The selective estrogen receptor downregulator GDC-0810 is efficacious in diverse models of ER+ breast cancer

Comments regarding the work on mechanisms focused on two points: First, there is an incomplete analysis of some of the results, which presumably could be readily improved. Second, a point of considerable discussion among reviewers involved the need to analyze the relationship between ER SERD and ER antagonist activity. This needs to be done in the context of the work by others in the field, but more importantly in the context of the in vitro vs. in vivo results presented in the manuscript itself: It is notable that GDC-0810, a SERD, and tamoxifen, an antagonist with no SERD activity, show comparable efficacy in suppressing xenograft growth. Hence, the statement, "Together, these in vitro results support the discovery strategy of optimizing ERα degradation in addition to ERα antagonism." appears to be contradicted by your results from the in vivo models. From what you present, one could conclude that beyond its superior PK and oral availability (which are clearly significant advances allowing higher dosing), GDC-0810 does not have mechanistic superiority over other SERDs and ER antagonists. This whole issue requires careful clarification and could benefit from additional work.

We thank the reviewers for bringing to our attention what might be considered a disconnect between what we highlight as being a key feature in vitro (i.e. ER degradation) versus what appears to be sufficient for anti-tumor activity in vivo (i.e. ER antagonism). We have made significant investments in addressing the relationship between ER degradation and antagonist activity, including experimentally, by removing some data to de-emphasize some points, and by better placing our data and discussion in context of what has been previously reported. We demonstrate, using additional experimental data, that GDC-0810’s ability to suppress ER signaling can be uncoupled from its ability to deplete ER, and acknowledge here, and in the Discussion that its potent ER antagonism coupled with good PK and bioavailability are likely all major contributing factors to robust in vivo efficacy, potentially independent of degradation. We also note though, in the context of newly included uterus data, that ER degradation may be associated with preventing ER agonism, which may be more relevant for long-term/acquired resistance in vivo. We believe the manuscript has greatly benefited from clarification surrounding this important issue (specific details below).

Summary:

The authors report an interesting and quite thorough study characterizing the orally active SERD GDC-0810 as a potential drug treatment for metastatic breast cancer driven by mutant estrogen receptors. Their strategy is based on first identifying compounds that work as estrogen receptor degraders (SERDs) that were then studied further in a variety of ways to examine their efficacy, pharmacokinetic properties and oral bioavailability. There is a clear need for such drugs, since currently used agents that also degrade ER (Fulvestrant) have poor pharmacokinetic properties and must be given to breast cancer patients by intramuscular injection, a route that is painful and probably still gives an insufficient exposure of the tumors to the drug.

Major Revisions Requested:

1) Analyze and discuss more incisively the relationship between ER SERD and ER antagonist activity, including how in vitro results relate-or stand in contrast-to in vivo outcomes. GDC-0810 is likely to have activity in several contexts of hormone-resistance where ER dependence is retained. Work by other investigators using other SERDs and ER antagonists in these types of resistance models should be cited and discussed (Wardell, Ellis et al. 2015 in Clinical Cancer Research; Couillard et al. 1998 in Cancer Research). In particular, Discussion should address in some detail the Wardell paper as well as previous work by those authors (Wardell et al. 2013 Clinical Cancer Research, Wardell et al. 2011 Biochemical Pharmacology) regarding the question of whether degradation of ER by GDC-0810 is necessary for the drug's antitumor effects.

We have used the “ER over-expression strategy” described by Wardell et al. (2013) Clinical Cancer Research and Wardell et al. (2011) Biochemical Pharmacology in order to test whether degradation of ER by GDC-0810 is necessary for the drug’s antagonistic properties in vitro. These data are provided as new panels in Figure 2C, D, E, and are described and discussed in the last paragraph of subsection “GDC-0810 induces a distinct ERα conformation versus tamoxifen and other ER therapeutics, and does not exhibit tamoxifen-like ER agonism in MCF7 cells”, and in the second and last paragraphs of the Discussion. We find, similar to what had previously been reported for fulvestrant and bazedoxifene, that the depletion of ER is neither required for the ability of GDC-0810 to suppress the expression of PR, as a read-out for ER pathway activity, nor for its ability to suppress the proliferation of MCF7 cells. This is entirely consistent with the demonstration that GDC-0810 acts directly as an ER antagonist, by outcompeting E2 and changing the ER conformation such that the PGC1α co-activator, for example, is displaced, in cell free assays where degradation does not occur. We note here that while over-expression of ER had previously been seen to block ER degradation mediated by fulvestrant, in our experiment fulvestrant retained its degradation activity. We speculate that the overexpression of ER in our doxycycline-inducible cells may not be as high as what was reported previously – though these levels are sufficient to prevent GDC-0810’s ability to deplete ER, consistent with distinct mechanisms of degradation driven by fulvestrant versus GDC-0810.

2) Provide a conclusion in which the factors of SERD and antagonist efficacy and potency, as well as PK, oral activity and accessible dosing are presented in a balanced way. In Figures 1 and 2, a distinction is made between GDC-0810 and Tamoxifen-like compounds on the basis of GDC-0810 inducing a different conformation and causing ER degradation, but then, in vivo, there is little difference between Tamoxifen and GDC-0810 in several xenografts (3A, 3C, 3D, 6B). It would be valuable if you could show a context (e.g., resistance due to Y537S) where the degradation matters in vivo. If not, you could simply conclude that at this point there is no correlation between ER degradation and in vivo activity, and that in vivo efficacy represents mainly potency of antagonism and improved PK.

We have made great efforts to perform dose concentration curves where appropriate, and where logistically reasonable. The rationale for selecting a single fixed dose for Figure 2A is described above. For Figure 2B, we likewise selected a single high dose, across a number of genes, with a collection of positive controls in an attempt to screen for agonistic activity of GDC-0810 in MCF7 cells. It seems that at least 1 reviewer is particularly interested in whether we see any agonistic activity for GDC-0810 at low concentrations (see also point 10 below) perhaps triggered by the observation that some ER ligands can exert a biphasic response on ER signaling. We have therefore included additional data here, showing results from a gene expression analysis (Taqman) as a dose response curve. Unlike what we find for 4OH-tamoxifen, we do not observe evidence of agonism triggered by GDC-0810 in MCF7 cells, regardless of concentration or presence versus absence of E2 (see Author response image 1). Though such lack of agonism may be tissue context dependent, since we do see weak agonism in Ishikawa endometrial cells.

Author response image 1

Download asset Open asset

8) Figure 1C should include Fulvestrant as direct comparison.

We have included this as Figure 1D.

9) Related to Figure 2A, it would be worth mentioning whether GDC-0810 is agonistic in other tissues such as uterus or bone.

We have provided additional data, as Figure 3A–F, detailing the activity of GDC-0810 in Ishikawa (endometrial) cells in vitro, and in the rat uterus in vivo. We find that GDC-0810 exhibits mild estrogenic activity in this context. We note that in this context, GDC-0810 does not exhibit robust ER depletion, though fulvestrant, which acts as an inverse agonist does degrade ER. These observations are described in the subsection “GDC-0810 displays mild estrogenic activity in uterine models in vitro and in vivo” and in the second paragraph of the Discussion.

10) Is the data shown in Figure 2C generated in the absence of presence of E2? The authors should include a dose response curve of GDC-0810 in the absence of E2. This should clearly be shown. The data shown in Figure 2C are not conclusive – it seems that GDC-0810 has some agonist activity at low doses (?). It can't be compared to Fulvestrant since that data looks inconsistent at low doses. This experiment should be repeated.

We have provided a dose response curve for GDC-0810 (as well as fulvestrant and 4OH-tamoxifen) in both the absence and the presence of E2, as requested, in Figure 1E, F. We see no clear evidence of agonism using cell proliferation as a read-out, including for 4OH-tamoxifen, for which we see some agonism at the ER pathway level, though only for a sub-set of target genes.

11) For in vivo treatment (e.g. Figure 3, and Figure 3—figure supplement 1B), it is assumed that the GDC-0810 treatment is given in the presence of E2 (it is described in more detail later in the manuscript). This should be more clearly indicated in both the figure and the legend.

This information had been provided in the figure legend as, for example “in the presence of 60-day release 0.36 mg 17β-estradiol pellets”, and “HCI-003 patient derived xenograft tumors were implanted in mice containing a 1 mg 17β-estradiol beeswax pellet”. The presence of E2 is indicated in the figure itself as “Vehicle (+E2)” versus “Vehicle (-E2)”.

12) It is not clear why the investigators choose to study interaction of the ESR1 mutant with PGCalpha (Figure 5A), which is a less well described co-activator in the context of ER and its mutants. Why didn't they test SRC1, SRC3 or other better-described co-activator interaction? (Or is the rationale for this experiment linked to data shown in Figure 1E? If that is the case, it should be stated).

We originally profiled a collection of ER ligands (not including GDC-0810), against a number of co-activator peptides in the context of WT and mutant ER. Though trends were similar across these peptides, the absolute signal and dynamic range varied. We thus selected PGC1α for further studies since it reflected the general trends, had robust signal and dynamic range, and was also commercially available in a format amenable to our studies in a large number of compounds.

13) It would be worth evaluating mutant and WT expression at the end of study in Figure 6A to evaluate transgene expression (mRNA) and degradation.

Since this model (previously Figure 6A, now Figure 8A) is an artificially engineered model, over-expressing mutant ER under the non-endogenous EF1 promoter, we chose to perform pharmacodynamic/response studies in the WHIM20 PDX model instead. These new data are presented in Figure 8B–D.

14) It is unclear when biological or technical replicates of experiments have been done. This should be more clearly stated (and possibly performed if not in some circumstances). Some figures lack statistical analysis (e.g. Figure 5).

We have tried to be explicit about replicates and have added the requested statistical analyses.

15) IC50s for GDC-0810 and Fulvestrant for WT and mutant cells should be given.

We have added IC50 values to all of our dose response curves.

16) What are PR levels in the ESR1 mutant tumor -/+ GDC-0810 treatment?

We have included western blot, RPPA as wells as IHC data, evaluating levels of PR as Figure 8C, D.

17) Wardell et al., 2013 reference should be completed.

References have been updated.

18) The Methods are quite well written and thorough; however, it would be helpful if figure legends stated the number of days after cell injection or tumor fragment implantation that treatment with ligands was begun.

Our standard procedure for efficacy studies is to group tumor bearing animals based on tumor size (an average of 150–300mm³), prior to initiating dosing, rather starting dosing based on time since implantation, since the time for tumor initiation can vary widely from animal to animal. This information is included for specific models in the Methods, last paragraph,Figure 8 legend, subsections “TamR1” and “Xenograft Studies” and Figure 4—figure supplement 2 legend.