Combining deep mutational scanning and imaging to explain mutation pathogenicity and drug resistance in tubulins

Background

Microtubules, dynamic polymers composed of alpha and beta-tubulin subunits, are integral components of eukaryotic cells and are functionally conserved in bacteria through related essential cytoskeletal elements. These structures facilitate diverse cellular processes, such as cytoplasmic transport, cell polarity, and cell division via mitotic spindle assembly, and comprise cilia and flagella. Alpha and beta-tubulins serve as crucial targets for chemotherapeutic agents and antifungal drugs. Mutations in several different tubulin genes cause a range of inherited genetic disorders, the tubulinopathies, most commonly associated with neurodevelopmental disorders. Thus, there is a strong need to develop methods to identify which mutations observed in the human population are most likely to have clinically relevant effects. However, as we have recently shown [1], the effects of tubulin mutations tend to be very poorly predicted by current computational approaches, most likely due to their association with gain-of-function and dominant-negative mechanisms [2]. As an alternative, high-throughput deep mutational scanning (DMS) have recently emerged as strategy for experimentally probing variant effects on a large scale. In some cases, DMS experiments can far outperform computational predictors in the identification of disease-causing genetic variants [3].

Aims and Methodology

Aim 1: Application of DMS to tubulin genes using a yeast-based approach

We will use a yeast-based approach, with which we have recently had success for probing human disease genes [4], to measure the effects of all possible single-amino acid substitutions in alpha- and beta-tubulin genes. Our preliminary analysis shows that most human pathogenic tubulin mutations occur at sites that are conserved in their yeast orthologues. Thus, these results are likely to be directly relevant for understanding human genetic disease. Importantly, our experimental strategy, involving coexpression of mutant yeast tubulins alongside the genomically encoded wild-type proteins, will be particularly sensitive to dominant-negative and gain-of-function effects, which are the reason tubulin computational predictions currently fail [1].

Aim 2: Assessment of human tubulin mutations and drug resistance using a human cell assay

We will use a human cell assay to probe the effects of human tubulin mutations, focusing specifically on the TUBA1A isotype, associated with lissencephaly and expressed in most cell types, and TUBB4A, associated with dystonia and leukodystrophy. Using a HAP1 cell model, we will use saturation genome editing to construct tubulin missense mutants, covering protein regions identified by the yeast assays to be most sensitive to gain-of-function and dominant-negative effects. These will then be assayed using a high-throughput microscopy screen to assess the effects of variants on cytoskeletal morphology. In addition, we will repeat the assays in the presence of small molecule drugs, such as paclitaxel, vincristine, and nocodazole, to evaluate the impact of specific mutations on drug binding and resistance. By understanding the relationship between tubulin mutations and drug resistance, we can contribute to the development of more effective therapeutic strategies for cancer and tubulinopathies.

Aim 3: Integration of experimental outputs and machine learning for tubulin-specific prediction

Finally, using the outputs of the yeast and human cell experiments, and combining these with state-of-the-art variant effect prediction strategies based on protein language models, we will use supervised machine learning to train predictors specific to human tubulin isotypes. This approach will enable us to enhance the accuracy and relevance of tubulin mutation predictions, which will ultimately contribute to a better understanding of drug resistance in cancer and tubulinopathies, such as neurodevelopmental disorders.

Training Outcomes

1. High-throughput experimental techniques and microscopy: The student will gain experience in DMS, yeast-based assays, human cell assays, and microscopy-based screening. This will provide a strong foundation in the design, execution, and analysis of high-throughput experiments, as well as expertise in advanced microscopy and imaging techniques for assessing cellular structures and morphology.
2. Genomic editing and molecular biology: The student will learn saturation genome editing techniques and other molecular biology methods, which are vital for constructing and analyzing tubulin mutants in various cellular models.
3. Computational biology and machine learning: The student will gain experience in computational variant effect predictors, including state-of-the-art protein language models. The student will also use supervised machine learning techniques to integrate experimental data with evolutionary and structural features to develop tublin-specific predictive models.

References

[1] Attard TJ, Welburn JP, Marsh JA. Understanding molecular mechanisms and predicting phenotypic effects of pathogenic tubulin mutations. PLOS Computational Biology. 2022 Oct 7;18(10):e1010611.

[2] Bergendahl LT, Gerasimavicius L, Miles J, Macdonald L, Wells JN, Welburn JP, Marsh JA. The role of protein complexes in human genetic disease. Protein Science. 2019 Aug;28(8):1400-11.

[3] Livesey BJ, Marsh JA. Using deep mutational scanning to benchmark variant effect predictors and identify disease mutations. Molecular systems biology. 2020 Jul;16(7):e9380.

[4] McDonnell AF, Plech M, Livesey BJ, Gerasimavicius L, Owen LJ, Hall HN, FitzPatrick DR, Marsh JA, Kudla G. Deep mutational scanning quantifies DNA binding and predicts clinical outcomes of PAX6 variants. bioRxiv 10.1101/2023.07.25.550478