Celyad — Update 28 November 2016

NKR-2 CAR T-cell therapy involves adding a new set of genes into the patient’s CD8 T-cells so they have an artificial natural killer (NK) cell receptor as well as their natural T-cell receptors (TCR). The NK receptor recognises a wide variety of cell surface proteins (ligands) which are expressed at high levels on cancer cells; these ligands are also produced by stressed normal cells. In theory, this targets the NKR-2 construct to a wide variety of cancer types. Most other CAR T-cell therapies in the clinical target CD19, a ligand only found on immune cells associated with antibody production.

The completed Phase I safety study involved four cohorts (of three patients each) treated sequentially with a rising single dose of NKR-2 cells. The dose was increased to 30m cells for the last cohort. Given that Celyad ended the Phase I at a relatively low single dose it would be unexpected on current CD19 CAR criteria to detect significant signs of efficacy. What efficacy signs were reported have not yet been disclosed. However, they were noted at a dose between 50 and 1,000 times lower than Celyad had observed in animal experiments.

A major drawback with the current CAR CD19 focus of other companies is that while it is excellent for B-cell tumours (like leukaemias and lymphomas) CD19 will not target other cancers. Each solid tumour type has a different profile of antigens and often these are also found on normal tissues, albeit at much lower levels.

Of the leading CAR companies Novartis, Juno, Kite and Bellicum, only Bellicum has a single solid tumour trial running: a Phase I in advanced pancreatic cancer with an anti-PSCA CAR construct. This is a 30-patient dose ranging and safety study due to complete in 2020. A National Cancer Institute melanoma study by the Rosenberg group reported in 2016. It used CD4 CAR T-cells targeted to the MAGE antigen. Melanoma responds well to immunotherapies. The Rosenberg group has taken the view that CD4 T-cells will be better for solid tumours than CD8 T-cells used in CD19 CAR therapies.

The main effort in solid tumours in immune oncology is based around checkpoint inhibitors like the marketed products Yervoy (ipilimumab), which targets cytotoxic T-lymphocyte-associated protein 4 in melanoma, and Opdivo, an anti-PD-1 monoclonal antibody used in melanoma, gastric cancer and renal cell carcinoma. It is possible that checkpoint inhibitors will be combined with CAR therapy at some point in future but this is currently some way off clinical development.

With the advantage of the different mechanism of cell targeting used by NKR-2, Celyad plans to use NKR-2 in multiple cancer types. NKR-2 attacks multiple targets found on severely stressed cells, a profile shown by most cancer types. A risk is that NKR-2 may attack normal but stressed cells so some side effects are to be expected. Celyad does not report any significant side effects to date but the safety study was small.

The next Phase Ib study, THINK (THerapeutic Immunotherapy with NKR-2), has been designed by Celyad as an open-label, multiple-dose US and European study. It will assess higher dose levels, safety and clinical activity of autologous NKR-2 cells in seven refractory cancers:

In the THINK Phase Ib, Celyad intends to increase the dose starting at 3x108 (adjusted for body weight) then 1x109 then 3x109. At each dose, the patients will receive three successive administrations, two weeks apart, of NKR-2 T-cells. There will be 24 patients and eight per dose group. They can be from any of the above cancer types.

In the expansion phase to test efficacy, Celyad intends to enrol up to 86 more patients to evaluate each tumour type independently. Combined with patients in the dose escalation phase with the same tumour type, this should give at least 14 patients per cancer.

Celyad plans to use three doses because it has been shown in CD19 CAR T-cell dosing that giving two high cell doses rather than one very big cell dose gives fewer side effects like cytokine release syndrome (CRS) and neurotoxicity. It is also apparently more effective. There are some crucial differences between NKR-2 and CD19 CAR T-cell therapy however:

Currently, the THINK study is only approved in Belgium; approval in other European countries and the US will be required. The overall timelines have not been disclosed by Celyad and there is no record as yet posted to ClinicalTrials.gov. Edison estimates that if the FDA approves the trial design by the end of 2016, the dose escalation phase may have completed by approximately springtime. In that event, the expansion phase should run for the rest of 2017 implying data possibly by late 2017 or more realistically from early 2018.

Timelines beyond this point are inevitably unclear at this time but Edison is currently assuming that potentially pivotal studies start in 2019 or earlier and complete by 2021 or earlier. This implies that NKR-2 products could receive approvals from 2022 onwards, particularly if breakthrough and fast-track designations are received. There is insufficient data to form an opinion about this at this time.

A recent review in Molecular Therapy – Oncolytics by Newick et al (2016), based at the University of Pennsylvania, a leading centre for CAR therapy development, clarifies the multifactorial difficulties involved in solid tumour therapy.

The grafted CAR T-cells have to move from the injection site into the solid tumour. To do this, the CAR T-cells have to have the right adhesion molecules to even bind to the cancer cells. Once inside the tumour, the CAR T-cells will be in a difficult and hostile environment due to low oxygen and nutrient levels and high levels of acid. It is also full of other immune cells, some of which will down regulate a killer T-cell response. The tumour environment is also believed to contain multiple inhibitors of T-cells. Finally, the tumour itself is often heterogeneous so not all tumour cells may be recognised by the CAR construct. Generally, haematological cancers like leukaemia are relatively homogeneous. The review indicates that about 30 solid tumour antigens are being evaluated for use in CAR therapy. Finally, because of the lack of highly specific solid tumour antigens, there is a risk of the CAR T-cells attacking healthy tissues that express the antigen at a low level. This is not theoretical; a case study from 2010 reported that a HER2 CAR patient died due to off-target toxicity.

It is too early to develop detailed predictive models for solid tumour NKR-2 sales. Exhibit 1 looks at the number of US deaths (SEER database) in each for the two haematological cancers and the five solid cancer types. The cancer incidence is also given. The Edison value assumes $150,000 per treatment. This may be perceived as very low in the putative acute lymphoblastic leukaemia (ALL) market, where $500,000 is talked about currently as affordable. However, if and when a therapy treats many cancer types, the overall cost to the healthcare system becomes crucial. Exhibit 1 values are also based on deaths as a proxy for refractory late-stage cancers. Earlier-stage cancers could be treated as well, which would magnify the market potential in some cases.

In context, the CD19 favourite, ALL, has about 1,500 deaths per year, all tragic, mostly children, so a worthwhile indication but a small market.

The probabilities used reflect a best estimate at this point but have no substantive evidence base. Ovarian scores highest as there is published preclinical work. Pancreatic cancer is notoriously hard to treat, so a high market share could be expected if NKR-2 had an impact on survival, but a very low probability is assigned.

The global market opportunity is based on 75% of the potential US sales as the US is the main market for high-value biological therapies.

TCR-Inhibitory Molecules (TIMs) are an approach used by Celyad, currently in preclinical development, to manufacture allogeneic (from another person) CAR T-cells that will not attack the host tissue triggering graft-versus-host (GvH) disease. GvH occurs when allogeneic T-cells are injected (grafted) into a cancer patient and start to attack healthy tissue as the patient (host) is detected as non-self to the graft. As T-cells recognise foreign tissue through the TCR, eliminating functional TCRs in an allogeneic graft will avoid GvH disease. Autologous therapy, the current paradigm, avoids this as it uses “self” cells, but it is expensive as every dose is custom made. Allogenic T-cells would be cheaper to produce and could be “out of the freezer”, enabling rapid treatment. However, it might also be necessary to have a wide bank of different tissue types to ensure some level of tissue matching as otherwise the host immune system would attack and eliminate the grafted cells.

TIMS are either short hairpin RNA (ShRNA) or dominant negative proteins. These genes are linked to the CAR gene construct used to transfect the cells. ShRNA stop the production of TCR proteins; dominant negative proteins are synthesised by the T-cell and assembled into non-functional TCR. These T-cells now only attack patient tumour cells as they cannot “see” any healthy tissue.

A successful patent requires that the idea is novel, that there is an inventive step, and that the invention has a practical function. Most of any patent is taken up with a description of what was already known and a detailed description and examples of the invention and its application. What is protected is stated in the claims. It is usual to start with a very broad Claim 1 and then move to more specific claims that are more easily defended if challenged.

In November 2015, the US patent office granted US9,181,527; this was filed by Prof Sentman in 2009. Claim 1 covers any TCR deficient CAR T-cells to prevent graft-versus-host disease. No international filings were made so Claim 1 only applies in the US. There is no direct prior art cited in the literature so the US examiner in October 2015 granted the patent including Claim 1.

Claim 1 was challenged in February 2016 and is being re-examined by the US patent office; this will take many months and follows a set procedure. Claim 1 remains in force during the process.

The position notified by the examiner in August 2016 was that there was a case that the invention was obvious so an initial opinion has been issued that would, if implemented, strike out Claim 1 only. The basis of the opinion is that diverse elements of the concept of TCR-disabled CAR modified T-cells have now been identified in literature. These are from three diverse literature references separated widely in time and from different journals. The examiner mostly disregarded the prior art cited by the challenger. However, no one prior to Professor Sentman in 2009 had made a specific invention with the function of avoiding graft-versus-host disease. There is always a danger of hindsight in that what now appears obvious can be selectively picked out retrospectively and claimed to be obvious to the idealised “person skilled in the art”.

Again, as is normal Celyad did not respond to the initial examination but waited for the initial findings. Celyad will now respond to the examiner’s initial report. This may cause a reconsideration or maybe a reaffirmation of the original finding. Whatever the outcome, it is likely that one of the parties will appeal. Eventually, the US patent office or courts will either uphold Claim 1 or amend the patent and remove it. Other claims are not affected but may depend upon Claim 1.

While this process proceeds, the patent with Claim 1 remains in force. This could, in theory, block a number of other companies from developing allogeneic therapies. However, allogeneic CAR therapy is many years from any possible regulatory approval and launch in the United States. Patents do not prevent research work from occurring and using the invention but they can block commercial exploitation. Celyad has indicated that it may initiate an allogeneic NKR-2 trial in H217.

Celyad holds other IP covering the TIM technology in detail. In the US, granted patent US9,273,283 specifies the way in which the allogeneic cell is made TCR deficient. In Europe (EP2844742) and elsewhere, the related patent is still an application. These patents do not have the broad scope of Claim 1 in the 2009 patent, but are more specific and potentially therefore stronger patents. Edison assumes that Celyad is filing further intellectual property. Any patent applications will normally be published about 18 months after their filing date.

At the current time, the patent challenge is of minimal relevance to investors because of the early nature of the allogeneic CAR technology space. It does not for example affect the Ono deal on allogeneic CAR therapy in Asia as the patent does not apply there. The key aspect for Celyad investors is the scope and speed of development of the current autologous NKR-2 product.

C-Cure remains as a large part of the Celyad value. It is currently at a 35% US and European probability as the CHART-1 study did not meet its overall primary endpoint. This includes the probability that a partner will be found to fund the CHART-2 study. A presentation on the data in Rome on 28 August focused on a responsive subgroup based on end-diastolic volume (EDV); the full data set has not yet been published. Edison notes that this probability may be too low given the EDV-defined subgroup response. It is anticipated that the CHART-2 trial will focus on this group. Celyad management has commented that in this EDV-defined subgroup, “a statistically significant positive difference was seen in all individual elements of the composite primary endpoint”.

In H216, Celyad will receive an €11.25m upfront fee from Ono in respect of the allogeneic CAR deal signed in July. Celyad’s H116 results showed increased R&D spend leading to a cash outflow from operations of €20.8m. This includes a working capital outflow of €3.6m in H1 largely due to a $5m payment in respect of the Oncyte acquisition in January 2015. Edison now estimates that the cash outflow over 2016 as a whole will be €29.6m, indicating year end cash of about €78m, previously €86m.

In 2016, Edison forecasts R&D expenditure of €32m and cash admin costs of around €5m. This fits with the €9m per quarter cash use rate in 2016 and €87m 30 September 2016 cash level after the Ono payment. Celyad management has stated (Q316 report) that it believes that Celyad has cash to mid-2019. Given that any other deal payments are not clear, Edison has adjusted its forecast of the cash burn to about €34m per year in 2017, 2018 and 2019. This gives a 2017 year end cash estimate of about €46m (formerly €27.9m). This implies R&D spend will have to average no more than €29m per year or that additional cash will be obtained from deals and milestones, for example, from partnering of C-Cure.

Celyad has made some adjustments in its H116 accounts. It acquired BMS SA for €1.5m (biological manufacturing services); BMS is a separate but closely linked Belgian company that owns the clean rooms used for NKR-2 manufacturing. It has also adjusted the fair value of Belgian government loans provided for C-Cure development. These loans are repayable in the main part if C-Cure is commercialised. As the earliest C-Cure launch date is now in the 2020s and depends upon partnering, management has reduced the fair value of the loans by €2m, giving an exceptional gain in the P&L. In addition, the fall in the share price has led to an accounting adjustment to share-based payments recorded in the P&L as a loss of €2.2m.

The valuation structure has now been revised to put NKR-2 CAR T-cell therapy into the lead position in Exhibit 2. This gives rise to the following changes:

In July, we increased on an interim basis the NKR-2 probability from 17.5% to 18.5%. A nominal €10m value for allogeneic technology was added after the ONO deal. This gave an indicative value of €41/share. The new indicative value on the revised interim basis is €45 per share. No further dilution is expected until 2019 and possibly later depending on any deals in the interim.