A noninvasive screening tool for cell fate determination using raman spectroscopy and machine learning in tissue engineering and microbiology applications

Hematopoietic stem cells (HSCs) are a small heterogeneous pool of self-renewing cells responsible for maintaining the blood and immune system in the human body. The hematopoietic systems of patients suffering from immune or hematological disorders may be reconstituted through the transplantation of healthy HSCs. Transplant success rate could increase if the number of HSCs transplanted per patient increased. Consequently, much research focuses on identifying the cues that would direct these stem cells to self-renew in artificial cultures. Though various cellular, biochemical, and mechanical cues regulate the fates of HSCs in their native microenvironment, attempts to use niche-inspired cues to expand and maintain HSCs in vitro for extended times have had limited success. Therefore, miniature biomaterial platforms for screening thousands of putative stimuli have been devised to identify the combinations of factors that drive self-renewal in vitro while reducing the numbers of rare HSCs required for screening. To correlate the extrinsic cues presented in these biomaterial platforms with the fates they elicit, the fate decisions of specific stem cells must be accurately identified. However, current methods are invasive, destructive and they lack location specificity, which precludes temporal analyses as well as use of valuable stem cells for further studies after determining their fate. The efficiency of screening HSC responses to extrinsic cues would improve if HSC responses (e.g., differentiation, quiescence) could be identified objectively, noninvasively, with location specificity and in situ within a miniature biomaterial platform. Raman spectroscopy enables noninvasively assay the biomolecular composition of individual, living cells to determine their phenotype and assess differentiation stage. Combining Raman spectroscopy with machine learning methods such as partial least-squares discriminant analysis could enable objective classification of stem cell fate outcomes on screening platforms, monitoring differentiation stages of cells over time and investigate the heterogeneity of stem cell fate decisions using the biochemical information in the cells’ Raman spectra. However, the miniature biomaterial platforms reported to date are not compatible with the acquisition of Raman spectra from individual cells in situ. This thesis describes the development of a miniature culture platform that is compatible with the acquisition of Raman spectra from individual cells in situ. Furthermore, by applying machine learning tools to the single-cell Raman spectra acquired in situ, the fates of stem and progenitor cells in response to the cues presented on the platform may be noninvasively tracked. We developed a novel microwell platform that facilitates repeatedly relocating the same cells, and noninvasively tracking their differentiation over time in response to 3D interactions with bioinspired hydrogels. We first established that Raman spectroscopy accurately identifies lineages of living cells cultured within hydrogels (3D cultures) as described in a proof-of-concept study in Chapter 2. Next, development and application of the microwell platform to noninvasively track the differentiation of hematopoietic-derived progenitor cells from the THP-1 line in response to 3D interactions with a hydrogel matrix is presented in chapter 3. Our findings demonstrate the utility of this tool to unveil the heterogeneity in the dynamics and trajectories of differentiation in the cell population. The data also suggest the presence of an intermediate cell population along the route from undifferentiated THP-1 to fully matured M2 cells in cultures supplemented with a cytokine. This platform may be employed to pre-screen for cells and time points prior to resource intensive analyses. Finally, Chapter 4 details a study that probes the feasibility of using Raman spectroscopy and machine leaning for non-medical application. In a proof-of-concept study, machine learning models of Raman spectra obtained from algal cells are used to identify changes in algal cell health that were triggered by environmental stress. Overall, the research described herein demonstrates in situ Raman spectroscopy and machine learning provide biochemical information that complements that provided by traditional bioassays. This approach may enhance the efficiency of research endeavors in tissue engineering and microbiology applications.

Name thesis:degree_name

Ph.D.

Level thesis:degree_level

Dissertation

Discipline thesis:degree_discipline

Chemical Engineering

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2024

Author dc:creator

• Godbole, Apurva Rajeev

Contributors dc:contributor

• Kraft, Mary
• Harley, Brendan
• Underhill, Gregory
• Kong, Hyun Joon

• Raman Spectroscopy
• Single-cell Tracking
• Pls-da
• Microwell Platform
• Noninvasive

Statement dc:rights

• Copyright 2024 Apurva Rajeev Godbole

Language dc:language

eng, en

Handle dc:identifier

https://hdl.handle.net/2142/124541

OAI identifier oai:identifier

oai:www.ideals.illinois.edu:2142/124541