Assessment of machine-learning- versus experimental structural models of sex-specific isoforms of linker histone H1

Uça, Merve
Dimitrov, Stefan
Kale, Seyit
The nucleosome is the repeating unit of chromatin and it is composed of an octamer histone core that holds together around 145 base pairs of DNA in a disc-shaped configuration. In higher eukaryotes, an additional histone, known as linker histone H1, provides further structural order by stabilizing the nucleosomal DNA termini. Histone H1 family includes seven somatic variants (H1.0 to H1.5 and H1.10) and four variants expressed only in germ cells, three of which are testisspecific (H1T, H1T2m and HILS1), and one oocyte-specific (H1.8). These isoforms are expressed differentially across cell types and stages, suggesting diverse functional roles in various physiological processes including gene expression, differentiation, aging, and cancer. In this work, we employed molecular modeling and molecular dynamics simulations to gain insights into the dynamic behavior of these sex-specific isoforms in free- and chromatin-bound states. We surveyed the Protein Data Bank (PDB) for available linker histone structures in nucleosome-bound states. For isoforms where an experimental structure is not available or is not sufficiently reliable, we used AlphaFold predicted structures as well as SuperLooper2 structural corrections. We solvated each H1 isoform in a cubic water box to mimic a physiological buffer solution. We used the Charmm36m force field together with the OPC water model. We collected production trajectories sub-microsecond using Gromacs molecular dynamics suite, version 2018.3. We used VMD and in-house Python scripts for analysis. During molecular dynamics simulation of H1.8 isoform in chromatin-bound state, we observed local structural instabilities such as loss of secondary structure features (graphical abstract, left panel). To resolve this issue, we generated multiple different starting models of H1.8 using conventional molecular modelling approaches as well as the more recent machine-learning based structure prediction tools. Among all the structural models we tested, the AlphaFold predicted initial structure provided the best overall stability (graphical abstract, right panel), excelling not only in comparison to the original, cryo-EM experimental model but also in comparison to the conventional molecular modelling proposals. Encouraged by this result, we are now in the process of exploring the possible binding modes of the testis-specific H1.T to chromatin, for which only machine-learning predicted atomic models are available, albeit in isolated (not nucleosome bound) state. This endeavor is beneficial not only from a computational structural biological standpoint but also in terms of validation of machinelearning based protein structure predictions which yet lack a consideration of the effect of binding partners.


Design of an emission ratiometric biosensor from MerR family proteins: A sensitive and selective sensor for Hg(II)
Wegner, Seraphine V.; Okesli, Ayse; Chen, Peng; He, Chuan (2007-03-25)
A ratiometric fluorescent biosensor for Hg2+ is constructed from the MerR protein and a duplex DNA containing a pyrene excimer.
Structural properties of an engineered outer membrane protein G mutant, OmpG-16SL, investigated with infrared spectroscopy
Yilmaz, Irem; Yildiz, Ozkan; KORKMAZ ÖZKAN, FİLİZ (Informa UK Limited, 2019-05-31)
The structural and functional differences between wild type (WT) outer membrane protein G and its two mutants are investigated with Fourier transform infrared spectroscopy. Both mutants have a long extension to the primary sequence to increase the number of beta-strands from 14 (wild type) to 16 in an attempt to enlarge the pore diameter. The comparison among proteins is made in terms of pH-dependent conformational changes and thermal stability. Results show that all proteins respond to pH change but at dif...
Investigation of SNX3 in colon cancer
Cingöz, Harun; Erson Bensan, Ayşe Elif; Department of Biology (2019)
Sorting Nexin 3 (SNX3) is part of the retromer complex that recycles cargo receptors back to plasma membrane or to Trans Golgi Network. WNT ligand carrier protein Wntless (WLS) is a known SNX3 cargo protein. Our earlier data suggested over expression of SNX3 in colon cancer cells. Considering its importance in receptor recycling, we hypothesized SNX3 to be a potential modulator of cancer related receptors. To begin understanding the role of SNX3, we developed RNAi models of SNX3 in SW480 colon cancer cells ...
Experimental Generation of Carcinoma-Associated Fibroblasts (CAFs) from Human Mammary Fibroblasts
Polanska, Urszula M.; Acar, Ahmet; Orimo, Akira (2011-10-01)
Carcinomas are complex tissues comprised of neoplastic cells and a non-cancerous compartment referred to as the 'stroma'. The stroma consists of extracellular matrix (ECM) and a variety of mesenchymal cells, including fibroblasts, myofibroblasts, endothelial cells, pericytes and leukocytes(1-3).
Realization of polarization-angle-independent fishnet-based waveguide metamaterial comprised of octagon shaped resonators with sensor and absorber applications
Sabah, Cumali (Springer Science and Business Media LLC, 2016-05-01)
A new fishnet-based waveguide metamaterial structure for the microwave region is introduced and investigated both numerically and experimentally. The proposed model is designed and fabricated on both sides of the substrate and exhibits strong metamaterial behavior (such as negative material parameters: i.e. negative permittivity, negative permeability, and negative index of refraction) at the resonance. Only one single slab is used in the simulation and experiment which provides a reduction in the number of...
Citation Formats
M. Uça, S. Dimitrov, and S. Kale, “Assessment of machine-learning- versus experimental structural models of sex-specific isoforms of linker histone H1,” Erdemli, Mersin, TÜRKİYE, 2022, p. 2023, Accessed: 00, 2023. [Online]. Available: