Poster L18 - Using Automated Methods to Quantify Macrophage Phenotypes After Volumetric Muscle Loss Injury to Rat Gastrocnemius

Saturday, October 26, 2024

10:00 AM – 11:00 AM EST

Location: Exhibit Hall E, F & G

Presenting Author(s)

Cecilia Fetterolf

Undergraduate Student, Department of Biomedical Engineering
University of Virginia
Leesburg, Virginia, United States

Co-Author(s)

Sydney Shriver, BS

Graduate student
Department of Biomedical Engineering, University of Virginia, United States
Olivia Sergent

Graduate Research Assistant
Department of Biomedical Engineering, University of Virginia, United States
Kavita Parekh, B.S.

Graduate student
Department of Bioengineering, University of California, Berkeley
Berkeley, California, United States
Morgan Pfaff (she/her/hers)

Graduate student, Department of Bioengineering
University of California, Berkeley
Berkeley, California, United States
KH

Kevin E. Healy

Professor
Departments of Bioengineering and Materials Science Engineering, University of California, Berkeley, California, United States

Primary Investigator(s)

GC

George Christ

Professor of Biomedical Engineering and Orthopaedic Surgery
University of Virginia, United States

Introduction: Volumetric Muscle Loss (VML) is defined as an injury that exceeds the body’s capacity for endogenous repair, resulting in a permanent loss of muscle structure and function and a corresponding accumulation of fibrotic tissue. VML injuries are commonly caused by severe traumatic injury, surgery (e.g., tumor resection) and inflammation. These injuries are prevalent amongst military and civilian personnel. There are currently no treatments for VML injuries that can fully restore form and function.

We have developed a hyaluronic acid (HA) based biomaterial, both a hydrogel and a cryogel (MuscleMatrix™), as novel therapeutics for VML repair. The MuscleMatrix™ platform is biocompatible, biodegradable, nonimmunogenic, and easily modiﬁed through its modular HA chemistry. Importantly, creation of a pro-regenerative immune environment is crucial to treating VML as these injuries are characterized by aggressive fibrosis and scar formation due to aberrant inflammatory signals.

To this end, we characterized the impact of MuscleMatrix™ on the presence of M1-like and M2-like macrophages in the early weeks after a VML injury using an automated cell detection technique. M1s are often considered pro-inflammatory as they phagocytize debris and deposit collagen, while M2s can be considered anti-inflammatory/pro-regenerative as they align existing collagen and encourage remodeling. Macrophages are most active during the first 7-10 days of muscle wound healing, however VML-injuries result in a prolonged presence of macrophages. The goal of the current study was to begin to evaluate the impact of MuscleMatrix™ on the early immune response (1-4 weeks post-injury).

Materials and

Methods: A 20% VML injury (by weight, ~10 x 7 x 2 mm defect size) was created in the lateral head of the gastrocnemius (GN) muscle in 12-week-old Lewis rats. Injuries were subsequently treated with hydrogel, cryogel, or left untreated (n=3 per treatment per timepoint, n=27 animals total). Animals remained on study for 1, 2, or 4 weeks, after which they were euthanized and their GNs were explanted and snap frozen in liquid nitrogen for histologic analysis. Cryosections (10 µm) underwent a three-panel macrophage stain for each animal. The stain consisted of DAPI (nuclei), CD68 (pan-macrophage marker), and CD163 (M2-like macrophage marker). Injured regions of each tissue were captured using a 20x oil immersion objective on a Leica DMi8 Confocal microscope.

For data analysis, images were imported into QuPath, a software used for biological image analysis, where a 300 µm by 300 µm region was selected based on its ability to accurately represent the entire area. In these regions, the automated cell detection feature was used to identify nuclei based on the blue DAPI channel. Following nuclear detection and channel thresholding, all macrophages within the region were counted by hand; where cell detections with a strong blue-green fluorescence were classified as M1-like and cell detections with blue-green-red fluorescence classified as M2-like. An automated macrophage classifier, previously trained on tibialis anterior (TA) muscle sections, was used and results were compared to hand counted values to determine the accuracy of the corresponding measurements on the GN muscle (Figure 1).

Results, Conclusions, and Discussions: Reported here are the 1-week hand counted measures compared to the classifier counts for a small sample size (n=1 for unrepaired and hydrogel treated animals, n=2 for cryogel treated animals). The hand counted M1:M2 ratio for the cryogel treatment was 4.26, 1.22 for hydrogel, and 1.27 for no repair. Using the TA-trained classifier, these ratios were 1.56, 0.79, and 0.92 for cryogel, hydrogel, and no repair groups, respectively (Figure 2). On average, the classifier had a 23.27% error when classifying M1-like macrophages and a 58.59% error when classifying M2-like macrophages determined by comparing quantification values obtained from the classifier to those manually counted.

The pretrained classifier performed worse for M2-like macrophage classification, as compared to the average resultant error obtained from the M1-like phenotype. Additionally, the classifier had increased error when attempting to predict phenotypic ratios in cryogel treated animals.

The automated classifier used herein was trained on TA muscle sections. The training data did not include GN muscle samples, which exhibit a different morphology from the TA. Cryogel repair groups, in which material autofluorescence must be accounted for, were also not included in training for the classifier. When combined, these factors create a potential source of error suggesting the possibility of muscle-specific and treatment-specific training requirements.

Continued optimization is needed to improve accuracy of the classifier to predict macrophage phenotype. As such, a new machine-learning based classifier will be trained, validated, and tested on all treatments utilized in the GN muscle. Using QuPath as the primary source of classification, hand counted and classified macrophages will be used to train the classifier on a small region. The classifier will then be tested on distinct previously counted regions to evaluate accuracy, with additional training performed if necessary. Upon completion of classifier training, it will be deployed on a larger region for each animal timepoint and treatment. Ultimately, this will allow for an accurate depiction of how the macrophage-facilitated immune response progresses following VML and provide a metric for quantifying repair following injury.

Acknowledgements (Optional):