, in this case, axitinib, has produce a result that dwarves what we see with epacadostat, the IDO inhibitor that would be expected to aid in blocking tumor immunosuppression. That is not to say that the epacadostat combination will not bring real benefit (again, it’s the durability plus the response that counts). It may very well do so, and it may do so with less toxicity (we’ve not discussed adverse events, which can be a differentiating feature).  However, the concept that pairing immune checkpoint inhibitors would unleash the anti-tumor immune response to bring synergistic activity is not robustly supported in this indication (not yet anyway).

That brings up the second interesting point, which is: against an ORR of 55% (using the avelumab plus axitinib data), how should we judge novel agents? If this is ‘noise’ arising from combination with SOC, how will novel agents overcome this and show a positive signal? The answer, simply, is in randomized, controlled trials (RCTs). Dr James Mule made this point at our IO combinations panel at the Sach’s pre-ASCO conference, and he used this slide from Lerrink to show how few immunotherapy combination studies are randomized (h/t to Dr Mule and to Lerrink):

Screen Shot 2017-06-11 at 11.20.31 AM








So despite the need for RCTs, we don’t see many yet. Another tactic maybe to drive a biomarker forward alongside a novel agent, in order to be able to select patients and differentiate in that manner. Easier said then done, but a possibility. The magnitude of potential noise is illustrated in this graph from EvaluatePharma’s nice IO combo report:

Screen Shot 2017-06-11 at 12.20.41 PMTheir report uncovered 765 immunotherapy combo trials across every conceivable indication. That is a lot of data to sift through.

Ok, the 3rd point. What I love about this example (renal cell cancer) is the focus on biology. Note here that we’ve not tried to be comprehensive about the RCC field, the treatment landscape, other novel targeted agents in development, the oncogenic drivers, the mutation burden, the tumor microenvironment, or the microbiome (all of these rooks will come home to roost in time).  Instead we focused on two classes of therapeutics that are playing well together – anti-angiogenic drugs (anti-VEGF, VEGFR inhibitors) and anti-PD-(L)-1 antibodies. That’s all. But this gives us a new way to think about the treatment and competitive landscape in immunotherapy – specifically, how are companies building on this early data.

To look at this I used an IO combo database that I’m beta-testing for Beacon-Intelligence, and an interesting theme emerged. So, a quick share (this is a quick look and again, not a comprehensive one by me: one can quickly find more studies in the database or online)

Screen Shot 2017-06-11 at 11.38.21 AM

So, not suprisingly, Roche/Genentech has a slew a trials designed to pair it’s anti-VEGF antibody bevacizumab with atezolizumab plus SOC in a range of indications including RCC.  What is a bit surprising is the appearance of Eli Lilly, bringing along it’s own anti-PD-L1 (LY3300054). While Lilly is not really considered an immunotherapy player, it does have some keen assets to deploy: anti-VEGFR2 antibody (ramacirumab) and, although not shown here, the multi-kinase inhibitor galunisertib that potently inhibits TGFBR1. The VEGF and TGFbeta pathways are intricately intertwined and both have profound impact on tumor biology, stromal biology, and immune biology. From a biology perspective, Lilly suddenly looks like an immunotherapy player in the making. Here I think is an interesting lesson, that is, follow the biology, not the molecule (btw, a search on TGFbeta combination therapy leads down another rabbit hole, best saved for later).

stay tuned