Researchers from the University of Texas have used data-driven approaches to identify a possible explanation for why some cancer drugs that are effective in PDX models may be ineffective at killing tumour cells in human trials.
Patient-derived xenograft (PDX) models
Preclinical evaluation using in vivo and in vitro platforms is essential for developing new cancer treatments. Patient-derived xenograft (PDX) models are an important in vivo platform in drug development. They are developed by implanting human tumour tissues into immune-deficient mice. PDX models have been considered a useful representation of the in vivo microenvironment for tumour growth. As of July 2020, over 4,031 PDX models were deposited in the PDX finder, and at least 19,242 publications related to mouse models of cancer have been deposited in the Mouse Tumor Biology Database.
Despite their critical role in cancer drug development, PDX models also face some challenges. Viral infections have been reported in PDX tumours and PDX-derived cell lines. However, not much is known about the extent of viral infections and how they impact PDX systems at the genome-scale.
RNA-seq analysis of PDX models
Next-generation sequencing has provided the opportunity to evaluate viral infection of PDX tumours and their impact on the PDX genome. RNA-seq data, generated from PDX tumours, capture expressed RNA from PDX tumours and viruses that could infect murine stromal and PDX tumour cells. In typical RNA-seq data analysis, sequencing reads are mapped to the human genome to evaluate expression levels of human genes of genome DNA variations. Unfortunately, unmapped reads (or the ‘dark matter’) of the sequencing data, which most likely includes mouse stromal cells and viral sequence reads, are disregarded.
By analysing these disregarded reads with human sequences, researchers can quantify the levels of gene expression for PDX tumours, mouse stromal cells and murine viruses. This will allow assessment of murine viral infection in PDX and its impact on the transcription profile of PDX tumours. Furthermore, RNA-seq data can be used to detect any viral sequences that have integrated into the PDX tumour genome. Thereby facilitating evaluation of whether the integrity of the PDX tumour is compromised by the viral infection.
The researchers behind this study compared the ‘dark matter’ of RNA-seq data from 184 datasets generated from PDX tumours and single-cell tumour cells to corresponding primary tumours and cells directly obtained from patients with no exposure to murine viruses. This data-driven approach assessed the sequences that had non-human origin and characterised the landscape of murine viruses in PDX models.
Data-driven approach to understanding the effect of murine viruses in PDX models
The researchers used sequencing data from different sources, including:
- PDX tumours or PDX
- Primary tumours – tumour explants directly obtained from patients without any type of culture or treatment who had not been exposed to murine viruses.
- PDX cell lines, which are cell lines derived from the PDX tumour.
- Primary cell cultures – cell lines derived from patient tumour explants with not exposure to murine viruses.
- Cell lines, which are lab-cultured cell lines obtained from the American Tissue Culture Collection primarily.
Raw reads for RNA-seq analyses were obtained from the NCBI Sequence Read Archive (SRA). The researchers then trimmed the raw reads to remove adapter sequences, ambiguous nucleotides and potential contamination. The reads were then mapped to the human and mouse reference genomes using STAR or Bowtie2. Viral sequences were detected with the pathogen discovery program READSCAN. Reads were considered viral if they had at least 10% coverage of the reference virus sequences. Overall, the researchers collected the data used in this project from 184 different experiments.
Results
By analysing sequence reads, the researchers showed that there is mouse stromal cell contamination in PDX tumours. They predicted that it is probable that some of the murine viruses they detected were from these cells. However, the researchers found a lack of correlation between the number of mouse stromal cells and murine virus load. They identified an extensive presence of chimeric sequence reads with high read depth containing both viral and human genome sequences. As well as, viral sequences in single-cell sequencing data from PDX tumour cells without any indication of mouse sequence contamination. This suggests that human tumour cells may also be infected by murine viruses and support viral replication.
In addition to that, the researchers also observed strong correlations between murine virus loads and the expression of many important human genes or pathways that are crucial in cancer research and treatment. For example, in PDX tumours with a high viral load, the researchers observed downregulation of immune genes. This included CD80, which plays an essential role in immune therapy deployment. Moreover, if this downregulation happens in lymphocytes within PDX tumours, it could result in suppression of PD-L1 receptor expression. This would suppress restriction of T cell activation, leading to high activity of T lymphocytes in viral-infected PDX models.
A direct outcome of this would be successful immune therapy that would not work on virus-free tumours in patients, if the observed gene expression change is related to murine viruses in the PDX environment. This is consistent with the finding that some PDL1 inhibitors, such as durvalumab, suppressed tumour growth in human tumour xenograft models but failed in clinical trials.
What does this mean for cancer drug drug development
The findings of this study indicate a possible way of recreating the high virus load tumour environment in cancer patients by oncolytic or other viruses, since the researchers observed some shared gene expression changes in tumour cells after infection and in PDX models infected by the murine viruses. If such an environment causes patient tumour cells to have similar gene expression changes as observed in PDX tumours, it may modify tumour sensitivity to drugs. Subsequently, this could help to recover some drug leads that worked well in PDX models, but not in clinical trials.
Summary
This study identified extensive presence of murine viruses and a strong association between virus load and gene expression changes. Comprehensive evaluation of the impact of viral infection on PDX should occur to fully understand their effect on drug development. To prevent viral infection and transmission animal models should be regularly tested for infection, which could have a profound impact on the success of cancer drug development using PDX models.
Image Credit: FreePik