Gene therapy and engineering is a promising area of research for the treatment of various genetic diseases. Therapy involves manipulating stem cells and transplanting them into recipients allowing the modified stem cells to differentiate into blood cells of all lineages. Understanding hematopoiesis, or the differentiation of blood cells from stem cells, is critical to improving gene therapy. One way to understand the development of blood cells is to track genetically engineered clones based on integration of the virus used to modify the stem cells. Currently, there are two methods of studying hematopoiesis following modified stem cell transplant: integration site analysis (ISA) and DNA barcoded sequencing (DBS). Both methods rely on tracking clones overtime but do so in very different ways. ISA identifies the integration site of the virus by shearing genomic DNA, amplifying outward from the viral sequence into the surrounding genomic sequence. The sequencing results are then aligned to a reference human genome. This method requires extensive manipulation of the cells and DNA, alignment requires a well-defined reference genome, and sequence variability can make tracking clones over time difficult. DBS uses sequencing of unique DNA barcodes within the viral construct, requiring much less cell manipulation. However, DBS is limited by the number of unique barcodes within the viral library and is not currently approved for use in humans. Dr. Adair and Dr. Kiem (Clinical Research Division) wanted to better understand which of these methods was more reliable for tracking modified cell clones following stem cell transplant. The results of their research were recently published in Molecular Therapy Methods & Clinical Development.
The authors used a non-human primate model to best recapitulate human hematopoiesis. Additionally, the authors developed a viral system that could be amplified using both methods to test them within the same samples. Upon sequencing the plasmid library prior to viral prep, the authors recovered over 1 million unique barcodes. When compared to sequencing the viral RNA, the barcodes largely overlapped indicating no barcoding bias during viral production. The authors received 14-16 blood or bone marrow draws per animal, which were further sorted into distinct blood cell types, for a total of 102 genomic DNA samples over a 2-year period. When comparing head to head, DBS led to higher sequence read numbers than ISA, which was expected - ISA requires much more DNA manipulation leading to potential losses. Additionally, the quality of the reads was significantly different; over 80% of DBS reads could be mapped to a barcode, while less than 20% of ISA reads mapped to an unambiguous integration site. One potential issue with DBS is a high rate of false positives due to the random nature of barcode sequences. However, nearly all barcodes found in vivo were found in either the plasmid or viral sequence library. Additionally, the authors analyzed barcodes over time and found that only 289 out of 19,853 barcodes (animal 1) and 356 out of 41,935 barcodes (animal 2) were not found in additional samples from other timepoints. This high recapture of barcodes indicated a low frequency of false positives in barcoded samples.
To further compare the methods, the authors designed primers for 14 of the top 26 integration sites identified by ISA. Using these primers, they amplified the integrated virus and sequenced the barcodes. Barcodes identified from these top integration sites identified by ISA matched barcodes identified by DSB within the top 37 most abundant clones, indicating each method was able to identify the top clones. The authors wanted to calculate the efficiency of capturing a specific clone with either method. They compared the number of barcodes or integration sites captured from sorted modified cells marked with GFP compared to unmanipulated samples where the percent modified was analyzed by expression of GFP. They determined that the efficiency of DBS was approximately 45% while ISA was less than 10%. Furthermore, due to low number of sequence reads, ISA led to much more inconsistent clonotype tracking, even when analyzing the most abundant clones over time. DBS showed that these abundant clones were quite stable over time and showed that some early clones were clear stem cells from which multiple blood cell lineages arose over time. ISA might lead to incorrect assumptions about clones being non-stable over time, while DBS shows this is not true. This direct comparison of methods shows that DBS is more consistent, higher efficiency, and requires much less cell manipulation than ISA, which may only offer limited data on clonal dynamics post-transplant.
This study was supported by the National Institutes of Health, the Cuyamaca Foundation, the Hartwell Foundation, the Bill & Melinda Gates Foundation, the Laurie Kraus Lacob Faculty Scholar in Pediatric Translational Research, the Amon Carter Foundation, the NIDDK Cooperative Center of Excellence in Hematology and Fred Hutch.
UW/Fred Hutch Cancer Consortium members Jennifer Adair and Hans-Peter Kiem contributed to this work.
Jennifer E Adair, Mark R Enstrom, Kevin G Haworth, Lauren E Schefter, Reza Shahbazi, Daniel R Humphrys, Shaina Porter, Kenric Tam, Matthew H Porteus, Hans-Peter Kiem. 2020. DNA Barcoding in Nonhuman Primates Reveals Important Limitations in Retrovirus Integration Site Analysis. Mol Ther Methods Clin Dev. Mar 30;17:796-809. doi: 10.1016/j.omtm.2020.03.021.