Chlamydomonas Cellular Phenotypes Enable in vivo Validation of Computationally Designed Therapeutic ADA1 Variants

A NOTE ON SHARING WITH US! Part of our mission is to share as much useful research as we can. If you choose to share a protocol or other useful information with us after viewing this poster, please understand that we may act upon this knowledge and share it when we publish our work. We publish quickly on an independent platform, so this may happen soon after you share, and we cannot wait for you to publish elsewhere. If you decide to share anyway, yay! That's what science is all about. If your input is useful, we will include you as a contributor to the publication and explain that your role was in providing "Critical Feedback," likely with an additional description of what you shared. tl,dr —If you're not ready for everyone to know about something, please refrain from sharing it with us. Prachee Avasthi • Supervision, Conceptualization Audrey Bell • Visualization Brae M Bigge • Supervision, Conceptualization Ben Braverman • Investigation, Methodology Christopher Bulow • Supervision, Critical Feedback Tara Essock-Burns • Investigation, Formal Analysis Megan Hochestrasser • Editing Evan Kiefl • Software, Formal Analysis Ilya Kolb • Investigation, Formal Analysis, Methodology Ryan Lane • Software, Formal Analysis Cameron Dale MacQuarrie • Investigation, Formal Analysis, Methodology, Conceptualization David G Mets • Supervision, Conceptualization Austin Patton • Software, Formal Analysis Sunanda Sharma • Investigation 1. Avasthi P, McGeever E, Patton AH, York R. (2024). Leveraging evolution to identify novel organismal models of human biology. https://doi.org/10.57844/arcadia-33b4-4dc5 2. Sharma S, Mets DG, Kolb I, Braverman B. (2025). AutoOpenRaman: Low-cost, automated Raman spectroscopy. https://doi.org/10.57844/arcadia-7vbd-n3ry 3. Essock-Burns T, Lane R, MacQuarrie CD, Mets DG. (2024). Rescuing Chlamydomonas motility in mutants modeling spermatogenic failure. https://doi.org/10.57844/arcadia-fe2a-711e 4. Braverman B, Mets DG, York R. (2024). The phenotype-o-mat: A flexible tool for collecting visual phenotypes. https://doi.org/10.57844/arcadia-112f-5023 5. Hu Q, Sommerfeld M, Jarvis E, Ghirardi M, Posewitz M, Seibert M, Darzins A. (2008). Microalgal triacylglycerols as feedstocks for biofuel production: perspectives and advances. https://doi.org/10.1111/j.1365-313X.2008.03492.x 6. Ji Y, He Y, Cui Y, Wang T, Wang Y, Li Y, Huang W, Xu J. (2014). Raman spectroscopy provides a rapid, non-invasive method for quantitation of starch in live, unicellular microalgae. https://doi.org/10.1002/biot.20140016 7. Lunardon A, Patena W, Pacini C, Warren-Williams M, Zubak Y, Laudon M, Silflow C, Lefebvre P, Jonikas M. (2024). The Chlamydomonas reinhardtii CLiP2 mutant collection expands genome coverage with high-confidence disrupting alleles. https://doi.org/10.1101/2024.12.16.626622 Mutating ADA1 in Chlamydomonas induces distinct cellular phenotypes, characterized by increased starch accumulation, enhanced swimming longevity, and elevated linear velocity. We successfully rescued these phenotypes by co-expressing ML-designed ADA1 variants, with several candidates surprisingly surpassing wild-type function. However, because standard Chlamydomonas transformation results in random genomic integration, we observe significant variability between clones expressing the same variant. Consequently, it remains unclear which clonal rescues are intrinsic to the designed enzyme kinetics or artifacts of high-copy expression and off-target genomic disruption. To validate these findings, we are currently mapping insertion sites to confirm that phenotypic recovery is driven strictly by the ADA1 variant rather than positional effects. Additionally, we are performing quantitative molecular assays to confirm protein expression levels directly. Simultaneously, we are developing open-source methods for targeted gene integration in C. reinhardtii to standardize expression across clones Additionally, we are performingin vitroexperiments on purified ADA1 variants to characterize properties like stability and in vitro activity. We are also using this pipeline to test potential enzyme replacement variants for additional lysosomal disorders caused by other enzymes, including UBE2A, PPT1, and DNASE2,in S. pombe, S. cerevisiae, C. reinhardtii, and C. elegans. • Adenosine Deaminase (ADA) deficiency leads to SCID due to the accumulation of toxic purine metabolites that destroy T-cells. • Current Enzyme Replacement Therapy (ERT) relies on bovine (cow) ADA. This is frequently immunogenic, leading to antibody neutralization and treatment failure. • Machine Learning (ML) can design stabilized human ADA variants to evade immune rejection. • We lack a rapid, high-throughput in vivo system to screen these variants. • We leveraged our Zoogle dataset (zoogle.arcadiascience.com) to establish C. reinhardtii as an ideal model based on high ADA1 evolutionary conservation. • We obtained and characterized an ADA1-deficient mutant (CLIP2 library) to establish baseline disease phenotypes. • We engineered a multi-modal phenotyping pipeline to quantify motility and metabolic health in high-throughput. • We transformed the mutant strain with ML-designed human ADA1 variants to assess their ability to restore wild-type function in vivo . • ADA1 mutants exhibit significantly altered motility and metabolic attributes compared to wild-type controls. • Treatment with ML-generated variants results in phenotypic recovery, shifting the mutant behavior back toward WT levels in some clones. • Establishes Chlamydomonas as a robust, non-mammalian host for modeling human metabolic errors. • Validates a rapid, cost-effective pipeline to screen large libraries of ML-designed variants • Bridges the gap between computational design and in vivo validation 0 2 4 6 8 Trait Distance Score H. sapiens C. reinhardtii X. tropicalis C. milii M. murinus Symbiodinium sp. A.amoebiformis P. pacificus S. mediterranea S. mansoni D. discoideum A.aegypti C. hemisphaerica C. vulgaris H. vulgaris D. rerio P. troglodytes N. crassa S. pombe P. falciparum A.nidulans N.gruberi P. tetraurelia S. arctica H. miamia C. jacchus A.carolinensis C. elegans M. leidyi P. tricornutum E. diaphana C. intestinalis P. marinus O. tauri T. thermophila S. cerevisiae E. gracilis B. saltans E. histolytica M. commoda N. vectensis U. maydis D. papillatum A.bisporus D. melanogaster C. syrichta M. musculus C. albicans T. guttata P. marinus V. carteri I.galbana P. chrysogenum G. gallus S. rosetta M. mulatta Nannochloropsis sp. Cre05.g246377 Cre03.g161000 Figure 2) Linear Discriminant Analysis (LDA) of filtered Raman spectra. Spectra were collected using a Wasatch 785 nm system. Figure 9) Bars represent the mean linear (left) and angular (right) swimming speeds across all recovered clones for each strain or construct. Each dot corresponds to the clone-level mean speed, reflecting variability among independent transformation clones. Error bars indicate the standard deviation. 0 200 400 600 800 E1_65 E1_67 CLS-HsADA1 CrADA1 CrADA1-6xHIS ADA1 Mutant WT None None Fluorescein-5-thiosemicarbazide (FTSC) Fluorescent Intensity, A.U. Mean Single Cell FTSC Intensity 0 200 400 600 800 ADA1 Mutant WT None None E2_83 GV_107 Empty Vector 0 1 2 3 4 5 Mean Linear Speed PM_19 E1_67 GV_107 E1_65 PM_ 34 E2_83 HsADA1 ADA1 Mutant WT None None Mean Angular Speed 0 2 4 6 8 10 PM_19 E1_67 GV_107 E1_65 PM_ 34 E2_83 HsADA1 ADA1 Mutant WT None None −2 0 2 −1 0 1 2 3 LD-1 (59.2%) LD-2 (20.8%) Strain Nitrogen Peak Intensity at 2917 cm -1 (a.u.) WT WT ADA Mutant + 150 100 0 50 -+- ADA Mutant Chlorophyll Starch (FTSC) wild-type ADA1 Mutant ADA1 Varient DNA Genomic DNA Chloroplast Pyrenoid Nucleus Figure 3) Representative images of fluorescent staining of starch (FTSC, green) and intrinsic chlorophyll fluorescence (magenta). Maximum Intensity Projections of Z-stack images. Figure 4) A) Nitrogen-starved cells stained with 25% Lugol’s iodine. B) Quantification of whole cell brightfield reciprocal intensity. ***p<0.001. Figure 6) A) Schematic of the imaging assay from [3] depicts the experimental setup. B) Mean Center Contrast Score (Center Intensity Divided by Edge Intensity) over time of WT (n=132, blue) and ADA1 mutant (n=96, orange) wells. Representative wells are included at the appropriate time point. 0 2000 4000 6000 8000 10000 12000 14000 Mean Single Cell Chlorophyll Intensity E1_65 E1_67 E2_83 GV_107 CLS-HsADA1 PM_19 PM_34 CrADA1 CrADA1-6xHIS Empty Vector HsADA1 ADA1 Mutant WT Transformant None None Strain Chlorophyll Fluorescent Intensity, A.U. Figure 7) Bars represent the mean FTSC fluorescence across all recovered clones for each strain or construct. Each dot corresponds to the mean single-cell fluorescence measured from an individual transformation clone, reflecting clone-to-clone variability. Error bars indicate the standard deviation. Figure 8) Bars represent the average chlorophyll fluorescence intensity across all recovered clones for each strain or construct. Each dot corresponds to the mean single-cell chlorophyll signal measured from an individual transformation clone, reflecting clone-to-clone variation. Error bars indicate the standard deviation. Human ADA1 Sequence ADA1 Variants ESM1 ESM2 ProteinMPNN GNN ADA1 mutant cells show distinct Raman spectral signatures relative to WT. 0 200 400 600 800 1000 1200 1400 1600 Brightfield Reciprocal Intensity, A.U. WT ADA1 Mutant *** WT ADA1 Mutant 25% Lugol’s Iodine Variant Design Random Genome Integration Each dot below represents the clonal mean. Clones from the Same Transformation are Genetically Diverse Figure 5) Peak intensity at 2917 cm⁻¹—associated with starch-rich biochemical features—is elevated in ADA1 mutants relative to wild-type cells, both under nitrogen-replete and nitrogen-starved conditions. ADA1 mutants exhibit disrupted starch sheaths. wild-type ADA1 Mutant ADA1 mutants show increased starch accumulation. A B Increased starch in ADA1 mutants is detectable by Raman spectroscopy. C. reinhardtii ranks better than super model organisms in similarity to human ADA1 Figure 1) ADA1 orthologs were detected in 56 of the 62 organisms analyzed in the Zoogle dataset. Trait distance scores quantify similarity to human ADA1; lower scores indicate greater similarity. The Chlamydomonas reinhardtii gene with the highest similarity, Cre05.g246377, has been previously annotated as ADA2; here we refer to it as ADA1 for simplicity. ADA1 mutants consistently accumulate excess starch, and several ADA1 variants show trends toward phenotypic rescue despite clonal variability. ADA1 mutants have increased chlorophyll autofluoresnce, with several ADA1 variant transformants showing trends of phenotypic rescue. ADA1 mutation alters swimming behavior, with ADA1 variant transformants showing variable trends toward resuce The Results The Approach The Problem The Impact Rescuing ADA1 with Designer Variants 0 10 20 30 40 Time, Hr 0.960 0.965 0.970 0.975 0.980 0.985 0.990 Center Contrast Score ADA1 Mutant Wild-Type Camera captures top-down view (shown at right) Motile Non-motile Non-motile cells sink to bottom of well, allowing light to pass *Linescan position Motile cells swim throughout well, blocking light 384-well, V-bottom plate Camera Transillumination LED * Background Next Steps References All published work: research.arcadiascience.com Download the Poster: Contributors (A–Z) Conclusions/Summary ADA1 Mutant Phenotypes bit.ly/chlamy-ADA A B ADA1 mutants show prolonged swimming endurance compared to WT Chlamydomonas Cellular Phenotypes Enable in vivo Validation of Computationally Designed Therapeutic ADA1 Variants Cameron Dale MacQuarrie @cdmacquarrie Presented by @cdmacquarrie.bsky.social [email protected]