August 9, 2021
Facing a pandemic that posed limitations unprecedented in living memory, HuBMAP’s first cohort of student interns and their advisors persisted and succeeded. In a webinar to mark the end of their summer program, the students presented projects focused on topics including the female reproductive tract; the bowel and colon; collection of metadata to validate research results; and investigating diversity in HuBMAP’s datasets. These projects provided a unique education in bioinformatics and its affiliated disciplines for the interns, while also expanding the consortium’s capabilities.
An Investigation of the Upregulated Genes Found in the Female Reproductive Tract
Stephanie Bobadilla-Regalado, working with the Gregory laboratory at the University of Pennsylvania, described using single-cell RNA sequencing to monitor gene expression within the human female reproductive tract. She began the project with tissue samples from a patient undergoing female-to-male (FTM) sex reassignment.
Bobadilla-Regalado discovered that the high-testosterone hormone treatments of the procedure caused a unique shift in gene expression in a population of cells (termed “cluster 12” by the laboratory) in the cervix. The top five upregulated genes coded for immunoglobulin proteins, suggesting a shift toward “male” expression that may mirror the known differences between the sexes, and which shift in women through the menstrual cycle.
Next for the project will be repeating the result with vaginal, fallopian and ovarian tissues, employing FISH to map out the changes in gene expression spatially and carrying out a more in-depth analysis of the effects of hormone treatment.
Effects of Hormonal Therapy on the Single Cellular Level
Tatiana Gonzalez, also working with the Gregory lab, focused on cervical gene expression changes downstream of FTM sex-reassignment hormone therapy. She did so in a different population of cells, “cluster 13.”
Gonzalez found three genes—MIR31HG, MUC16 and RHEX—to be most highly upregulated in correlation with hormone therapy. The genes were downregulated in other cervical cells during the therapy.
All of these genes are associated with malignant or benign cell growth. MIR31HG is a cancer-causing oncogene; MUC16 is another oncogene associated with increased metastasis of cancer cells as well as with ovarian cysts; and RHEX is associated with expansion of erythroid (red blood) cells, a process sometimes associated with chronic inflammation and tumor formation.
These findings shed light on health risks associated with hormone therapy as well as more general lessons on how the female reproductive tract responds to sex hormones. Future goals for the project include analysis of other organs in the female reproductive tract.
Spatial Transcriptomics: Regional Importance in the Female Reproductive System
In the third presentation of the webinar, Ogechukwu Etuazim, who worked with the O’Neill laboratory at UPenn, described a project to decode the biochemical signals that drive oocyte development; follicular migration; ovulation; and implantation of the fertilized oocyte, the zygote.
The process from oocyte maturation through zygote implantation is fundamental to reproduction, but does not always proceed successfully. One particular mode of failure is ectopic pregnancy, in which the egg implants in a part of the reproductive tract other than the uterus’s inner layer, the endometrium. Sometimes life-threatening, ectopic implantations can span the reproductive tract but are most common (95 to 96 percent of cases) in the fallopian tubes connecting the ovary to the uterus.
In her project, Etuazim will explore both targeted analysis of specific genes, through quantitative PCR (qPCR) expansion of a specific mRNA sequence, and untargeted single-cell RNA sequencing of all the mRNA species in a given cell to decipher gene-expression differences associated with successful versus ectopic implantations.
Challenges include the high monetary cost of the technologies, the limited spatial resolution of qPCR and compatibility with archival tissue samples whose fixation may interfere with the signals.
3D Printed Ovarian Molds
In a turn from biology to fabrication, Kate da Silva described creating a mold to allow replicable, sample-independent sectioning of human ovaries regardless of the size or shape of the organ. Da Silva worked in the Penn Image Computing and Science Lab (PICSL) in conjunction with the O’Neill lab.
Sectioning the ovary into biologically relevant thin sections has posed a challenge. Due to individual differences in organ size as well as the spheroidal shape of an ovary, the standard Thomas Stadie-Riggs tissue slicer poses difficulties in positioning and creating replicable cutting.
Da Silva employed 3D printing using Fusion 360 software and the Python API programming interface to create a mold with pre-set slicing guides out of the hard plastic acrylonitrile butadiene styrene (ABS). The ABS provides strength and rigidity not possible with the more standard PLA (polylactide). The custom shape of the mold, fitting the specific organ to be dissected, overcomes the positioning issues. These molds have now been used to successfully and rapidly section individual ovaries into 12 sections of equidistant thickness dependent on ovary size, providing enough sections for studies of differential gene expression across the organ.
Limitations to the technology to be addressed, stemming from the ABS material, include inability to autoclave the mold; shedding of plastic during slicing; and poor conductivity, making freezing of mold and organ together not possible.
3D Modeling of the Uterus with Magnetic Resonance Imaging
Casey Henson worked with the Penn Image Computing & Science Laboratory to apply 3D segmentation and animation to magnetic resonance images of the human uterus.
Variations in the dimensions and relative height to width of the uterus stemming from individual variation as well as age and reproductive history make replicable sectioning of images of the uterus challenging. These anatomical distinctions may be important, as the roughly 1.6 height-to-width ratio typical at age 21 associates with maximal fertility.
Using the ParaView visualization tool and the open-source ITK-SNAP software, Henson created an animated sectioning of uterine MRI images following the sectioning strategy previously developed by the O’Neill lab for dissecting biological samples.
The method holds promise for anatomical mapping and spatial orientation of data generated throughout HuBMAP.
Small Bowel and Colon Codex Segmentation
Another segmentation project, by Injyil Gates working with the Nolan lab at Stanford University, focused on multiplex imaging of cells from a series of samples taken from eight sites in the small bowel and colon.
Gates used antibody probes conjugated with unique DNA sequences to create a sequential bar code of fluorescent DNA staining for proteins present in each cell in a given sample—the CODEX multiplex imaging procedure. One key to assigning signals from the 44 studied proteins to individual cells was successful segmentation of the cells and their nuclei—time-consuming when done by humans, and error-prone when automated. Gates learned how to use the Annotator J and Image J software packages to segment the images, manually correcting for segmentation errors to create final images ready for analysis.
One future goal for the project will be to make manual correction unnecessary.
Kim’s Collection—A Metadata Collection App
Working with the Kim lab at University of Pennsylvania, Oluwafolajinmi Olugbodi investigated ways of retrieving biologically relevant metadata from HuBMAP’s data collections.
Currently the tools associated with HuBMAP’s data do not allow collecting metadata without intensive labor and time commitment. Such collection is necessary to establish provenance and assure apples-to-apples data comparisons.
Olugbodi devised a hierarchy of batch, group and sample attributes to allow ongoing capture of metadata before, during and after experimental procedures. The framework will allow user-friendly and mobile input of the metadata with minimal additional effort, as researchers work.
Using Apple’s CloudKit tool, Olugbodi was able to create a system with pre-set entities, relationships and class definitions that allow a high-level abstract mapping of the objects in a given dataset that should enable just such a framework.
Visualization of Diversity in HuBMAP Data
Roselkis Morla Adames, working with the Gehlenborg lab at Harvard Medical School, described the collection and analysis of medically relevant measures of diversity, including age, sex, race, ethnicity and other factors.
Currently, Adames said, there is no easy way to overview the diversity of HuBMAP’s donor pool. To overcome this limitation, she created a new page for the HuBMAP portal to allow visualization of diversity data. Starting with a user survey of the most scientifically important combinations of donor characteristics, Adames used the Elasticsearch engine to search and analyze these characteristics. Her initial work included visualization of race information in the datasets correlated with donor sex.
The results, she added, suggest that donor diversity in HuBMAP can be improved, and that inconsistent data collection for the diversity measures may have left gaps in the database. She would like to see the web page she has developed serve as a tool for monitoring and encouraging progress in recruiting a more diverse donor set.
The inaugural HuBMAP summer internship program provided the opportunity for students interested in scientific research to work with HuBMAP member groups and gain invaluable mentorship and training. This webinar gave HuBMAP members a chance to meet these talented students and hear about their research progress before they return to the classroom.
For more on the internship program, please visit the website.