DeepTM: A deep learning algorithm for prediction of melting temperature of thermophilic proteins directly from sequences

Thermally stable proteins find extensive applications in industrial production, pharmaceutical development, and serve as a highly evolved starting point in protein engineering. The thermal stability of proteins is commonly characterized by their melting temperature (Tm). However, due to the limited availability of experimentally determined Tm data and the insufficient accuracy of existing computational methods in predicting Tm, there is an urgent need for a computational approach to accurately forecast the Tm values of thermophilic proteins. Here, we present a deep learning-based model, called DeepTM, which exclusively utilizes protein sequences as input and accurately predicts the Tm values of target thermophilic proteins on a dataset consisting of 7790 thermophilic protein entries. On a test set of 1550 samples, DeepTM demonstrates excellent performance with a coefficient of determination (R2) of 0.75, Pearson correlation coefficient (P) of 0.87, and root mean square error (RMSE) of 6.24 ℃. We further analyzed the sequence features that determine the thermal stability of thermophilic proteins and found that dipeptide frequency, optimal growth temperature (OGT) of the host organisms, and the evolutionary information of the protein significantly affect its melting temperature. We compared the performance of DeepTM with recently reported methods, ProTstab2 and DeepSTABp, in predicting the Tm values on two blind test datasets. One dataset comprised 22 PET plastic-degrading enzymes, while the other included 29 thermally stable proteins of broader classification. In the PET plastic-degrading enzyme dataset, DeepTM achieved RMSE of 8.25 ℃. Compared to ProTstab2 (20.05 ℃) and DeepSTABp (20.97 ℃), DeepTM demonstrated a reduction in RMSE of 58.85% and 60.66%, respectively. In the dataset of thermally stable proteins, DeepTM (RMSE=7.66 ℃) demonstrated a 51.73% reduction in RMSE compared to ProTstab2 (RMSE=15.87 ℃). DeepTM, with the sole requirement of protein sequence information, accurately predicts the melting temperature and achieves a fully end-to-end prediction process, thus providing enhanced convenience and expediency for further protein engineering.

Explanation of inconsistencies in the determination of human serum albumin thermal stability

Thermal denaturation of human serum albumin has been the subject of many studies in recent decades, but the results of these studies are often conflicting and inconclusive. To clarify this, we combined different spectroscopic and calorimetric techniques and performed an in-depth analysis of the structural changes that occur during the thermal unfolding of different conformational forms of human serum albumin. Our results showed that the inconsistency of the results in the literature is related to the different quality of samples in different batches, methodological approaches and experimental conditions used in the studies. We confirmed that the presence of fatty acids (FAs) causes a more complex process of the thermal denaturation of human serum albumin. While the unfolding pathway of human serum albumin without FAs can be described by a two-step model, consisting of subsequent reversible and irreversible transitions, the thermal denaturation of human serum albumin with FAs appears to be a three-step process, consisting of a reversible step followed by two consecutive irreversible transitions.

Domain-wise dissection of thermal stability enhancement in multidomain proteins

Stability is critical for the proper functioning of all proteins. Optimization of protein thermostability is a key step in the development of industrial enzymes and biologics. Herein, we demonstrate that multidomain proteins can be stabilized significantly using domain-based engineering followed by the recombination of the optimized domains. Domain-level analysis of designed protein variants with similar structures but different thermal profiles showed that the independent enhancement of the thermostability of a constituent domain improves the overall stability of the whole multidomain protein. The crystal structure and AlphaFold-predicted model of the designed proteins via domain-recombination provided a molecular explanation for domain-based stepwise stabilization. Our study suggests that domain-based modular engineering can minimize the sequence space for calculations in computational design and experimental errors, thereby offering useful guidance for multidomain protein engineering.

Characterization of hydroxymethylpyrimidine phosphate kinase from mesophilic and thermophilic bacteria and structural insights into their differential thermal stability

The hydroxymethylpyrimidine phosphate kinases (HMPPK) encoded by the thiD gene are involved in the thiamine biosynthesis pathway, can perform two consecutive phosphorylations of 4-amino-5-hydroxymethyl-2-methyl pyrimidine (HMP) and are found in thermophilic and mesophilic bacteria, but only a few characterizations of mesophilic enzymes are available. The presence of another homolog enzyme (pyridoxal kinase) that can only catalyze the first phosphorylation of HMP and encoded by pdxK gene, has hampered a precise annotation in this enzyme family. Here we report the kinetic characterization of two HMPPK with structure available, the mesophilic and thermophilic enzyme from Salmonella typhimurium (StHMPPK) and Thermus thermophilus (TtHMPPK), respectively. Also, given their high structural similarity, we have analyzed the structural determinants of protein thermal stability in these enzymes by molecular dynamics simulation. The results show that pyridoxal kinases (PLK) from gram-positive bacteria (PLK/HMPPK-like enzymes) constitute a phylogenetically separate group from the canonical PLK, but closely related to the HMPPK, so the PLK/HMPPK-like and canonical PLK, both encoded by pdxK genes, are different and must be annotated distinctly. The kinetic characterization of StHMPPK and TtHMPPK, shows that they perform double phosphorylation on HMP, both enzymes are specific for HMP, not using pyridoxal-like molecules as substrates and their kinetic mechanism involves the formation of a ternary complex. Molecular dynamics simulation shows that StHMPPK and TtHMPPK have striking differences in their conformational flexibility, which can be correlated with the hydrophobic packing and electrostatic interaction network given mainly by salt bridge bonds, but interestingly not by the number of hydrogen bond interactions as reported for other thermophilic enzymes. ENZYMES: EC 2.7.1.49, EC 2.7.4.7, EC 2.7.1.35, EC 2.7.1.50.

Effects of ultrasound-assisted freezing at different power levels on the structure and thermal stability of common carp (Cyprinus carpio) proteins

This study investigated the effect of ultrasound-assisted immersion freezing (UIF) at different ultrasonic power levels on the myofibrillar protein primary, secondary and tertiary structures of common carp (Cyprinus carpio). Furthermore, protein thermal stability, electrophoresis pattern, and microstructure of the muscle tissue were also studied. Compared with a control, an ultrasonic power of less than 175 W had no significant negative effect on protein primary structure (P > 0.05), including total sulfhydryl, reactive sulfhydryl, carbonyl groups, free amino groups, dityrosine content, and surface hydrophobicity. UIF at 175 W (UIF-175) minimized the changes in protein secondary and tertiary structures. There were no obvious changes in the SDS-PAGE patterns of the control and frozen sample proteins. Microstructure analysis showed that an appropriate ultrasonic power (UIF-175) promoted the formation of smaller and more uniform ice crystals, reduced the damage of muscle tissue by ice crystals, and maintained the sarcomere integrity. In addition, UIF-175 samples had higher protein thermal stability. Overall, ultrasound treatment at a proper power (UIF-175) effectively minimized the changes in protein structure and protected the protein thermal stability during freezing process.

The comparison of ultrasound-assisted immersion freezing, air freezing and immersion freezing on the muscle quality and physicochemical properties of common carp (Cyprinus carpio) during freezing storage

This study investigated the impact of ultrasound-assisted immersion freezing (UIF), air freezing (AF), and immersion freezing (IF) on the ice crystal size, protein thermal stability, and physicochemical properties of common carp (Cyprinus carpio) muscle during frozen storage. UIF samples had smaller ice crystals throughout the storage period than AF and IF samples did, which led to less damage to the muscle tissue. Low-field nuclear magnetic resonance analysis revealed that UIF reduced the mobility and loss of immobilized and free water. The thawing and cooking losses in the UIF samples were significantly lower than those in the IF and AF samples (P < 0.05). The AF samples had a higher shear force (P < 0.05) than UIF and IF samples did at the beginning of storage, and then the shear force reduced rapidly. During the 90-180 days, the shear force of the UIF samples was higher than that of the AF and IF samples (P < 0.05). Decreases in the T_max and enthalpies were observed for all of the treatments during storage, and the UIF samples had a higher protein thermal stability than AF and IF samples did. The UIF samples showed lower thiobarbituric acid reactive substance and total volatile basic nitrogen values during storage than the AF and IF samples did (P < 0.05). Principal component analysis showed that there were significant correlations between the freezing methods and the ice crystal size, protein thermal stability and physicochemical characteristics of frozen muscles. Overall, UIF was an effective way to inhibit the deterioration of frozen fish during frozen storage.

Thermal proteome profiling: unbiased assessment of protein state through heat-induced stability changes

In recent years, phenotypic-based screens have become increasingly popular in drug discovery. A major challenge of this approach is that it does not provide information about the mechanism of action of the hits. This has led to the development of multiple strategies for target deconvolution. Thermal proteome profiling (TPP) allows for an unbiased search of drug targets and can be applied in living cells without requiring compound labeling. TPP is based on the principle that proteins become more resistant to heat-induced unfolding when complexed with a ligand, e.g., the hit compound from a phenotypic screen. The melting proteome is also sensitive to other intracellular events, such as levels of metabolites, post-translational modifications and protein-protein interactions. In this review, we describe the principles of this approach, review the method and its developments, and discuss its current and future applications. While proteomics has generally focused on measuring relative protein concentrations, TPP provides a novel approach to gather complementary information on protein stability not present in expression datasets. Therefore, this strategy has great potential not only for drug discovery, but also for answering fundamental biological questions.

Differential binding of monovalent cations to KcsA: Deciphering the mechanisms of potassium channel selectivity

This work explores whether the ion selectivity and permeation properties of a model potassium channel, KcsA, could be explained based on ion binding features. Non-permeant Na⁺ or Li⁺ bind with low affinity (millimolar K_D's) to a single set of sites contributed by the S1 and S4 sites seen at the selectivity filter in the KcsA crystal structure. Conversely, permeant K⁺, Rb⁺, Tl⁺ and even Cs⁺ bind to two different sets of sites as their concentration increases, consistent with crystallographic evidence on the ability of permeant species to induce concentration-dependent transitions between conformational states (non-conductive and conductive) of the channel's selectivity filter. The first set of such sites, assigned also to the crystallographic S1 and S4 sites, shows similarly high affinities for all permeant species (micromolar K_D's), thus, securing displacement of potentially competing non-permeant cations. The second set of sites, available only to permeant cations upon the transition to the conductive filter conformation, shows low affinity (millimolar K_D's), thus, favoring cation dissociation and permeation and results from the contribution of all S1 through S4 crystallographic sites. The differences in affinities between permeant and non-permeant cations and the similarities in binding behavior within each of these two groups, correlate fully with their permeabilities relative to K⁺, suggesting that binding is an important determinant of the channel's ion selectivity. Conversely, the complexity observed in permeation features cannot be explained just in terms of binding and likely relates to reported differences in the occupancy of the S2 and S3 sites by the permeant cations.

Optimization of protein samples for NMR using thermal shift assays

Maintaining a stable fold for recombinant proteins is challenging, especially when working with highly purified and concentrated samples at temperatures >20 °C. Therefore, it is worthwhile to screen for different buffer components that can stabilize protein samples. Thermal shift assays or ThermoFluor(®) provide a high-throughput screening method to assess the thermal stability of a sample under several conditions simultaneously. Here, we describe a thermal shift assay that is designed to optimize conditions for nuclear magnetic resonance studies, which typically require stable samples at high concentration and ambient (or higher) temperature. We demonstrate that for two challenging proteins, the multicomponent screen helped to identify ingredients that increased protein stability, leading to clear improvements in the quality of the spectra. Thermal shift assays provide an economic and time-efficient method to find optimal conditions for NMR structural studies.

Thermal stability of purified and reconstituted CFTR in a locked open channel conformation

CFTR is unique among ABC transporters as the only one functioning as an ion channel and from a human health perspective because mutations in its gene cause cystic fibrosis. Although considerable advances have been made towards understanding CFTR's mechanism of action and the impact of mutations, the lack of a high-resolution 3D structure has hindered progress. The large multi-domain membrane glycoprotein is normally present at low copy number and when over expressed at high levels it aggregates strongly, limiting the production of stable mono-disperse preparations. While the reasons for the strong self-association are not fully understood, its relatively low thermal stability seems likely to be one. The major CF causing mutation, ΔF508, renders the protein very thermally unstable and therefore a great deal of attention has been paid to this property of CFTR. Multiple second site mutations of CFTR in NBD1 where F508 normally resides and small molecule binders of the domain increase the thermal stability of the mutant. These manipulations also stabilize the wild-type protein. Here we have applied ΔF508-stabilizing changes and other modifications to generate wild-type constructs that express at much higher levels in scaled-up suspension cultures of mammalian cells. After purification and reconstitution into liposomes these proteins are active in a locked-open conformation at temperatures as high as 50 °C and remain monodisperse at 4 °C in detergent or lipid for at least a week. The availability of adequate amounts of these and related stable active preparations of homogeneous CFTR will enable stalled structural and ligand binding studies to proceed.

A novel arsenate reductase from the bacterium Thermus thermophilus HB27: its role in arsenic detoxification

Microorganisms living in arsenic-rich geothermal environments act on arsenic with different biochemical strategies, but the molecular mechanisms responsible for the resistance to the harmful effects of the metalloid have only partially been examined. In this study, we investigated the mechanisms of arsenic resistance in the thermophilic bacterium Thermus thermophilus HB27. This strain, originally isolated from a Japanese hot spring, exhibited tolerance to concentrations of arsenate and arsenite up to 20mM and 15mM, respectively; it owns in its genome a putative chromosomal arsenate reductase (TtarsC) gene encoding a protein homologous to the one well characterized from the plasmid pI258 of the Gram+bacterium Staphylococcus aureus. Differently from the majority of microorganisms, TtarsC is part of an operon including genes not related to arsenic resistance; qRT-PCR showed that its expression was four-fold increased when arsenate was added to the growth medium. The gene cloning and expression in Escherichia coli, followed by purification of the recombinant protein, proved that TtArsC was indeed a thioredoxin-coupled arsenate reductase with a kcat/KM value of 1.2×10(4)M(-1)s(-1). It also exhibited weak phosphatase activity with a kcat/KM value of 2.7×10(-4)M(-1)s(-1). The catalytic role of the first cysteine (Cys7) was ascertained by site-directed mutagenesis. These results identify TtArsC as an important component in the arsenic resistance in T. thermophilus giving the first structural-functional characterization of a thermophilic arsenate reductase.