Prevalence and utility of pharmacokinetic data in preclinical studies of mRNA cancer vaccines

Abstract

In this brief report, we provide insights into current practices in preclinical messenger RNA (mRNA) cancer vaccine characterization. To enable a more automated and thorough survey of mRNA cancer vaccine data in the literature, we implemented natural language processing to mine abstracts from PubMed followed by annotation of the selected articles. Through this analysis we identified a gap in the literature wherein pharmacokinetic (PK) characterization is not reported in mRNA cancer vaccine‐focused articles. As a result, the field is unable to evaluate and discuss cross‐study PK and pharmacodynamic (PD) relationships nor the translation of these relationships from preclinical species to humans. As the field of mRNA cancer vaccines is rapidly evolving, there is value in expanding our understanding of preclinical PK/PD relationships and how they relate to PK/PD in humans.

INTRODUCTION

The advent of new drug modalities prompts translational scientists to consider what information is necessary to enable a therapy to transition to, and be successful in, a clinical setting. The pharmaceutical industry has witnessed many transformative innovations in the last century that have facilitated a better understanding of both drug development and translation. Each novel drug modality comes with new challenges, and it is important to consider what characterization is most useful for understanding and predicting drug behavior. In this brief report, we focused on pharmacokinetic (PK) and pharmacodynamic (PD) relationships and their utility in the burgeoning field of messenger RNA (mRNA) cancer vaccines. mRNA based therapeutics have demonstrated their ability to provide clinical impact for prophylactic infectious disease vaccines, most notably coronavirus disease 2019 (COVID‐19) vaccines, and hold much promise for therapeutic cancer vaccine applications. ¹

PK and PD are measures of the effect of the body upon a drug and, in turn, the effect of the drug upon the body. PK characterization typically involves measuring the level of an administered drug in the body across relevant timepoints and tissues (most often plasma or blood). These measures are often paired with measurements of PD, which can include biomarkers and other efficacy markers, to allow for the exploration and elucidation of PK/PD relationships.

For drug modalities, such as small molecules and antibodies, characterization of PK/PD relationships is a key component of the drug development process. These data can help inform clinical study design as well as help address questions from global regulators and study teams with respect to dose amount, frequency, and relationships with safety and efficacy. Notably, whereas elucidation of PK/PD is routinely incorporated into preclinical and clinical translational strategies for small molecule and antibody drugs, ² exactly how PK/PD characterization is incorporated into the drug development process for novel modalities is less clear. The US Food and Drug Administration (FDA) guidance for therapeutic cancer vaccines does not mention the evaluation of biodistribution or PK and recommends that the kinetics of the immune response can be used to provide insights into dose and dose frequency selection. ³

In this regard, some newer modalities, such as mRNA vaccines, have evolved from fields that have historically relied less upon PK/PD during their drug development process. Although mRNA vaccines have benefited from developments across a range of fields, their history is deeply rooted in infectious disease. ⁴ In the 1800s, when vaccinology was in its infancy, scientists were unable to measure PK or even closely control the dose that was administered. Similarly, PD measurements were not performed, as scientists did not have the tools nor the understanding of how to measure a patient's immune response. Although vaccinology has advanced immensely in the years since, it is helpful to remember its trajectory as we consider what assessments should be performed when developing and evaluating a new vaccine.

Since their emergence, mRNA vaccines have proven to not be as straightforward as other drugs, such as protein therapeutics, with respect to PK determination. For example, because a monoclonal antibody is often administered directly into the bloodstream as a single active ingredient, PK determination is primarily focused on measuring the antibody concentration in the blood (Figure 1a). Unlike a monoclonal antibody, mRNA vaccines often include mRNA cargo that is packaged in a specific carrier (e.g., liposome or polymer nanoparticle). After administration, the carrier (with its mRNA cargo) can travel as an intact entity, or it may become unstable, and the components may be distributed and eliminated separately, therefore each of these forms are candidates for PK measurements.

FIGURE 1.

PK considerations for mRNA cancer vaccines. (a) Following intravenous administration of a monoclonal antibody, the most common modality for cancer immunotherapies, the drug distributes into peripheral tissues from the vasculature. The antibody has its desired effect upon binding to its antigen. This effect may take the form of several different mechanisms, from directly activating a pathway to recruiting innate cells through the Fc domain. In contrast, (b) depicts the path of an mRNA cancer vaccine, which goes through many steps following drug administration to achieve the final effect of activating T cell that will attack the tumor. After administration, the cancer vaccine may remain intact or have some dissociation of the mRNA and carrier, all of which are able to distribute to tissues. Once in the tissue, the vaccine will be taken up by innate immune cells, where the mRNA must be translated into a protein, processed, and finally presented on the surface of antigen presenting cells. This peptide presentation, along with immune stimulation signals, will activate antigen‐specific T cells that are able to attack the tumor. The additional complexity of the mRNA cancer vaccine is highlighted with a gray background for emphasis. mAb, monoclonal antibody; mRNA, messenger RNA; PK, pharmacokinetic. Created with biorender.com.

Beyond consideration of the administered drug material, it is critical to think of the active drug in its final form. In the case of an mRNA cancer vaccine, a series of steps must successfully transpire: (1) the vaccine must transfect cells, (2) the mRNA must be translated, (3) the resultant cancer‐associated peptides must be presented on innate immune cells such as dendritic cells, and (4) these innate cells must activate cancer antigen‐specific T cells that can target and kill the antigen‐expressing tumor cells (Figure 1b). Each player in this process is a candidate for further characterization of changes in concentration and dynamics over time, which adds layers of complexity when considering the most relevant PK to measure.

In recent years, the extensive utilization of natural language processing (NLP) for literature mining has showcased its capability to automate the extraction of large volumes of textual data, empowering researchers to uncover insights at an unprecedented scale and speed. Leonardelli et al. exemplifies the successful application of NLP to identify key vaccine features using literature mining while reducing manual effort and increasing efficiency. ⁵ Here, we surveyed current practices in preclinical mRNA cancer vaccine characterization by implementing NLP to mine abstracts from PubMed. We then annotated the identified articles for key features of their study design and the availability of PK data. As our focus was understanding the availability of data that would enable translation from the preclinical to the clinical setting, we only included publications from organizations and institutions that have sponsored mRNA cancer vaccine clinical trials. Additionally, we chose to narrow our analysis to focus only on publications that included in vivo studies which administered vaccines encoding cancer antigens. Thus, publications that focused on platform development and used a tool antigen rather than cancer antigens were excluded. Although characterization of non‐cancer vaccine platforms is critical for the development of mRNA vaccine technology, these studies did not allow for linking the measured PK and biodistribution with the biological effects of interest, and thus were not included in our analysis. For a complete list of all publications that were annotated, please see Table S1.

METHODS

Described here is a two‐phase framework for identifying abstracts from PubMed and extracting scientific evidence from full‐text articles (Figure 2). Phase 1 involved article retrieval from PubMed wherein the text mining tool I2E (Linguamatics) and NLP technologies were utilized to identify abstracts that mention specific mRNA cancer vaccines. ⁶ Phase 2 involved manual review whereby articles were selected and manually annotated with our in‐house tool, Highlighting Annotation Web Kit (HAWK). For additional details, please see Appendix S1.

FIGURE 2.

Natural language processing (NLP) framework for mining data from PubMed abstracts and full text articles. Articles were retrieved from PubMed wherein the text mining tool I2E (Linguamatics) and NLP technologies were utilized to identify abstracts that followed our specified inclusion and exclusion criteria. Several removal steps, both automated and manual, followed. Last, manual annotation with our in‐house tool Highlighting Annotation Web Kit (HAWK) was completed on the selected articles by the study team.

RESULTS

Our two‐phase framework for identifying abstracts from PubMed and extracting scientific evidence from full‐text articles allowed for a more qualitative, expedited, and automated process of reviewing the literature, resulting in a significant reduction in the time needed to manually extract the necessary data by reading through each document. This methodology also provided clear inclusion and exclusion criteria for visibility and interpretability of our findings in a consistent manner. We were able to proceed with our review of the selected literature having captured all relevant articles that met our prespecified criteria.

Of the six preclinical articles included in our analysis, only one included in vivo PK characterization as well as data on tumor treatment efficacy. ⁷ This article included a study measuring the protein expression kinetics of luciferase‐encoding mRNA (not a cancer antigen encoding mRNA) to compare the expression of different vaccine formulations. Whereas the mice in this study were naive, non‐tumor bearing mice, the authors also conducted a separate study wherein they explored the response to a cancer vaccine in tumor‐bearing mice. No articles included in vivo PK characterization of mRNA molecules or the carriers themselves, and none measured protein expression from a vaccine encoding cancer antigens. Similarly, of the six articles describing the results of clinical trials, none reported PK.

It is important to note that even in the aforementioned article, which showed protein expression in the same paper as tumor efficacy, PK and PD measurements were not conducted with the same mRNA cargo, which is in contrast to PK studies for monoclonal antibodies or other biologics where the molecules administered in the preclinical studies are either the final clinical product or a close surrogate.

Another useful measure is the evaluation of dose–response. This is simpler than PK characterization, as it only requires the dosing information to quantify drug concentration. However, the inclusion of several doses in an efficacy study can make these studies challenging to execute and requires a greater degree of animal euthanization. As such, only one article included a dose–response study where more than one dose level was administered. ⁸ In this study, the mRNA vaccine was used to expand CAR‐T cells, and the dose–response study allowed them to see a potential saturation of the expansion at doses above 10 mg. Although their methods are not translatable to the expansion of endogenous T cells, one can envision performing a similar study while looking at the expansion of T cells specific to the antigens encoded by an mRNA cancer vaccine.

DISCUSSION

In conclusion, we found that it is not common practice to characterize PK/PD measurements in published preclinical mRNA cancer vaccine studies. The published studies focused only on the extent and kinetics of the immune response along with tumor efficacy. A broader search of the literature demonstrated that PK or biodistribution were more commonly incorporated in studies during the “platform development” stage of the mRNA carriers; however, this does not allow for characterization of PK/PD relationships, as there is no relevant PD to measure without cancer vaccine cargo.

Considering that there has not been a clinically approved mRNA cancer vaccine to date, it is clear that the field still has much more to learn about this modality to enable it to reach its full potential. Addressing this major gap in the availability of data in published studies may be one way to increase learning. Additionally, extracting and annotating other types of documents, such as summary basis of approvals, regulatory guidances, and meeting abstracts, using HAWK could provide further learnings and relevant information; however, given there is not a way to systematically download these types of documents, accessing and extracting these data for meta‐analysis is more difficult. Furthermore, although none to minimal PK/PD data were found in the published literature with respect to mRNA cancer vaccines, it may be useful to investigate dose response relationships for mRNA based COVID‐19 vaccines for any shared and translatable learnings.

Evaluating PK/PD relationships during mRNA vaccine drug development process could provide additional insights into the mechanisms of mRNA cancer vaccines. mRNA PK/PD should be evaluated preclinically across a wide range of dose levels. It is also recommended to evaluate mRNA disposition across multiple clinically relevant animal models. These data could then be combined with other measurements (e.g., immune response stimulation/dynamics) to translate these results from preclinical studies to clinical settings. In addition, empirical methodologies and mechanistic modeling approaches should also be used to better characterize the PK/PD of mRNA cancer vaccines. Examples of these approaches include interspecies allometric scaling of PK/PD parameters (e.g., mRNA clearance), compartmental modeling, and physiologically based PK modeling.

Alternatively, incorporating PK/PD measurements may reveal that PK is not a useful measure for development of mRNA cancer vaccines or, rather, that other measurements should be prioritized to help with pharmacological predictions. Coming to a data‐driven conclusion to not include PK/PD measurements in the standard assessment of mRNA cancer vaccines would be an important finding and warrants an initial investment in these types of studies. Importantly, increasing the availability of data in the published literature will help researchers to better understand the nature of how mRNA vaccines interact with the immune system to fight tumors. Finally, it is important that we take the lessons from the broader pharmacology field and intentionally decide what types of characterization will be most valuable as we improve our ability to develop new drugs, like mRNA cancer vaccines, and better enable collaboration with other drug developers and regulators.

Overall, although this work focused on published literature available in PubMed and annotated a focused set of PK/PD‐related questions, this is the first paper, to our knowledge, which has surveyed the availability of PK/PD data for mRNA cancer vaccines using semi‐automated approaches, such as text mining tools and NLP.

AUTHOR CONTRIBUTIONS

All authors wrote the manuscript, designed the research, performed the research, and analyzed the data.

FUNDING INFORMATION

This work was funded by Roche/Genentech, Inc.

CONFLICT OF INTEREST STATEMENT

All authors are employees and stockholders of Roche/Genentech, Inc.

Supporting information

Appendix S1

Roth GA, Vora B, Kim C, Wu M, Kuruvilla D. Prevalence and utility of pharmacokinetic data in preclinical studies of mRNA cancer vaccines. Clin Transl Sci. 2023;16:1554‐1558. doi: 10.1111/cts.13586