ImmunoGen’s ADC Platform and Toxicity – Not So Simple After All

November 11th, 2013
Last week ImmunoGen’s candidate Antibody-Drug Complex (ADC) IMGN901 (aka lorvotuzumab mertansine)blew up in a Small Cell Lung Cancer (SCLC) phase 2 clinical trial. ADCs are designed to bind to proteins expressed at a high level on tumor cells. Upon binding these antibodies are internalized and release their payload, a cytotoxic drug. The idea is to load the tumor cells with a lethal poison while sparing normal cells.
The problem with IMGN901 seems a simple one. The antibody recognizes and binds to its target (called CD56) expressed on tumor cells including SCLC, Multiple Myeloma (MM), and Merkel Cell Carcinoma. In the SCLC trial the ADC was given either alone or with standard-of-care chemotherapy, which is the combination of etoposide and carboplatin (E/C).
It can be inferred from the press release that the ADC treatment, either alone or in combination therapy (ADC plus E/C) did not work: tumor cells were not killed off to a degree that resulted in measureable clinical benefit. Further, “an imbalance in the rate of infection and infection-related deaths was noted between the arms” and it was “recommended that all patients discontinue IMGN901 treatment. Infection-related death is a recognized risk in SCLC trials, including trials with E/C. Among the 198 patients receiving IMGN901 as a single agent in early trials, there was one incidence of infection-related death; it was deemed possibly drug related”. The quotes here are from the Business Wire version of the press release.
The tumor protein targeted by ’901, CD56, is also expressed on immune cells called NK cells. NK cells are essential for immune responses to infection, and it is possible that they were diminished enough by ’901 treatment to allow infections to occur at a rate higher than normal. CD56 expression on human NK cells is bright by flow cytometric analysis, generally giving a 2-log shift in staining intensity with standard reagents. This suggests somewhere between 10,000 and 50,000 copies of CD56 on each NK cell. This is similar to the expression of CD56 on multiple myeloma cells. We do not have comparative data for SCLC but immunohistochemical staining on these solid tumors shows extremely dark CD56 reactivity, implying robust expression. On balance it is plausible that there was enough expression of CD56 on human NK cells to cause their depletion, and this might have been associated with a higher infection rate. However, proving this is the case is currently not possible.
The combination of lack of efficacy and high rate of infections led the company to suspend the clinical trial.
Of note, a high infection rate was not noted in the ’901 MM phase 1 trial. This is from the presentation given at ASH in 2012, as summarized by The Myeloma Beacon web site: “Of the 39 patients evaluated for response, 57 percent of the participants responded to the combination therapy, including 3 percent who achieved a stringent complete response, 28 percent a very good partial response, and 26 percent a partial response. The most common side effects of all grades (both moderate and severe) included peripheral neuropathy (56 percent), fatigue (42 percent), low white blood cell counts (32 percent), and low platelet counts (32 percent).” In this trial the combination therapeutics used were Revlimid (lenalidomide) and dexamethasone, and the side effect profile is consistent with the use of those agents.
On top of the increased infection observed in the ’901 SCLC trial, there is also the larger question of whether ImmunoGen’s ADCs are inherently unsafe. This hypothesis has been brought forward based on multiple reports of ocular toxicity associated with ImmunoGen’s drugs, thought to be related to the “linkers” used by ImmunoGen to attach its antibodies to the toxin. The linker is a biochemical construct – these are designed to be stable while in circulation, and then cleaved once the ADC is internalized by the target tumor cell. Much has been made of the hypothesis that some linkers are “leaky”, that is, they undergo some cleavage even while in circulation and that this leakiness contributes to toxicity. This is a very difficult hypothesis to evaluate, as ImmunoGen has used multiple different linkers for different ADCs. Also there is anecdotal reference to ocular toxicity seen with Seattle Genetics ADCs, which use a different linker and toxin combination.
Some insight can be gleaned from ImmunoGen’s recent patents in the area of linker development. US 2012/0282282 “relates to charged or pro-charged cross-linking moieties and conjugates of cell binding agents and drugs comprising the charged or pro-charged cross-linking moieties and method of using the same to reduce ocular toxicity associated with administration of antibody drug conjugates”. The benefit of charged groups in the linker appears to be to increase solubility, which implies that the ocular toxicity is a drug precipitation or drug aggregation effect seen at higher doses. Example 8 of the patent states that  “administration of the substituted charged sulfo-SPDB linker for the uncharged SPDB linker greatly decreases ocular toxicity in a rabbit model.” Example 9 of the patent further states that “MGN242 is an antibody drug conjugate for the treatment of CanAg-expressing tumors … Forty-five patients have been treated with IMGN242 at 8 different dose levels (18 to 297 mg/m2) in two clinical trials. Dose limiting toxicities (DLTs) included decreased visual acuity, corneal deposits and keratitis, which appeared to improve in patients where follow-up data is available. A two-phase pharmacokinetic profile was observed for IMGN242 in plasma from patients with low circulating CanAg levels (1000 U/mL) compared to patients with low levels (
So, at least in this case, a high level of circulating drug is associated with ocular toxicity.
Similarly, ImmunoGen reported ocular toxicity associated with IMGN853, an ADC targeting folate receptor alpha on diverse tumor types. Toxicity was managed by dose reduction (from 7mg/kg to 5mg/kg, as reported at ASCO this year – abstract #2573). Related to this observation is the finding that changing the dose schedule can reduce ocular toxicity, as reported for with SAR3419, an ADC targeting CD19 (Beck et al. 2012. mAbs 4: 637-647).
ImmunoGen appears to have suffered a perfect storm of negative publicity on the heels of the ’901 efficacy/toxicity results. The data relevant to a specific drug (’901) in a specific cancer (SCLC) have been extrapolated to reflect on the entire pipeline. As we’ve seen from even this brief look under the hood, it’s not that simple. The bigger issue of ocular toxicity may be addressable by adjusting dose and dosing schedule, although I’ll grant this is not an ideal situation. ImmunoGen is continuing the development of next generation linkers as discussed above. Finally, ImmunoGen is seeing continuing success with T-DM1/Kadcyla, an earlier generation ADC with its own issues and history, approved for treatment of Her2-positive breast cancer. However, StartUp, writing in 2011, disclosed that ImmunoGen’s upside from the Kadcyla deal with Genentech/Roche was very limited, with single digit royalties following the payment of ~50MM in milestones. At total worldwide current Kadcyla sales of ~200MM per annum this will not amount to much cash. I have no idea what ImmunoGen’s broader financial picture looks like, but investor’s were clearly spooked by last week’s news, driving the stock down over 20% on November 4th. It has regained about half of that ground since.
 
With at least 10 ADCs using 4 different linkers already in the clinic, trial results may inform ImmunoGen’s developmental strategy in a timely manner, and at the end of the day this technology may be refined to a point that it achieves broader application. In the meantime we should anticipate a bumpy ride, to be expected I think when innovation is the driver. ImmunoGen aside, Seattle Genetics, Roche, BMS, Takeda etc have programs running in this space, with 32 clinical programs listed by Credit Suisse in November 2012. A host of small companies (and investor’s dollars) are competing to add value, whether at the level of antibody technology, alternative scaffolds, novel linker technology or new ways to couple toxins in ADCs. Identifying the best chances for adding value is a discussion for another day.