In previous posts, I’ve covered several improvements of Tocagen’s regimen versus historical attempts. However, it’s been said over and over that this is a high-risk trial.  It’s important to discuss why. (Previous posts: https://msbio.tech.blog/category/tocagen/)

Pre-clinically

There are a few points of interest surrounding the retrovirus and FC combo.  First, several studies have shown adenoviruses alert the host immune system and generate IFN and potentially neutralizing antibodies (Alvarez 2000).  Retroviruses have not been shown to generate an IFN response, preclinical studies have found anti-RRV antibodies (Tai 2008).  There were no changes in overall transgene expression in cells, but these antibodies may slow replication.  Retroviruses also cause minor inflammation, which has led to the use of concomitant bevacizumab.  While there are no preclinical studies testing Toca 511 with bevacizumab, VBL-111’s activity was shown to be inhibited by co-administration of bevacizumab, which may partially explain VBL’s trial failure.  Bevacizumab is also known to phenotypically change glioblastomas into more invasive forms. A few other viruses tested with bevacizumab do not seem to lower viral activity, but it is not out of the real of possibility that bevacizumab or another agent has an effect on Toca 511 (Thaci 2013).  In fact, temozolomide + Toca 511 administration delays RRV spread, though the same amount of CD gene was eventually transduced (Huang 2013). Though there may be a more significant reduction effect in patients than in cells.

Finally, there is evidence in early studies of GCV resistance due to TK gene splicing with the HSV-tk+GCV combo.  The cells with the spliced gene survived treatment, and eventually spread.  After multiple rounds, GCV was no longer effective (Garin 2001).  While we have not yet seen this with Toca FC, it must also be considered.  To add to this, the bystander effect is critical to this treatment’s effectiveness.  Some preclinical studies showed a threshold of prodrug needed in order for the bystander effect to be meaningful (Barese 2012).  This specific threshold is not known for FC, but it does mean enough CD gene will have to be transduced for optimal anti-tumor effect.

Clinically

From the Phase 1 analysis, responders didn’t show a significant improvement in CD8 T cells vs non-responders, which is thought to be important for lengthy tumor immunity.  The tumor microenvironment is complex, and while the patient numbers were small and other biomarkers were significant, it’s hard to tell if the responses are a direct result of the Toca treatment.  

One major question IV administration didn’t work as well as the resection arm.  Toca 511 was first given via IV, then a resection was performed, and Toca 511 injected into the resection cavity.  Thus, it’s reasonable to expect a similar response rate from both arms.  While the IV arm mOS was comparable at 13.6 months, only 1 PR (in AA) and 1 CR (GBM) was seen. ** Correction: It was pointed out to me that the ORR% between IV and resection arms are consistent**

Several studies have mapped the recurrence pattern of GBM.  Distant tumor growth (away from the original site of resection) occurs in 3-15% of patients, and resection technique didn’t change this pattern (Zhou 2016).  I do not see a reason why Toca 511/FC would be effective in patients with distant tumor regrowth.

Phase 2/3 Trial 

The biggest danger is comparing the Phase 1 trial to historical results.  The baseline characteristics of patients between trials differs so much that the comparison can be confounded.  For example, historical data do not necessarily stratify by the same factors as TOCA 5.  TOCA 5 requires a smaller than 5cm3 tumor, which is not the case in many historical chemotherapy trials.  In fact, the TOCA 5 population is one of the more enhanced populations out of any previous rGBM trial.  Thus, the SoC comparator arm may outperform historical controls.  GBM patients eligible for resection may have a naturally better prognostic course course (Lu 2018). To add to that, cancer trials tend to show longer SoC survival in modern years.  As physicians become better at diagnosing and choosing treatments, patients tend to live longer.  In a specific subset of patients, mOS was seen to reach 10.4 months (Carson 2007).

Taking another glance at the responders in the Phase 1 trial, there is a slight imbalance in baseline.  There were a disproportionate number of patients with a KPS (Karnofsky Performance Score) of 90-100, and they are younger patients.  KPS and age are two of the biggest prognostic factors for longer survival.  The Phase 2/3 criteria were also decided based on a retrospective analysis, which could over-inflate the perceived benefit.  Furthermore, the SD and PD patients in the Phase 1 resection study also showed robust survival, and the mechanism behind that is unknown.  A combination of the SoC arm outperforming expectations (9.8 months) and a lessening effect of the treatment arm would make the success of the trial less likely.

Finally, is this combination best approach?  Tocagen tested a combination of temozolomide and Toca 511/FC in a preclinical model.  The two treatments showed synergy, and demonstrated significantly more potent anti-tumor activity (Huang 2013).  With so many pieces of the suicide gene + prodrug system, there may be a more optimal approach than what is currently being tested. A general theme we are beginning to see across cancer types is the use of combination treatments. A combination in rGBM may also post the best efficacy.

Conclusion

Overall, this has been an intriguing dive into a promising treatment for recurrent GBM – a field that has not seen any progress in decades.  The indication is tough, but any benefit seen in this trial would likely lead to approval in some aspect.  I stay cautiously optimistic in terms of a financial investment, but incredibly hopeful for the patients and families battling this disease.

So what were the key takeaways from last time? (Read part 1 here: https://msbio.tech.blog/2019/01/06/thoughts-on-toca-5/).  There have been previous, unsuccessful trials using suicide gene/prodrugs in the past.  Tocagen may overcome some of these challenges by converting the prodrug into a more potent molecule, and by using a vector that has increased stability and activity.  I’ll talk about some more tidbits I’ve learned on the improvements Tocagen has made vs past preclinical/clinical studies using vectors.  Additionally, I lightly touched on some trials in the previous post that were successful in small trials, but ultimately failed in larger trials.  I think it’s important to dig deeper into some of those trials to see if there are any takeaways.

Differences Between Past Oncolytic Virus Studies

Toca 511 is a non-lytic, retroviral replicating vector (RRV).  This is actually a pretty important distinction, if you comb through previous suicide gene/prodrug literature.  The failed Ark Therapeutics (HSV-tk + GCV) trial was tested with a non-replicating adenovirus.  Subsequent preclinical studies showed that transducing the tk gene with a replicating adenovirus showed increased transgene expression. This replicating adenovirus itself was lytic, and showed oncolytic capabilities.  Surprisingly, the addition of GCV did not significantly reduce tumor size (Wildner 1999, Lambright 2001).  To explain, a study done in 2000 showed in vivo that the cytotoxic nature of vectors actually is inversely correlated with suicide gene effectiveness.  The hypothesis is that the viral cytotoxicity interferes with the production of the TK gene, and thus with the bystander effect.  More importantly, the oncolytic power of the vectors, without a  prodrug, doesn’t consistently inhibit tumor growth on its own (Moriuchi 2000).

Preclinical studies show that RRVs, like Toca 511, do not naturally induce type I interferon response (IFN), while lentiviruses do, which slows their spread.  Exogenous IFN will inhibit RRV spread, as expected (Lin 2014). This signals that the nonlytic nature and the use of an RRV in Toca 511 may be less likely to trigger a response that could inhibit the spread of the virus.  Not to mention, Toca 511 does not rely on the lysis of host cells to spread, unlike other vectors.  Adenoviruses, which lyse the host cell, may alert the immune system to attack the virus.  Furthermore, human gliomas tend to have defects in IFN signaling, and it is theorized that Toca 511 spreads preferentially in IFN-defective cells.  Residual cancer cells may also serve as a viral reservoir and enable a longer-term protection against tumor growth.

Comparison with Non-suicide Gene Studies

Some of the more recent high-profile failures using different approaches were VBL Therapeutics and Celldex (and in a similar vein, ABT-414).  There are differences in each of the Phase 3 trials from the prior Phase 2 that may have contributed to the disappointing results of the studies.  Starting with VBL-111, the Phase 2 study began with a priming of VBL-111 only, followed by the addition of bevacizumab.  The Phase 3, however, had patients on the combination of VBL-111 and bevacizumab, which may have contributed to the vastly different results (15 months in Phase 2 vs 7.9 months in Phase 3).  In the press release, VBL noted: “the only significant change between the Phase 2 and Phase 3 treatment cohorts was in the treatment regimen”. 

Celldex (vaccine) and ABT-414 (monoclonal antibody + toxic payload) targeted EGFRvIII, which is often over-amplified in glioblastomas.  However, neither of these studies showed a treatment benefit, and there was no trend with EGFR expression and survival.  One explanation is that between 50-82% of patients lose EGFR expression upon recurrence (Sampson 2010).  Single-antigen targeting in glioblastomas may be unable to overcome the challenge of epitope escape.  Another difference in the Celldex Phase 2 and Phase 3 studies is the inclusion criteria for resection.  In the Phase 2, they defined gross total resection as < 1cm3.  However, in the Phase 3, resection was < 2cm3.  Celldex defended this notion by proposing that the vaccine should have worked better with more tumor, since more EGFR would be expressed (Weller 2017).

There are a few studies that lead me to disagree with Celldex’s defense.  Tumor size may play an important role in resistance to therapies.  After reaching a certain tumor size, poor T cell trafficking and higher immunosuppressive features negated any cytotoxicity in preclinical studies with an adenovirus + GCV.  CD8 cells were still being generated in the larger tumors, but were unable to inhibit growth (Predina 2012).  Even a partial resection of these larger tumors showed significant cytoreduction.  The less tumor, the better – hardly a surprising conclusion.  However, it’s hard to calculate what this threshold will be for human patients.  This study also potentially gives us an insight into why intratumoral trials (without resection) fail to see as lengthy of benefit.  Menei in the early 2000s attempted to treat GBM by implanting 5-FU microspheres.  One of the trials was attempted by implanting into the tumor, and the other into the cavity wall after resection.  The resection studies had patients that survived longer than the intratumoral (Menei 1997, 2003).  However, these were small studies (n=8, n=10) with varying baseline criteria, so I’m hesitant to draw too many conclusions from this. Of note, VBL Therapeutics also saw a trend for better outcomes in patients with smaller tumors. Tocagen has done its best to keep the trial design as consistent as possible with the Phase 1.

Conclusion

Previous suicide gene/prodrug trials used a different vector or a different prodrug.  Both of these differences may play a critical role in the efficacy of the therapy.  Perhaps Tocagen’s improvements will enable them to overcome the survival improvement hurdle. However, it’s important to stress that this is still a high-risk trial, and while I do think there’s a better chance of success because of an improved vector and consistent trial design, it’s no guarantee of success.  I’ll talk about some risks I’ve identified to the trial outcome next time.

Why could Tocagen be different from previous suicide gene/prodrug clinical trials?

Toca 511/FC is an elegant combo, but not a unique one. In fact, several trials have attempted this strategy, but none have succeeded. In 2000, a clinical trial using a similar concept in gbm failed, using a non-replicating adenovirus encoding herpes simplex virus thymidine kinase (HSV-tk) and ganciclovir (GCV), called Cerepro (Ark Therapeutics). Similar to Toca 5, the virus is injected after resection and GCV given by IV. GCV is converted into a toxic metabolite (GCV-triphosphate). Significant therapeutic effect was seen when ~10% of tumor cells were transduced. In a Phase 1, 21 rGBM patients were tested, and mean survival was 15 months vs 7.4 months control. However, in a Phase 3, 248 ndGBM patients were enrolled. Median survival was 365 days vs 354 days in the control arm (not sig). So what happened? One of the biggest benefits of prodrug therapy is the bystander effect, in which the converted toxic metabolite kills surrounding, non-transduced, cells. GCV does show the bystander effect, but is very reliant on gap junctional intercellular communication – meaning the cells have to be neighboring for the metabolite to be diffused between cells. Tumor tissues, however, rarely have such ordered cell junctions. Another bottleneck is the efficiency of GCV conversion, when HSV-tk also has a high affinity to its natural substrate, thymidine. Lastly, the efficacy of transducing enough cells via non-replicating vectors has been shown to be a hurdle, as others attempted to use adenoviruses with GCV, but also ran into similar issues. (Karjoo 2016, Tobias 2013, Rainov 2000).

Toca 511/FC is a replicating retroviral vector, and converts 5-FC (oral) to 5-FU via the inserted gene cytosine deaminase (not naturally occurring in humans). 5-FU has a neutral charge and small size, letting it diffuse into neighboring cells, which means it is not dependent on gap junctions. Using a bacterial CD gene does have a natural affinity to a substrate, cytosine. However, Tocagen designed their vector to use CD from yeast, which has a significantly higher affinity for 5-FC. A major limitation can still arise from % viral transduction, but Tocagen designed their vector to have improved stability and gene expression. Toca 511 was shown to have 3x increased activity over the prototype RRV (Perez 2012). What this translates to, in mice, is roughly 60x the amount 5-FU than what is seen in humans given 5-FU in colorectal tumors. In mice, clinical benefit was seen at 3% viral transduction, dose-dependent until about ~15%. Immune response seems to be the dominant mechanism of action, since mice in remission that were re-challenged with tumors still showed long-term protection (Hiroaka 2017).

Finally, why have so many early trials in rGBM shown promising results, but all falter in larger stages? This could be due to favorable baseline criteria, and of course, lack of power to show efficacy. In the Phase 1, of 6 patients, 2 were AA and had IDH mutations, 4 were less than 45 years old. All also had a KPS score over 80+ (most 90). However, responding patients tended to have low genomic burdens, which contrasts the idea that IO-type therapies works best in high mutational cancers. Importantly, Tocagen has stressed that the Phase 3 design is similar to parts of the Phase 1, which has not always been true, in the cases of Ark, Celldex, and VBL (Chiocca 2018).

The suicide gene/prodrug system is very complex, and much can go wrong. Is the new vector able to transduce enough cells? Will enough 5-FC cross the BBB and be converted to 5-FU? Is the amount of 5-FU enough to induce cytotoxicity and an immune response? Is there some other factor not seen that is preventing an immune response? There are a lot of variables, but when it goes right, it works. Past therapies have been inefficient and unreliable. Not to say this isn’t the case for Tocagen, but 6 patients in complete response for over 2 years is impressive, with signs pointing to an IO-like immune mechanism behind the response. Will the Toca 5 trial hit the interim or the final readout? After all this, the answer is, I have no idea. I sure hope it does, as not many therapies have been able to move the needle in decades. Given the benefits of the vector and trial design, it certainly has a better chance than previous trials.