!DOCTYPE html SYSTEM “about:legacy-compat”>
Jukka H. Meurman (ed.)Translational Oral Health Researchdoi.org/10.1007/978-3-319-78205-8_7
7. A Critical Reason for Conducting Clinical Trials Is that Results with Humans Are Not Necessarily Predictable from Preclinical Studies
In the biomedical sciences, when developing any new therapy for all types of human diseases, the typical sequence pathway followed progresses from laboratory bench research (e.g., studies in a test tube or cell culture), to experiments in an animal model, and finally to human clinical trials. The most often considered and obvious reasons for the early-stage (phase I/II) clinical trials are to ensure adequate safety and efficacy before allowing widespread administration of the therapy. However, there is another, critically important reason, not often discussed, which is that results from human clinical trials often are not entirely predictable from the preceding preclinical studies. Herein, this important reason is addressed using examples from the developing field of gene therapy targeting three quite different clinical disorders: cystic fibrosis, severe combined immunodeficiency X1, and irradiation-induced salivary hypofunction.
Human studiesGene therapyTranslational researchAnimal modelsCystic fibrosisSCID X1Salivary hypofunction
The last 30 years has seen an explosion in the amount of biological knowledge gained. As someone who earned a Ph.D. in biochemistry in 1974 without any training in what has become known as molecular biology, this explosion of information is part of my personal experience. While I was in graduate school, the polymerase chain reaction (PCR), reverse transcriptase, restriction endonucleases, routine and easy DNA sequencing, and the like were just simply not part of scientists’ knowledge base or even in the general imagination of more than a rare few. At present these and many other molecular biological tools are mundanely used in all types of laboratories around the world. Furthermore, in many cases these tools have delivered substantial benefits related to the treatment of human disease.
I began to work in the field of gene therapy in 1991 shortly after the first paper was published demonstrating the feasibility of its use in humans . It was a heady time for those working in the field, as there was unbridled optimism that the methods available generally were straightforward and that various successes in animal models would soon translate to comparable successes in the clinic. However, that beneficial translation was much slower to be achieved than expected and serious problems also occurred (e.g., [2, 3]). Even today, gene therapy can be considered a biological treatment that is just beginning to show return on its promises .
While there are now several excellent examples of gene therapy benefiting patients, for me, during the time I have worked in this field, there have been three major lessons that will be evident throughout the discussion herein: (1) it is critically necessary to appreciate the many nuances of human physiology and pathology, (2) at the current time relatively little (only a small fraction) is known about molecular biology-nucleic acid biochemistry, and (3) there are significant limitations to most of the molecular tools now available to treat patients, even those applicable to patients with “straightforward” single gene defects. Most importantly, these lessons present an overarching and critical take home lesson for all individuals trying to develop novel therapies for humans—it is essential to study humans. Results from preclinical animal studies may not predict what will be observed in early-phase clinical trials. Simply put, in order to treat human patients better, it is necessary to conduct experiments on humans. This need exists despite (1) an incomplete knowledge base about the disease and the molecular or other biological therapies being performed and (2) a lack of ideal tools with which to perform the desired treatment.
The approach employed herein will be to give specific and relevant examples from three areas of gene therapy research that have transitioned from bench toward the bedside: cystic fibrosis (CF) , severe combined immunodeficiency X1 (SCID X1 ; “boy in the bubble” disease), and my own work on irradiation (IR)-induced salivary hypofunction . Each example will, to varying degrees, demonstrate the three lessons described above, as well as the bottom-line message of the critical importance of conducting, and learning from, human clinical trials.
7.2 Cystic Fibrosis
CF is a fairly common inherited disorder. While it affects many organs, its major and most life-threatening manifestations occur in the lung . CF results from a mutation in a single gene, which encodes a chloride transport protein termed the CF transmembrane conductance regulator (CFTR) . Mutations that impede CFTR’s physiological role lead to altered water movement and mucociliary dysfunction in lung epithelia. What results is an accumulation of a sticky mucous, which promotes bacterial infections and pulmonary inflammation . CF patients have a markedly shortened lifespan, typically <40 years, and a significantly reduced quality of life . Consequently, it is no surprise that CF was one of the earliest diseases targeted for gene therapy (e.g., see [6, 7]). As reviewed recently by Sondhi et al. , “eleven different mouse models of CF have been developed,” by knocking out the CFTR gene. However, most of these models have little to no lung disease nor do they develop the spontaneous lung inflammation characteristic of the human disease .
It is not difficult to transfer a gene into the lungs. For example, the human CFTR gene has been delivered directly by intratracheal instillation in many species, from mice to humans. This method provides a straightforward way to perform in vivo gene therapy to the lung, using either a viral or non-viral vector. It has been employed both for evaluating vector-directed expression in normal mice (e.g., ) and for use in “curing” a mouse model of CF (e.g., ). However, as we now know, these studies in mice had little value in predicting the success, or in this case lack of success, of this gene therapy approach in humans.
Larger animal models, e.g., pigs with knocked out or mutated CFTR genes, also have been developed [10, 11] and fortuitously exhibit lung pathology similar to that seen in humans . We know from other studies (e.g., ) that fairly efficient transfer and expression of the human CFTR gene in porcine lung epithelial cells can be accomplished in normal pigs using an aerosolized, helper-dependent (totally defective) adenoviral vector. Furthermore, it’s clear that this procedure was generally safe in pigs. However, it is not yet clear that such gene transfer maneuver will be effective in “curing” a pig exhibiting CF-like disease.
While gene delivery to the airways is reasonably efficient in normal human lungs, via intratracheal administration (e.g., ), in a CF lung clogged with sputum, as well as highly inflamed and infected with bacteria, it is quite inefficient . Indeed, efforts to develop strategies to remove the viscous sputum have been modestly successful in the laboratory; however, they have not yet been translated to useful clinical applicability. Sondhi et al.  point out that there are numerous challenges to achieving successful gene therapy for CF, via the commonly used clinical approach of intratracheal delivery (e.g., as used in bronchoscopy), including targeting the correct cell population, efficient gene expression in those cells, the normal turnover of lung epithelial cells, the vectors available for gene transfer, as well as the recollection that CF affects many other organs besides the lung. Thus, despite the existence of a huge research effort focusing on CF gene therapy for >25 years now, all that has been learned from these studies, and the initial success found with mouse models, is that the human pathology of CF and the limitations of existing gene transfer vectors have precluded any level of clinical gene therapy success for this condition .
7.3 SCID X1
SCID X1 is an X-linked inherited disorder in which patients have a mutation in the γc cytokine receptor subunit gene. Since the γc chain is common to many different hematopoietic cytokine receptors, e.g., interleukin-2, interleukin-4, interleukin-7, interleukin-9, and interleukin-15 receptors, its mutation results in a block in the development of T lymphocytes and natural killer (NK) cells . Also, as a consequence of this mutation, SCID X1 patients have dysfunctional B lymphocytes. Since SCID X1 patients represent 30–40% of all SCID patients, it also is not surprising that this single gene disorder became an early target for developing a gene therapy . In fact, the first report of any successful gene therapy in humans was for SCID X1 .
After years of preclinical studies , Cavazzana-Calvo et al.  developed the methods to perform what is called ex vivo gene therapy . With their approach they took T lymphocytes from two SCID X1 patients and inserted the correct γc receptor subunit gene into the DNA of those cells using a defective, and thus presumably safe, gamma retrovirus (Moloney Murine Leukemia Virus, MoMLV). Afterward, they amplified the T lymphocytes in cell culture in vitro and then infused a large number of the genetically modified cells back into the corresponding donor SCID X1 patients . The original report in 2000 was greeted as essentially a miracle for the two, treated patients, each of whom had failed a bone marrow replacement therapy and, thus, were destined to spend their greatly abbreviated lives within the confines of an airtight bubble in a hospital room. After the treatment , both patients had normal levels of T, NK, and B lymphocytes, all of which also functioned normally. The two patients were able to leave the hospital and live the lives of normal children, e.g., play in a park, go to school, etc. .
In aggregate, this research group treated a total of nine SCID X1 patients using the MoMLV-mediated γc cytokine receptor subunit gene therapy strategy , and the therapy was successful in eight of these children. Of note, all additional patients, beyond the initial two, were apparently treated within in a few years of the 2000 publication. The reason for the limited number of patients being treated, and for the narrow time frame when the treatment occurred, was because in 2003 this group reported a very serious adverse event in four of the treated patients—the development of acute lymphoblastic leukemia [18, 19]. Three of those four patients were treated successfully for the leukemia, while the fourth died. Overall, seven of the nine originally treated SCID X1 patients, including the three surviving post-leukemia patients, remained alive and experienced a successful immune cell reconstitution .
Thus, despite the significant risk of developing an acute leukemia , this gene therapy represented a reasonable treatment risk for the children (and their parents) with SCID X1; there literally was no other viable alternative. However, these investigators, and many of their colleagues around the world asked what led to the serious adverse events and what, if anything, could be done to eliminate or at least minimize the risk of their future occurrence . The resulting studies were important not just for the SCID X1 patients but also for patients with many other disorders, hematopoietic and non-hematopoietic, for which an ex vivo gene therapy employed the MoMLV vector. It turned out that with the lymphocytes of the children who developed leukemia, the MoMLV vector had inserted the γc cytokine receptor subunit gene into a site in chromosome 11 within what is called the LMO-2 gene locus . This locus codes for the expression of a proto-oncogene and the insertion of the foreign gene led to the aberrant expression of LMO-2. Aberrant expression of this gene previously was known to lead to acute lymphoblastic leukemia , so the reason for the serious adverse events became understood. The efforts to minimize such future occurrences led to studying retroviral integration patterns and, subsequently, the development of much better and safer retroviral vectors . Indeed, at the time of this writing (May 2017), the European Union has given marketing approval for the treatment of another SCID (based on a deficiency of the enzyme adenosine deaminase; ), which utilizes the same general ex vivo gene therapy approach employed to treat SCID X1 patients but using a new, much safer, generation of retroviral vector.
7.4 Irradiation-Induced Salivary Hypofunction
As a group, head and neck cancers are the sixth most common malignancy worldwide, with ~500,000 cases occurring each year. Most patients are treated, at least in part, with therapeutic IR, which can damage normal tissues falling within the IR field, including the salivary glands. The end result of such salivary gland damage, for many patients, is markedly decreased saliva output and, consequently, dysphagia, a high risk for aspiration, increased oral infections (e.g., candida, caries), decreased oral mucosal wound healing, and considerable pain and discomfort. These patients, not surprisingly, experience a markedly reduced quality of life .
While there have been great advances in methods to deliver therapeutic IR and limit normal tissue damage, e.g., intensity-modulated radiation therapy, known as IMRT [22, 23], IR-damaged salivary glands and its sequelae still remain a significant clinical problem. There is roughly a 65% 5-year survival rate for head and neck cancer patients , which means many former patients are alive, thankfully, but suffering from IR-induced salivary hypofunction. Additionally, at present advanced treatment modalities, such as IMRT, are typically available at major medical centers in highly developed countries, which means many people worldwide are still treated with conventional radiation therapy and continue to be at risk for normal tissue damage.
In 1991, my research group began the long process of trying to use gene transfer technology to provide a “repair” of IR-damaged salivary glands that in turn would lead to patients with more saliva and improve their objective problems and symptomatic complaints. As indicated in the first paragraph of the Introduction, I had no background in molecular biology to support me on this endeavor. However, thanks to the help of a former postdoctoral mentor (Ronald G. Crystal, e.g., see ), and three outstanding postdoctoral fellows working in our group in the early days of this project (Brian C. O’Connell, Christine Delporte, Hideaki Kagami), we were able to make considerable progress in applying gene transfer to salivary glands and demonstrate proof of concept in a rat model for the “repair” of IR damage to salivary glands (e.g., see [26–28]). The general strategy that we employed used a first-generation serotype 5, adenoviral (Ad5) vector to deliver the cDNA encoding human aquaporin-1 (hAQP1), the archetypal water channel protein, via the cannulated main excretory ducts of salivary glands, a not dissimilar approach from the intratracheal administration of gene transfer vectors to the lung as described above. The vector created was termed AdhAQP1 . Later studies with wonderful colleagues in my research group (Changyu Zheng, Corinne Goldsmith) and a collaboration with an excellent former postdoctoral fellow (Songlin Wang) demonstrated the safety of AdhAQP1 and the delivery approach used (e.g., see ), as well as its extension to a large animal model of IR-induced salivary gland damage (e.g., see ).
A good thing about first-generation Ad5 vectors is that they are relatively easy to create and use, which was advantageous for a group with minimal molecular biological and virological experience such as ours in the 1990s. Another good feature of these Ad5 vectors is that they lead to high levels of functional gene transfer in the targeted tissues. However, there is an important negative feature of such Ad5 vectors: they can elicit potent immune responses (innate, cellular, and humoral) after administration. Because of the latter, first-generation Ad5 vectors, in a wide range of animal models and tissues, and in many clinical studies targeting non-salivary gland tissue, yield only transient expression of the delivered transgene, typically for no more than a week or two, with a peak response on days 2 or 3. Indeed, our studies using Ad5 vectors, including AdhAQP1, with salivary glands of mice, rats, miniature pigs, and macaques demonstrated a similar, short time course of transgene expression.
Despite the above-described shortcomings of first-generation Ad5 vectors, based on the results of our preclinical studies with AdhAQP1 , and after extensive toxicology and biodistribution studies , we developed a protocol for a phase I/II clinical trial to test the vector in IR-damaged parotid glands of human subjects (www.clinicaltrials.gov/ct/show/NCT00372320?order=). The subjects enrolled exhibited grade 2 or 3 damage to their parotid glands according to the criteria of the Radiation Therapy Oncology Group , i.e., they had some epithelial tissue remaining in their parotid glands, but were not responsive to conventional pharmacological treatment with Salagen or Evoxac. Interestingly, because of the anticipated short expression time from AdhAQP1 as observed in our preclinical animal studies, the original purpose of the clinical trial (www.clinicaltrials.gov/ct/show/NCT00372320?order=) was considered essentially to be a proof of concept. We fully expected that hAQP1 gene transfer to the IR-damaged parotid glands would lead to increased fluid secretion for at most 2 weeks, with a peak response on days 2 or 3. Consequently, we thought the subjects in the trial in the event of positive results unlikely would experience long-term benefit (osp.od.nih.gov/sites/default/files/RAC_minutes_12-05.pdf).
Eleven subjects were treated with AdhAQP1 in this first in human clinical trial . The initial findings of that trial, through day 42 post-AdhAQP1 delivery, identified 5 of 11 treated subjects as responding positively to the gene therapy. The positive response was defined as both increased salivary flow from the targeted parotid gland, as well as the improvement of two key symptomatic benefits (a subject’s perception of the amount of saliva, and the level of dryness, in their mouth). Interestingly, the peak increases of parotid salivary flow rates for all responder subjects were observed at much later times than seen in animal models, from 7–42 days after AdhAQP1 delivery. The originally approved clinical protocol required subjects to be followed for 360 days. However, the protocol was amended because the first responder subject, whose initial peak response to gene transfer occurred on day 7, exhibited a second, later elevation in parotid salivary flow rate, on days 180 and 360, well above his baseline. Accordingly, the original protocol was modified to allow the evaluation of all responders to AdhAQP1 for two additional times, at least 1 and 2 years following their completion of the original 360-day protocol .
The AdhAQP1 clinical trial was a human study involving gene transfer to a salivary gland, and the results were unexpected based on (1) all previous clinical trials in humans with Ad5 vectors in other tissues and (2) our own studies delivering Ad5 vectors to the salivary glands of multiple animal models. Not only did the initial results show peak expression times much later than seen previously, but all five responder subjects after long-term follow-up displayed substantially elevated levels of parotid saliva flow 3–4.7 years after the AdhAQP1 administration. Furthermore, most subjects experienced relief from two key xerostomic symptoms for at least 2 years after treatment .
We think there were two key reasons for this unusual result. First, it is widely thought that the immune response to a first-generation Ad5 vector delivery leads to complete removal of the vector from the targeted tissue. However, we have shown that is not the case in rat salivary glands . For example, after a dose of 109 vector particles/rat submandibular gland, 0.1% of the delivered dose was still present in gland tissue 6–12 months later . Secondly, we recently showed that the human cytomegalovirus promoter used in AdhAQP1 is substantially methylated in rodent salivary glands, a modification that inhibits its ability to function as a promoter and lead to transgene expression. This does not occur in human cells . The combined results of both Zheng et al. [35, 36] studies imply that the AdhAQP1 vector will be (1) present in human parotid glands long after its administration and (2) able to direct the expression of functional hAQP1 . Thus, the results of the AdhAQP1 clinical trial could not have been clearly predicted from preclinical animal studies, including those with miniature pigs and nonhuman primates.
7.5 Key Differences Between Human and Animal Models
Aside from the obvious physical differences between humans and the animal models used in preclinical studies, there are some key biological differences that doubtless influence responses to viral vector-mediated gene transfer such as described above. Of particular note are immunological and genetic differences. This point can be clearly appreciated when comparing humans with mice, since the latter are the most commonly employed preclinical animal models of disease treatment. However, as noted by Davis , while mice have been extremely useful for developing an understanding of basic immunology, they have been less helpful in understanding human disease. Following that perspective, by studying human immunology directly and not just extrapolating from results obtained with inbred strains of mice, Su et al. have found numerous differences in T-memory cell mechanisms . Similarly, Benitez et al. have shown key differences in the dynamics and mechanisms for nonmemory B cells between mice and humans . Thus, it would not be surprising to find that immunological responses to a viral or non-viral gene transfer vector could be markedly different when studied in mouse models from that found in actual clinical trial subjects .
A comparable pattern of differences emerges from genetic studies of mice and humans. For example, inbred strains of laboratory mice share most of their protein-coding genes with humans . However, as reported by Yue et al. , “the Mouse ENCODE Consortium mapped transcription, DNase I hypersensitivity, transcription factor binding, chromatin modifications and replication domains throughout the mouse genome in diverse cell and tissue types,” and compared those results with data from humans. They found considerable “divergence of sequences involved in transcriptional regulation, chromatin state, and higher order chromatin organization” . Many other studies lead to the same general conclusion, i.e., that there are quite important differences related to genetic regulation between mice and humans, such as in the patterns of gene expression seen during development  or RNA expression profiles for both coding and noncoding regions , which the authors state, “likely reflects fundamental physiological differences…”.
While it certainly is not surprising that differences exist between humans and mice in key aspects of immunological and genetic regulation, similarities in these biological processes do exist between the species. What then becomes important for investigators is to understand which mechanisms are similar between humans and mice (or any other preclinical animal disease model) and which are different, so that preclinical studies aimed at developing novel therapies can be best designed for likely translation into effective clinical human treatments. For example, the work of Godec et al.  and Li et al.  have generated data sets of genetic patterns of immune reactivity in humans and mice for different conditions, e.g., sepsis and inflammation, that show both differences and similarities. These then can be used by investigators to target specific genes or mechanistic pathways that are best conserved between the species and, thus, likely to be most valuable in novel therapeutics discovery. The same is true when looking at specific genetic components of diseases, e.g., when trying to develop the best animal models for use as models of human neurological diseases  and human DNA repair ; there are both similarities and differences between humans and different animal models.
As noted at the outset of this chapter, results from human clinical trials often are not as expected from the preceding preclinical animal studies. Animal experiments are still necessary in the process of developing new therapies. However, there is a critical need to study humans not only to determine if therapies developed in animals are safe and effective but also to appreciate physiological differences, manifested in genetic, immunological, and other characteristics, that will permit the development of better therapies in the future. Human clinical trials are experiments, and in order to move forward, we need to understand human biological mechanisms with as much detail as we can possibly garner from experiments in mice and other animal models.