Additional Risks of Gene Therapy
Hepatocellular Carcinoma
The Adeno-associated virus was discovered in the 1960s and for decades was not thought to cause any human disease.29 For this reason, it was an ideal candidate for use as a gene therapy vector. However, in 2015, Nault et al reported wild type AAV (wAAV) insertions near the TERT promoter in human hepatocellular carcinoma (HCC) specimens.30 Additional wAAV insertions near 4 other oncogenes were found, and the insertions were clonal.30 These same oncogenes are targeted by hepatitis B, which is known to contribute to HCC.31There remains debate as to whether the wAAV insertions are drivers of oncogenes or benign passengers.32 As far as hemophilia gene therapy goes, it really doesn’t matter if wAAV causes HCC or not. wAAV is not infused into patients for gene therapy, rAAV is. The question for hemophilia gene therapy is does rAAV cause HCC?
Unfortunately, the answer is yes, and here there is no debate.33 Pre-clinical Investigations of rAAV gene therapy date back over 2 decades. An early investigation of rAAV involved a mouse model of mucopolysaccharidosis VII (MPSVII). Neonatal mice with this disorder were successfully treated with rAAV. However, a small number developed HCC.34,35 Because MPSVII is fatal in mice, it was unclear if the rAAV was causing the HCC or the underlying disease was. A follow-up study included a normal control group. These neonatal mice also developed HCC following rAAV.34Investigation over the ensuing 2 decades showed that rAAV inserts and drives the RIAN locus, more specifically microRNA-341, in neonatal mice. This does not occur in adult mice. Humans do not have aRIAN ortholog, but dogs do.35 The combination of rAAV/neonatal mouse/RIAN locus are needed to induce HCC, or so it was thought.34
Numerous trials of rAAV gene therapy for a variety of genetic disorders have been carried out in adult mice without the development of HCC.36 One difference in neonatal and adult mice is that neonatal mice have more proliferating cells, notably, hepatocytes. A recent investigation of rAAV gene therapy tested if rAAV could induce HCC in adult mice with liver injury (proliferating hepatocytes).36 Investigators used either partial hepatectomy or fatty liver disease (high fat diet). Two rAAV vectors were used, one designed to drive the RIAN locus, and another vector control with a reporter gene. As expected, neither vector caused HCC above baseline in control adult mice. The rAAV vector designed to drive the RIAN locus caused HCC in both models of proliferating hepatocytes. Surprisingly, so did the control vector in adult mice with fatty liver disease (partial hepatectomy not tested).36 In addition, a clinical trial of rAAV for phenylketonuria was placed on hold due to the development of HCC in a mouse model.2Thus, rAAV can induce HCC in adult mice with proliferating hepatocytes. What about other animal models?
To date, HCC has not been observed in dogs or non-human primate rAAV gene therapy for hemophilia. However, clonally proliferating liver cells have been found.37 Nguyen et al. recently reported on a long term follow up of hemophilia A gene therapy in dogs using rAAV.37 Nine dogs were followed for up to 10 years. Liver samples were available in 6 of the dogs. rAAV integrations were found in 1741 sites, and between 1 and 130 cells per integration. Preferential integrations near oncogenes in clonally proliferating cells were seen. The dogs were otherwise healthy. The authors did not report on the presence or absence of fatty liver disease. A second long term study of 8 hemophilic A dogs receiving rAAV gene therapy did not find clonal proliferation in the liver.38 This manuscript did not report on rAAV integrations, and additional analysis may be forthcoming. Long term data for hemophilia B gene therapy in dogs is also available.39,40 Many canine gene therapy strategies for hemophilia B using rAAV targeted skeletal muscle and may not confer the same HCC risk as strategies targeting hepatocytes. There are no reports of HCC. However, detailed analysis for rAAV insertions and clonal proliferation have not been forthcoming.
One subject in a human trial of rAAV for hemophilia B developed HCC.35 This subject had risk factors for developing HCC including hepatitis B and C. His liver specimen has been extensively evaluated and did not reveal genetic changes related to rAAV carcinogenesis but did have genetic changes typically found in HCC.35 No other human subjects have been reported with HCC following gene therapy for hemophilia. A recent report of liver biopsy results from 5 subjects who participated in a hemophilia A gene therapy trial did not show evidence of HCC or clonal proliferation.41 Notably, 4 of 5 subjects had liver steatosis, which apparently was subclinical.
In summary, emergent data has shown that rAAV can induce HCC in adult mice with liver disease.36 Long term studies of small number (<15) of hemophilia A dogs have shown clonal proliferation in the liver.37,38 If rAAV causes HCC in a small proportion, or even 5-10% of dogs, it could easily be missed by the current canine studies. To date, no studies have looked at the risk of HCC development of dogs or non-human primates with fatty liver disease following hemophilia AAV gene therapy. No humans have developed HCC or clonal proliferation caused by rAAV. Since the latency for the development of human HCC following rAAV may be decades, the risk of human HCC following rAAV cannot currently be estimated.
Genome Integration
Another advantage of using rAAV as a vector is that it mostly remains episomal following insertion into a cell.42 Until recently, studies have suggested that 99+% of rAAV vector was episomal, and <1% integrates. Emergent data has shown a higher percentage of integrations. Up to 3% of rAAV may integrate into liver cells following gene therapy.42Most hemophilia gene therapy protocols infuse 1014 – 1015 viral particles and target 1011hepatocytes. Assuming a more conservative estimate of genome integrations of 0.1%, one could anticipate over 100 million integrations following gene therapy. Indeed, some experimental data confirms this notion. 42 Therefore, there are a massive number of integrations following rAAV gene therapy. Most are not intact vectors. With 1011 hepatocytes in a human adult, each with a genome of 3 X 109 base pairs, most integrations are likely to land in an inactive genetic region if they integrate randomly. However, the above referenced dog hemophilia gene therapy study suggests that integrations are not random, they tend to occur near active genes, including oncogenes.37 One driver integration near the wrong oncogene at the wrong time could start the cell toward clonal proliferation and eventually, over years or decades, overt cancer.
Other Cancers
In addition to the above-mentioned HCC following rAAV gene therapy for hemophilia, 3 additional cases of cancer have been reported in humans, and one in a dog.43-45 As with the HCC, the other cancers (tonsillar carcinoma, salivary gland carcinoma, leukemia) have been thoroughly investigated and shown not to be caused by rAAV.43-45 It is unknown if immunosuppression received during clinical trials may have contributed to oncogenesis. To date, several hundred subjects have participated in clinical trials involving rAAV for hemophilia with reported observation periods lasting several years. There is easily over 1000 person years of observation. Thus 4 reports of human cancer may not be unexpected. However, a proper epidemiological investigation seems indicated.
Unfolded Protein Response
The target cell for hemophilia gene therapy using rAAV is the hepatocyte. Although factor VIII is made in the liver, it is not made in the hepatocyte.46 Accordingly, hemophilia A gene therapy targets a cell that does not typically produce factor VIII. Factor IX is naturally made in the hepatocyte. Factor VIII is a large protein with complex folding. Misfolded factor VIII protein can lead to cellular toxicity via the unfolded protein response (UPR).47 This has been shown to occur when non-native cells are driven to express factor VIII. This occurs both in vitro (Chinese hamster ovary cells) and in vivo (mouse hepatocytes) for factor VIII.47,48 Factor IX expression is stable for years following rAAV gene therapy in animal models and humans.1,2 While factor VIII expression has been stable following rAAV gene therapy in animals, this has not been the case in humans when therapeutic levels are achieved. As reported above, several clinical trials of AAV gene therapy for hemophilia A have shown that hemostatic levels are not sustained.24,28 The etiology for the falling levels remains unclear. An immune response has been investigated and does not clearly seem to be the cause.35 UPR (Unfolded Protein Response) could provide another explanation. However, as above, liver biopsies from subjects following hemophilia A gene therapy failed to demonstrate evidence for UPR at the time of biopsies.41 Biopsies were performed 2.6-4.1 years following infusion of rAAV. So cellular toxicity/loss from UPR occurring prior to this would have been missed.
Although an immune response and UPR have been independently proposed as explanations for falling factor VIII expression following AAV gene therapy, an investigation by Butterfield et al. suggests that they may not be mutually exclusive.49 In addition, this study also suggested that translational shutdown (related to UPR and immune response) rather than loss of hepatocytes could lead to falling factor VIII expression following AAV gene therapy.
Concerns about UPR and HCC have also been raised. Kapelanski-Lamoureux et al. have investigated UPR and HCC risk in mice.50In their study, all mice fed a high fat diet following receipt of a B-domain deleted factor VIII gene therapy vector (non-AAV) via hydrodynamic tail vein injection developed liver tumors. This happened less so with a factor VIII variant vector less prone to misfolding and not at all with a control vector. This suggests that factor VIII misfolding in mice fed a high fight diet contributes to the development of HCC independent of viral vector integration.
Spinal Muscular Atrophy
One of the first rAAV gene therapy treatments to achieve regulatory approval was for the treatment of spinal muscular atrophy (SMA). This is a degenerative neuromuscular disorder that results in early death in those affected with severe (infantile) forms. rAAV gene therapy for this disorder has met with widespread success. 51 While affected infants treated with rAAV gene therapy are not normal, they are achieving developmental milestones with an extended lifespan. Like hemophilia rAAV trials, the principal toxicity seen during the SMA trials was mild liver inflammation that is controlled with steroids.52 Now that over 3000 infants have received gene therapy for this disorder, rare side effects not seen during the clinical trials are being observed. 52 At least 9 cases of thrombotic microangiography (TMA) have been reported in the medical literature following rAAV gene therapy for SMA, one of which was fatal.53 Thirty (two fatalities) cases of TMA are reported in the U.S (United States). Food and Drug Administration Adverse Event Reporting System (FAERS). 54 It is unknown if there is any overlap between cases reported in FAERS and the medical literature. The manufacturer of SMA rAAV has reported two cases of hepatotoxicity leading to fatalities. 56 FAERS reports 120 cases of hepatobiliary disorders and 8 fatalities. 54 One should interpret the FAERS data with caution, as the cause of death is not listed, only “Reactions”. Duplicate reports may also be present. Therefore, patients who died from causes unrelated to rAAV may be included in this database. Table 1 shows the number of reported relevant “Reactions” and deaths in the FAERS database for regulatory approved medications for SMA.54 54-58 A comparison of clinical trial results suggests that the event free survival rate is similar between Onasemnogene abeparvovec and Risdiplam, and higher than Nusinersen.58Post-marketing surveillance from the FAERS database also shows higher mortality reporting for Nusinersen. TMA and hepatobiliary disease following treatment for SMA seems to be relatively unique to Onasemnogene abeparvovec . Use of rAAV in other clinical trials has also resulted in hepatotoxicity related fatalities. 53 Because SMA and hemophilia are different diseases, different age groups were treated, and different doses used, similar toxicities may not occur in hemophilia patients following regulatory approval for rAAV gene therapy. However, the potential for rare, serious, and potentially fatal toxicities should be included in any risk/benefit calculation.
Dorsal Root Ganglion
Dorsal root ganglion (DRG) pathology has been commonly found in non-human primate gene therapy using rAAV.59 This has been found in a variety of vectors for a variety of diseases. It occurs more commonly with central nervous system administration but is also seen following intravenous administration of rAAV. Immunosuppression does not seem to ameliorate the toxicity. It is not seen with non-expressing vectors. 60 Accordingly, the toxicity seems related to recombinant protein expression. Fortunately, the toxicity is mild and short lived, and importantly, does not seem to cause obvious symptoms in the non-human primates. DRG toxicity has been reported in human trials of rAAV, but not for hemophilia.1,2,53,61 However, it does not appear that subjects enrolled in hemophilia rAAV gene therapy trials were carefully evaluated for DRG toxicity, or if they were, this data is not reported in published manuscripts.1,2,24,25