Applications of isothermal titration calorimetry in pure and applied research from 2016 to 2020

1 | INTRODUCTION

The applications for analytical technology are driven by the pressing research questions of the day. The first high sensitivity isothermal titration calorimetry (ITC) instruments to be commercially available were released in 1990 and the early publications focused on biological applications of this technology.1,2 In subsequent years, our biochemi- cal knowledge underwent massive growth as the human genome was sequenced and structural genomics determined the 3D structures of an enormous library of proteins. ITC played an ever-increasing role in characterizing proteins’ interactions with other proteins, other bio- macromolecules, and with small molecules including drugs. This was aided by the publication of a series of papers explaining the method, the mathematics, the interpretation of the data, and methods for overcoming technical limitations of the technology.3-7 From a scien- tific perspective, it was the use of ITC to study multi-protein com- plexes that has made the greatest contribution to our knowledge in cell biology. From a societal or commercial perspective, it was the use of ITC in drug discovery and understanding drug-target interactions that has arguably made the greatest contribution.

Since 1990, there has been a steady rise in publications citing iso- thermal titration calorimetry until 2018 when the output of publica- tions peaked (Figure 1). Protein chemistry continues to dominate ITC use in research (Figure 2). In the period 2016 to 2020, 30 years after ITCs release, it might be expected that microcalorimetry would be stagnating. This is not the case. In protein chemistry, there have been significant advances in data analysis of ITC study of multi-protein complexes, and the use of ITC to study interaction and reaction kinet- ics. In synthetic chemistry, ITC has played an ongoing role in deter- mining the functionality of supramolecular structures and has future scope to determine the mechanisms for binding interactions and the role of solvents in this process. The application of ITC, previously used to study biomacromolecules to studying synthetic macromolecules is relatively straightforward extrapolation. The application of ITC to study interactions between small molecules is less obvious. It is worth remembering that the ITC is simply a titration method that will detect any interaction where there is a sufficient change in enthalpy (ΔH) for the instrument to measure. ITC is not limited to studying macromole- cules or even aqueous systems. ITC can study interactions between small molecules, interactions that play a key part in determining the effectiveness of processes like liquid-liquid extraction and extractive distillation. For the cell biologist, interactions between small molecules are often overlooked and phenomena like molecular crowding and osmolyte control of osmotic pressure in halophiles are poorly under- stood. Interestingly, despite 30 years of undertaking ITC research using the very sensitive microcalorimetry instruments, the physical chemistry of intermolecular interactions still presents some chal- lenges. The apparent enthalpy/entropy compensation observed for binding interactions between a target and closely related ligands are still subject to debate.8-11

There were more than 3000 articles reporting the use of ITC between January 2016 and December 2020 so the authors have selected approximately 180 that they feel best to represent the field and acknowledge many excellent papers were omitted from the selec- tion. These references have been classified into the following broad categories:

2 | PROTEIN CHEMISTRY

2.1 | Complex formation and allostery

ITC has played a role in understanding the formation of multi-protein complexes, confirming binding interactions, determining the stoichi- ometry and order of assembly. The field of biomolecular self-assembly has benefitted from an orthogonal approach wherein ITC was the gold standard for determining binding, while NMR, SAXS, and X-ray crys- tallography were used to generate a detailed understanding of mul- tiprotein complex structure, dynamics, and function. The last 5 years have yielded a series of articles that are the culmination of many years of research on multi-protein complex formation. Endosomal transport is reliant on a series of multi-protein complexes and has been the sub- ject of a series of excellent reviews including caveolae formation,189 retromer and retriever function in endosomal trafficking,190 and retromer-independent endosomal cargo recycling.191

Three recent studies nicely illustrate the role ITC can play in teas-
ing out the details of protein complex formation. Proteolysis-targeting chimeras (PROTACS) are involved in selective protein degradation by bringing the target protein into proximity with E3 ubiquitin ligases, which in turn attach a ubiquitin molecule to the target, leading to sub- sequent degradation in proteasomes. ITC demonstrated that the for- mation of the ligase-PROTAC-target complex occurs cooperatively.41 The second example is the use of ITC alongside NMR to observe the role of phosphorylation of the mitochondrial autophagy receptor Nix in promoting the formation of complexes with LC3B. This is an impor- tant step in Nix interaction with LC3/GABARAP, which leads to the degradation of mitochondria in autophagosomes.70 In a third study, ITC was used alongside NMR, x-ray crystallography, and molecular dynamic simulation to study the tetradecameric caseinolytic protease (ClpP) complex. The proteasome inhibitor bortezomib mimics a pep- tide substrate by binding to the serine in the ClpP active-site inducing allosteric activation of the enzymatic complex.40

The methodology for performing ITC analysis of multiprotein complex formation has also advanced. Protein complexes and their interactions with ligands are often substantially more complicated than the simple 1:1 binding reactions frequently studied by ITC. For example, the energetic dissection of protein complex formation usu- ally requires multiple ITC experiments focusing on different steps in the assembly pathway. Interpretation of these experiments in terms of a global model of assembly is best achieved by fitting all the datasets simultaneously to a single set of thermodynamic parameters that describe the overall process. There are a number of different software packages capable to some extent of globally fitting multiple ITC datasets, including the native software provided with MicroCal (Malvern Panalytical) instruments, and the AFFINImeter package (S4SD).26 Notable among these is a suite of open-source software for simultaneous analyses of ITC and multi-modal datasets developed by Schuck and co-workers at the NIH. In 2016, they published an article in Nature Protocols20 with step-by-step instructions for combined analysis of ITC data using NITPIC (which performs an analysis of ITC peak shapes to improve the accuracy of integrated heats),192 SEDPHAT (which performs the simultaneous analysis),193,194 and GUSSI (which analyses the uncertainties in the extracted thermody- namic parameters).195 Together, these articles have been cited more than 500 times in 2016-2020, according to Clarivate Analytics Web of Science, pointing to considerable uptake by the community (Figure 2).

SEDPHAT-based simultaneous analysis was recently applied to better understand T cell activation in the immune response.64 One of the early steps in transducing the signal from T cell antigen receptors is the activation of the enzyme phospholipase-Cγ1, which is recruited to the plasma membrane in complex with three other proteins. This hetero-tetramer is of interest as a drug target for suppressing T cell activation with potential applications in treating autoimmune diseases and transplant rejection. Samelson and co-workers combined analyti- cal ultracentrifugation data together with three binary ITC titrations (monomer into monomer), four ternary titrations (monomer into dimer), and one quaternary titration (monomer into trimer). They found that the last step in the formation of the quaternary complex was accompanied by a large favorable binding enthalpy that was almost completely compensated by a large unfavorable binding entropy. This is a hallmark of coupled folding and binding in intrinsi- cally disordered regions (several of the components have large unfolded regions in the monomeric state) and is thought to help pro- tein/protein interactions achieve high specificity without becoming overly tight, among other advantages.196,197 ITC was uniquely able to detect this assembly-driven folding event since it directly gives the heats of binding; the phenomenon would be virtually invisible to methods that measure affinity only, because the enthalpic and entro- pic components cancel out leaving a net free energy change of almost zero.

Similarly, there have been recent advances in using ITC to unravel complex allosteric communication networks underlying the binding interactions of high-order multimers. Foster et al used ITC to investi- gate interactions of the trp RNA-binding Attenuation Protein (TRAP) with L-tryptophan.24 TRAP forms a ring-shaped complex of 11 identi- cal subunits, which binds up to 11 molecules of tryptophan and inter- acts with RNA to suppress transcription of the trp operon when tryptophan is in abundance. Previous work had shown that TRAP binds tryptophan with moderately positive cooperativity, raising the question of what kind of binding model is appropriate for this system. Combinatorial calculations show that there are 125 unique patterns of bound and unbound subunits in a circular complex with 11 binding sites.24 According to binding polynomial theory,197 the equilibrium is described with as many as 11 sets of change in enthalpy (ΔH) and change in entropy (ΔS) binding parameters. Thus, at first glance, the system could appear recalcitrant to detailed characterization. Never- theless, Foster and co-workers showed that ITC data collected over a range of temperatures could be well fit with a simple model compris- ing three distinct types of binding event: binding to a subunit flanked by either two empty subunits, one bound, and one empty subunit, or two bound subunits. Values of ΔH, ΔS, and change in heat capacity (ΔCp) were obtained for each of the three types of reaction. Binding to subunits adjacent to one or two bound subunits occurred with about 5-fold higher affinity than when both adjacent subunits were empty. Furthermore the magnitude of ΔCp was about 6-fold larger and nega- tive when at least one of the adjacent subunits was already bound, compared to the case when both adjacent subunits were empty, pointing to the existence of allosterically-modulated conformational changes linked to tryptophan binding that could trigger interactions with RNA.

2.2 | Drug discovery

The last 30 years have seen major changes in the way drugs are devel- oped. X-ray crystallography enabled the three-dimensional structures of thousands of proteins to be determined with atomic resolution. This in turn enabled drugs to be designed to interact with structurally well-defined binding sites. Biophysical techniques were applied to study the resulting target/drug binding. X-ray crystallography is rou- tinely used with crystals of the target/drug complex to derive the geometry of the binding interaction. Nuclear magnetic resonance (NMR) is also used to determine the structures of the drug targets with the candidates, in-situ. ITC gives the association constants and the thermodynamic parameters for the binding interactions and sur- face plasmon resonance (SPR) yields the association and dissociation constants for the binding interaction.18 Both ITC and SPR determine the equilibrium constant of the binding interaction from which the
change in free energy (ΔG) of binding can be calculated [Equation (1)]. ITC provides the observed ΔH and the calculated —TΔS of binding [Equation (2)].

Historically, enthalpy-driven binding was preferred over entropy- driven binding and was seen as the key to better drug candidates. The theory was that ΔH is due to hydrogen bonding or similar non-covalent bonding and quite location specific, entropy on the other hand was due to water displacement and quite nonspecific.199,200 In principle, this is correct but in practice, it is dependent on the binding site. An apolar part of the drug candidate that localizes to an apolar portion of the target binding site can be beneficial but contributes to —TΔS not ΔH. In a paper from 2010, Ladbury et al200 (all eminent ITC scientists) defined the ΔH of binding as “a direct measure of the net change in the number and/or strength of the non-covalent bonds on going from the free to bound state.” The paper then went on to illus- trate how statins and HIV protease inhibitors had been improved over time by shifting from entropy-driven to enthalpy-driven binding. If the interpretation of ΔH was simply a measure of the number and/or strength of the non-covalent bonds, then modifying the drug candidate to enable additional or stronger non-covalent bonds would result in greater ΔG of binding and stronger association. A complicating fac- tor was the observation of entropy-enthalpy compensation (see later section) where the improvement in ΔH was offset by a corresponding change in —TΔS. ITC data are used during the later stages of the drug development pathway where they are combined with X-ray crystal- lography, NMR, and SPR to aid in the choice of lead candidates from the panel of hits identified during screening, and subsequently during the lead optimization process. Decision-making based on ITC data in isolation would be a mistake.

One factor that is often overlooked when studying interactions like drug candidate binding to a target is the role of the solution com- position. Intriguing drug binding data produced at different tempera- tures and in the presence of different buffers demonstrates that the ΔG of a binding may be unchanged, while both ΔH and —TΔS can be affected.16 Entropy-enthalpy compensation was evident in all of these cases. It serves as a useful reminder to ensure solution conditions should be as “natural” as practical to make rational decisions on drug selection based on ΔH and —TΔS values.

The challenge of interpreting the thermodynamic data is worth meeting. Protonation states and water effects have been overlooked in the past but play a role in bind- ing that cannot be ignored. The beauty of the ITC is that it illustrates the differences between apolar and polar interactions as it can detect the change in water molecules and the transfer of protons, elements in drug binding that are invisible using X-ray crystallography, NMR, and SPR.

2.3 | Interaction and reaction kinetics

ITC is predominantly used to measure the thermodynamics of binding and assembly processes, however, there is growing interest in using it to measure the kinetics of reactions as well. These applications are based on the idea that the raw ITC output gives a nearly (but not exactly) real time readout of the rate that heat is being generated or absorbed in the sample cell. The heat flow due to a chemical reaction is directly proportional to its rate multiplied by the change in enthalpy. Thus, ITC datasets can be readily interpreted in terms of reaction velocities. As most reaction enthalpies are non-zero, ITC represents, in principle, a nearly universal way to characterize the rates of chemi- cal and biological processes. The primary complication in these types of analyses is that there are several factors that contribute to the shape of the ITC signal in addition to the kinetics of the reactions themselves. These include the rates of heat transfer within the sample cell and its contents, the electronic response of the feedback circuitry, and physical mixing of the injected titrant with the analyte.7,169,201 As a result, a rapid burst of heat in the sample, generated for example by a rapid (sub-second) injection of ligand into a concentrated solution of binding partner, is detected as a peak lasting tens of seconds. The instrument response is often represented by the single time constant, τITC, which describes the rate of exponential decay of the ITC signal following a reference heater pulse or calibration injection.7,202 Note that these methods can give slightly different results, likely because the heat transfer steps differ in the two cases: heat is generated in the wall of the sample cell during a heater pulse and in the sample solu- tion itself for the injection.203 While this description holds for longer time points, the shape of the ITC response at short times is more com- plicated.169 Regardless of the level of detail of the description used, the delayed response of the ITC instrument places an upper limit on the rates of reactions that can be characterized by ITC and must be accounted for quantitatively in any rigorous analyses of ITC kinetic data.

One particularly interesting aspect of ITC kinetics analysis is that a wealth of potentially informative kinetic data has been hiding in plain sight. Many ITC datasets for 1:1 binding interactions contain the distinctive feature wherein peaks become broader near the equiva- lence point of the titration (equal mole fractions of binding partners), with all other injection parameters held constant. Egawa et al202 and subsequently Dumas et al6 showed that this peak broadening could be quantitatively explained in terms of the association and dissocia- tion rate constants, kon and koff. For a simple reaction A(cell) + B(syringe) — AB, the A component is in excess early in the titration and B is consumed with a pseudo-first order rate constant kon[A]. The concentration of free component A drops throughout the titration as it is converted to AB, leading to successive reductions in the rate of binding and broader peaks. Near the end of the titration, the B com- ponent is in excess and A is consumed with the pseudo-first order rate constant kon[B]. As free [B] accumulates, the forward rate increases once again and peaks narrow. Thus, conventional ITC exper- iments can encode kinetic information that is untapped in typical ther- modynamics analyses in which only the peak integrals are considered. The researchers provided two alternative approaches to extracting kinetic information, which they refer to as kinITC6 and kinITC-ETC (Equilibration Time Curve).21 In the first, raw ITC data are fitted directly to chemical kinetic equations together with those accounting for reagent mixing and the instrument response. This approach has the advantage of being completely general in the type of reactions that can be studied. For example, kinITC was used to study thiamine binding to two different E. coli TPP riboswitches and was able to resolve the kinetics of both of the binding step and a sub- sequent RNA folding event.6,106 The drawback of this method is that the mixing and the instrument response must be well accounted for in order to faithfully reproduce experimental ITC peak shapes, which can present considerable challenges, as mentioned above. Possibly due to the relative complexity of the analysis, the examples of full kinITC peak fitting are thus far limited to HIV protease/Nevirapine and TPP riboswitch/thiamine interactions.6,21 Alternatively, the simplified kinITC-ETC method analyses only the lengths of time required for ITC peaks to return to baseline following the injections (τend). For systems giving useful kinetic data, the values of τend are shorter for the early and late injections and longest near the middle of the titration, for the reasons mentioned above.21 The τend vs injection number profile is governed exclusively by kon (or koff) and KD (along with concentrations, injection volumes, etc.) The value of KD is obtained from the tradi- tional thermodynamic analysis. Thus fitting of the τend profile gives one of the rate constants while the other is calculated from KD=koff/ kon. The advantage of this approach is that the complicated effects of mixing and instrument response primarily affect the early portions of ITC peaks, while the overall lengths of the peaks are dominated simply by the binding kinetics (provided they are sufficiently slow). As a rough guide, the authors recommend that in order for kinITC-ETC to be successful, the following inequality should hold: τITCpkffiffiffioffiffifffifffikffiffiffioffinffiffi[ffiffiAffiffiffi]ffi0ffiffi < 5, where [A]0 is the total amount of binding partner,that is, [A] + [AB].6 Representative values of τITC vary between about 3.5 seconds for the Malvern ITC200 to 15 seconds for the Malvern VP-ITC. For a Wiseman c value of 10 ([A]0 = 10koff/kon), this means that koff should be slower than about 0.5 to 0.1 s—1. Importantly, kinITC-ETC results are generally in good agreement with those of SPR6,21,29,39,106 and fluorescence spectroscopy,99 providing validation for the approach. The main disadvantage of this approach is that is only applies to simple 1:1 (or 1 to multiple identical sites) binding reac- tions, as much of the granular detail in the data is deliberately discarded.

Recently, Dumas has partnered with the S4SD company to include kinITC-ETC analysis as a standard feature of the AFFINImeter-ITC software package, extending the accessibility of the method. The program has been used to study the binding kinetics of aptamers,99 carbonic anhydrase inhibitors,21,43 HIV reverse transcrip- tase inhibitors,21 the nickel binding to the NikR transcription factor,39 aminopolycarboxylate ion binding,166 and the lectin domain of FimH, an E. coli virulence factor.29,49 This last case is notable in that ITC data had already been obtained for a panel of potential mannosidic inhibi- tors and used to elucidate structure-activity relationships based on binding affinity. Ernst et al re-mined the existing data to extract bind- ing kinetic information, discovering that slow on-rates correlated with H-bond formation while rapid on-rates correlated with long-range electrostatic forces and conformational restrictions.29 There is grow- ing awareness in the drug development field that efficacy is related to both the strength and the kinetics of drug binding,204 thus structure- kinetics relationships, such as those uncovered here, are of great interest. The fact that both thermodynamic and kinetic information can be obtained simultaneously in a straightforward manner has impli- cations for the future role of ITC in drug development campaigns.

2.4 | Enzyme catalysis of single reactions

Another area where the use of ITC to measure chemical kinetics has grown is in the characterization of enzyme kinetics. According to our search of the literature, there were about 10, 8, 28, and 37 papers published on this topic in the 5-year periods beginning in 2001, 2006, 2011, and 2016. In these applications, the heat measured by the calo- rimeter is generated by the chemical reaction catalyzed by the enzyme rather than by binding and concomitant structural changes, as is the case for the majority of ITC experiments. This has several implications for experimental design. Firstly, the fact that each enzyme typically performs multiple catalytic turnovers during the ITC experiment effec- tively amplifies the heat signal, and the amount of enzyme required is far less (as low as picomolar)168 than that required for binding experi- ments. In addition to the advantages of requiring less material and dis- couraging enzyme self-association, this also means that the heats generated by association and dissociation of substrates, products, and inhibitors can be safely ignored, as they are far smaller than the heats of catalysis. Secondly, the fact that an enzyme provided with suffi- cient substrate can continue to generate heat signals almost indefi- nitely opens the door to a wide variety of experimental designs. All ITC enzyme kinetics measurements rely on the simple relationship wherein the displacement of the ITC signal from the no catalysis base- line is directly proportional to the velocity of the reaction. Thus, the ITC output represents a (nearly) real-time readout of enzyme activity, subject to the caveats of mixing and instrument response times described above. Beyond this commonality, a diversity of approaches is possible.In many cases, enzyme kinetics are well described by the following modified Michaelis-Menten equation [Equation (3)]205 after each injection is known from the concentration of substrate in the syringe and volumes of all injections. The reaction velocity can be read directly from the vertical position of each step, tracing out a complete ν versus [S] Michaelis-Menten curve, which can be fitted to the modified Michaelis-Menten equation. In practice, we find that the condition of negligible substrate consumption is met when [E]0 ≤ (10—4 s) × Km/kcat.

In a single injection ITC kinetic assay, the amount of enzyme is typically chosen to be large enough so that the substrate can be fully converted to product on the timescale of minutes or tens of minutes. The concentration of substrate is chosen so that enzyme is initially saturated following (each) injection ([S]>>Km).206 Single injection assays can be initiated either by injecting substrate (syringe) into the enzyme (cell) or by injecting enzyme (syringe) into the substrate (cell). In either case, the ITC signal exhibits a large deflection immediately after the injection with large heat flows continuing as long as the enzyme remains saturated with substrate. The signal gradually returns to the pre-injection baseline as the substrate is consumed. When the substrate is present in the syringe, this procedure may be repeated several times within the same experiment. When enzyme is in the syringe, only a single injection can be made, as all the substrate is con- sumed after the first injection. Once again, the velocity of reaction (ν) can be read directly from the displacement of the ITC signal as a func- tion of time. The amount of substrate at each time point is known from partial integration of the ITC peak (the fraction of total substrate remaining at time t is equal to the fraction of the total area of the peak lying to the right of time t). Together the ν and [S] values trace
out a complete Michaelis-Menten curve. We find that substrate is where ν is the rate of the reaction (-d[S]/dt), [S], [E]0, and [I] are the concentrations of substrate, enzyme, and inhibitor, respectively. kcat is the maximum number of reactions catalyzed by each molecule of enzyme per unit time, Km is the concentration of substrate required to reach half-maximal velocity in the absence of inhibitors, and Ki and K’I are dissociation equilibrium constants for inhibitor binding to the enzyme and enzyme-substrate complex. Most enzyme assays aim to determine the values of these kinetic parameters. In their seminal 2001 paper, Todd and Gomez describe two general strategies, which are now generally referred to as “multiple injection” and “single injec- tion” experiments.4 In a multiple injection ITC enzyme kinetic assay, the sample cell contains an enzyme solution and the syringe contains the substrate. The enzyme concentration is chosen to be sufficiently low so that substrate depletion during the experiment is negligible. As a result, the ITC signal is ideally constant (horizontal) between sub- strate injections and resembles a series of steps, one per injection. The injections are designed such that early steps have [S]<<Km and the final injections have nearly saturated the enzyme with [S]>>Km. The displacement of each step relative to the initial baseline is directly proportional to the reaction velocity, such that each step is smaller than the one preceding it as the enzyme is gradually saturated with substrate. The concentration of substrate present in the sample cell that the return to baseline is difficult to distinguish. However, the enzyme must be at a low enough concentration so that the return to baseline takes at least seconds to tens of seconds. More rapid reac- tions start to become obscured by the instrument response, as men- tioned above.169 Compared to multiple-injection experiments, the single injection method is more rapid and allows kcat and Km to be extracted from as little as tens of seconds of raw ITC data. As well, a great deal more product is generated in a single-injection experiment, which can be advantageous for researchers seeking to characterize product inhibition but is a disadvantage for those seeking to avoid this complicating factor.77

The multiple-injection (dilute enzyme) and single-injection (con- centrated enzyme) approaches provide frameworks for developing new types of ITC kinetic assays. In a recent example, dilute enzyme and substrate were both placed in the sample cell with an inhibitor in the syringe, leading to a constant reaction velocity and horizontal ITC signal prior to the first injection. Each injection of inhibitor reduced the velocity and heat flow, but the change in reaction rate occurred over a period of several minutes as the inhibitor gradually bound to the active site of the enzyme. These time traces were fit to extract the inhibitor association and dissociation rate constants, providing an efficient method for rapidly determining inhibitor binding kinetics.168

In another example, in a modified single injection experiment, the sub- strate was placed together with inhibitor in the syringe and injected multiple times into concentrated enzyme in the sample cell.170 Each injection generated a peak, each of which was fitted to obtain values of kcat and Km. The inhibitor accumulated in the sample cell with each injection, thereby generating datasets of kcat and Km as functions [I], allowing determination of Ki and K’I (modified Michaelis-Menten equation) that is, completely characterizing the strength and mode of inhibition. The method has since been applied to inhibitors of epoxide hydrolase, which is of interest as a drug target.30 Notably, the previ- ous (non-ITC) assay for this enzyme involved hyphenated liquid chro- matography and mass spectrometry, which involves considerably more time and experimental uncertainty than the ITC-based method.

ITC offers many advantages compared to traditional enzyme assays. Firstly, ITC offers real-time monitoring of enzymatic reactions in cases where other types of continuous assays are unavailable, such as for epoxide hydrolase, above.30 It is also unique in that traditional enzyme assays measure concentrations of substrates and products, with rates determined indirectly. ITC directly detects the reaction velocity, enabling new applications such as the inhibitor binding kinetic experiments described above.168 The ability to employ natural substrates is another large asset for ITC, as traditional assays often involve modified colorigenic or fluorogenic substrate analogs that do not necessarily have the same kinetic parameters as the native sub- strate.58 ITC also offers advantages for enzymes where the standard assay involves indirect readout with a coupled-enzyme system.207,208 This is particularly true when adding co-solutes or inhibitors that affect enzymatic activity since the secondary enzymes can be affected as well as the enzyme of interest. In addition, testing spectroscopically active inhibitors or other effector molecules can become a challenge when using spectrophotometric assays, that is, with chromogenic or fluorogenic probes, or with coupled assays. In contrast, deeply colored inhibitors are fully compatible with ITC inhibition assays.209 Further- more, ITC’s ability to characterize opaque samples further extends the reach of this technique beyond spectroscopically-accessible systems. One particularly interesting example involves using ITC to perform enzyme kinetics experiments on suspensions of living cells.59,63,79,85 For instance, Yang et al used ITC to study the metallo-β-lactamase NDM-1 in living cultures of E. coli. NDM-1 cleaves carbapenems, pro- viding bacterial resistance to these “last resort” β-lactam antibiotics.85 Development of NDM-1 inhibitors has the potential to resensitize resistant bacteria and offers an avenue for treating these kinds of seri- ous drug-resistant infections. Single injection experiments with cefazolin as a test substrate in the syringe and either purified NDM-1 or live E. coli bacteria expressing NDM-1 in the sample cell gave very similar ITC heat signals. The authors went on to show that ITC could diagnose the presence or absence of active NDM-1 in clinical strains of pathogenic bacteria85 and used the in vivo ITC experiment to dis- cover a new inhibitor scaffold.53 A similar approach was used to iden- tify the presence of extended-spectrum β-lactamase activity in a panel of 69 clinical Enterobacteriaceae samples, yielding results in just an hour, compared to the 32 to 48 hours typically required for traditional growth inhibition assays.59

2.5 | Enzyme catalysis of multiple reactions

To understand the kinetics of hydrolytic enzymes (glycoside hydro- lase) that digest homopolymers like starch, cellulose, and chitin, you need to consider the sequence of events that occur during cataly- sis. Non-processive glycoside hydrolases usually follow the sequence, (a) attach to the polymeric substrate, (b) hydrolyze the bond, (c) separate from the polymer, (d) release the cleavage prod- uct, then reattach to the polymer and start the process again. These enzymes usually follow classic Michaelis-Menten kinetics, an exam- ple being the hydrolytic cleavage of maltoheptose by human saliva α-amylase.58 Substrates that are crystalline homopolymers, such as chitin or cellulose are often structurally difficult to attack chemi- cally or enzymatically due to the tightly packed polymers locked together by hydrogen bonds. The ability of processive enzymes to attach to the polymer prevents them from diffusing away from their target and makes them more efficient than non-processive enzymes at digesting these difficult substrates. Attachment of a processive enzyme to a substrate can be studied using ITC. Serratia marcescens chitinases, ChiA and ChiB binding to (GlcNAc)6 was demonstrated by ITC using catalytically inactivated enzyme.50 Processive enzymes generally don’t follow classic Michaelis-Menten kinetics as the rate of catalysis is partly determined by the level of crystallinity of the substrate, which can be heterogeneous across different regions of a single sample. Hydrolytic enzymes that break down heteropolymers can follow Michaelis-Menten kinetics acting on homogenous model substrates (eg, trypsin/Nα-Benzoyl-L-arginine ethyl ester BAEE). Digestion of heterogeneous substrates is more complicated. A single injection assay using ITC to study trypsin hydrolysis of casein predictably deviates from Michaelis-Menten kinetics.67 Small trypsin substrates with a single cleavage site (eg, BAEE) are the method of choice for studying phenomenon like molecular crowding as the interpretation of the results is easier than a complex substrate like casein.67

Digestion of phytate (myo-inositol hexakisphosphate) by Citrobacter braakii 6-phytase has five discrete cleavage steps and the reaction deviates radically from Michaelis-Menten kinetics.54 The thermogram produced using a single injection of phytate into phytase had two peaks consistent with two periods of high enzyme activity with a bottleneck at the third cleavage step.

2.6 | Thermostability of proteins

One of the more innovative uses of ITC published in the last 5 years was a paper describing the analysis of the thermostability of lysozyme using the ITC in isothermal mode without titration.71 Traditional calo- rimetric measurements of thermostability employ differential scanning calorimetry (DSC) where the instrument measures the heat flow required to increase the temperature of a protein sample at a constant rate. As the temperature rises, the protein unfolds, leading to an increased heat absorption (or heat capacity, Cp) over the transition region. The temperature corresponding to the peak (maximum Cp value) in the unfolding thermogram (Tm) is used as a measure of pro- tein stability.

The ITC research studied the irreversible unfolding and subse- quent aggregation of hen egg lysozyme in a buffer at pH 9.0, using a TA Instruments TAM IV.72 The temperature in the ITC was set at 57◦C, 58◦C, and 59◦C, that is, below the Tm of 71◦C, and followed the heat flow over 6 days. The experiment demonstrated an alternative
method for studying the thermal stability of protein formulations at temperatures that are more representative of physiological or ambient conditions than those used in a DSC. The technique has merit for the development of stable protein formulations and has also been demon- strated for immunoglobulins36 which are the leading family of thera- peutic proteins.

3 | SYNTHETIC CHEMISTRY

3.1 | Supramolecular structures

Great advances have been made in the field of supramolecular chem- istry and an incredible diversity of supramolecular structures have been synthesized. The molecular cages are an important family within the supramolecular structures. These three-dimensional hollow struc- tures include the crown ethers, cryptands, calixarenes, resorcinarenes, curcurbiturils, cyclodextrins, pillararenes, cryptophanes, and hemi- cryptophanes.155 The cavities within these structures provide envi- ronments with specific dimensions and electric fields that will only allow a limited number of molecules to enter. This ability to interact with specific atoms or molecules provides a degree of specificity that makes the supramolecular structures potentially useful as sensors, drug delivery vehicles, catalytic sites, and as smart interlocking materials.

A range of analytical techniques are used to study supramolecular structures including X-ray crystallography to determine the molecular structure, NMR to ascertain the positioning of binding partners within the supramolecular structure, and ITC to study the stoichiometry and thermodynamics of the interaction. ITC is usually confined to deter- mining the association constant, the stoichiometry, and the change in enthalpy and entropy for the interaction although in principle it also could be used to study the kinetics of interactions and reaction kinet- ics if the molecule has the catalytic capability (as described above).

Early work with ITC tended to focus on macromolecules of bio- logical origin such as proteins, nucleic acids, lipids, and complexes, which made it difficult to interpret experimental data at the molecular level. The chemical simplicity of supramolecular structures enabled them to be used as research tools to study intermolecular interactions and study specific interactions. Research using supramolecular struc- tures provides a unique opportunity to study the energies associated with hydrogen bonding, electrostatic interactions, π π-stacking, cation π and anion π interactions, and water displacement (hydrophobic effect).210 The cavities formed in supramolecular structures represent excellent model systems for studying phenomena like the exclusion of water from hydrophobic pockets or constricted spaces145 or interactions between low charge density ions and hydrophobic pockets.144 For example, the presence of low charge density ions in a hydrophobic pocket is not driven by electrostatic interactions, since the interacting partners are charge neutral, but is due to the energeti- cally favorable displacement of water from this interface. The knowl- edge gained by studying interactions with non-biological model structures can be transferred to more complex systems like proteins, nucleic acids, and macromolecular complexes.

Supramolecular structures that are luminescent or fluorescent with specific recognition capabilities for molecules such as gasses, ions, metabolites, drugs, etc. have potential applications as sen- sors.136,211 ITC is a useful in these efforts when the analyte is soluble and can be employed to characterize closely related molecules and to ascertain the selectivity molecular recognition.

Drug delivery is an attractive application for supramolecular structures. Many therapeutic agents are relatively hydrophobic and have low solubility in water or blood. Encasing the therapeutic agent in a molecular cage has been used to improve drug solubility in an aqueous environment.212 For example, a new, highly insoluble antidiabetes drug was successfully solubilized by forming inclusion complexes with β-cyclodextrin and hydroxypropyl-β-cyclodextrin.146 Similar cyclodextrin inclusion strategies have been used with a range of drugs, terpenes, and steroids.134 Cyclodextrins have also been pro- posed to deliver drugs for release in the intestine where bile salts help displace the drug from the complex, as demonstrated in vitro,137 and have also been developed for siRNA delivery.143

Supramolecular structures have been used to create artificial enzymes.148,150 ITC was used to study oxidase-like catalysis using the multiple injection technique developed for studying enzyme kinetics (see above).152 The possibility of designing synthetic enzymes with better stability and efficacy than biological enzymes is an obvious goal with great potential benefits.

Supramolecular polymer networks (SPNs) are polymers cross- linked by non-covalent interactions.213 These structures differ from typical covalently linked polymer networks in being capable of revers- ible self-assembly and disassembly, as dictated by physical stress and environmental changes. This leads to interesting properties such as superior toughness, durability, and self-healing. ITC is uniquely suited to measuring the stoichiometry of polymers cross-linked within the supramolecular structures.

4 | PHYSICAL CHEMISTRY

4.1 | Enthalpy/entropy compensation

There are many examples where chemical changes to either the host or guest in a binding interaction have resulted in an increase in ΔH but was accompanied with a corresponding increase in TΔS that minimized the change in ΔG.9 This enthalpy-entropy compensation is an interesting phenomenon that has puzzled scientists since it was observed in the 1970s.214 The introduction of ultrasensitive ITCs in
the 1990s made studying entropy-enthalpy compensation easier and raised its prominence within the scientific community.8,9 The consen- sus is that the solvent plays an important role in entropy-enthalpy compensation10,11 but definitive evidence for this hypothesis has yet to be found.

Schönbeck and Holm published research in 2019, where they used a relatively simple synthetic model system to experimentally study entropy-enthalpy compensation.140 They used adamantane derivatives binding to the cavity in cyclodextrins with hydroxypropyl groups attached to the rim of the cyclodextrin to increase hydropho- bic contacts made during the binding interaction. The interaction was enthalpy-driven partly due to the semipolar nature of the cyclodextrin and possibly due to the release of water constrained within the tight cavity. The experiment determined the ΔH, TΔS, ΔG, and ΔCp of the different binding interactions between the adamantane derivatives and the modified cyclodextrins. While the experiment did not provide a definitive confirmation that dehydration of apolar moieties is the cause of observed entropy-enthalpy compensation, it does illustrate a rational approach to this research. When combined with molecular dynamic simulation this approach could well solve the puzzle of entropy-enthalpy compensation.

5 | NEW AREAS OF ITC RESEARCH

5.1 | ITC analysis of small molecule solvent and solute interactions

For historic reasons, there is a tendency to think of ITC as being useful for studying interactions involving macromolecules usually of biologi- cal origin. This is partially due to the release of the first commercially available microcalorimeters at the time when structural genomics was emerging as an important area of research. The importance of interac- tions between small molecules is obvious to process engineers where these interactions affect processes like distillation and liquid-liquid extraction.

We have dedicated a section of this review on the research into small molecule interactions. For the cell biologist or protein chemist small molecules are important and often overlooked. An obvious example is the action of osmolytes in regulating in vivo osmotic pres- sure in halophiles, without disrupting cell function. In mammalian cells, small molecules play a role in regulating in vivo ionic strength and vis- cosity that helps maintain cell integrity, and they play a role in molecu- lar crowding. Protein chemists are aware that small molecules alter protein solubility and stability, and have created theories like prefer- ential hydration to explain their observations without studying small molecule-water interactions to verify these theories’ validity. It is interesting that chemical engineers have taken the lead in using ITC to investigate small molecule interactions where interactions like hydro- gen bonding, electrostatic interaction or proton exchange can influ- ence the success of distillation and liquid-liquid extraction processes.

In chemical engineering, the field of separations is an important field, responsible for approximately half of the energy demand of the entire chemical industry.215 Next to traditional distillations, solvent-based separations are common unit operations applied in many appli- cation areas, including oil refineries, bulk chemical production plants, the mining industries, and for purification of fine chemicals and phar- maceuticals. For any solvent-based separation, the separation perfor- mance depends largely on intermolecular interactions between the solvent and the various solutes in the mixture to separate.

Historically, information on intermolecular interactions has been mostly derived indirectly from biphasic equilibrium measurements and assumed interaction mechanisms, sometimes supported by spectro- scopic evidence. For example, by measuring compositions in vapor- liquid equilibria (VLE) or liquid-liquid equilibria (LLE), information on the activity coefficients of the solutes in the solvent phase can be obtained. VLE measurements are labor intensive, and don’t provide enough information to describe the interactions within the system. For example, when strong complexes are formed between the solvent and solute, additional insights are useful to aid solvent selection. Spec- troscopic approaches (such as NMR or FT-IR) have been used to aid understanding of the interactions within these mixtures but it is here that ITC can play an invaluable role.

5.2 | Liquid-liquid extraction and extractive distillation of organic chemicals

ITC is a technique that has been used to study fluid separations, as direct measurement of heat effects arising from titrating a solute into a solvent provides valuable information on the affinity between the solvent and the solute. Two types of solvents may be considered in fluid separations, single molecule solvents, and composite solvents comprised of multiple types of molecules. Typically, single molecule solvents are physical solvents, and although hydrogen bonding and polarity interactions may occur, strong complexations are hardly found. In composite solvents on the contrary, often an extractant is dissolved in a diluent, and the extractant is added with the purpose to form complexes with a solute that is aimed to be extracted or entrained (the typical term in extractive distillation). In the past few years, attempts have been made with ITC to characterize both com- posite solvents and single molecule solvents.183,184

The use of ITC to characterize liquid-liquid separations and link experimentally determined interaction energies directly to extraction performance was pioneered by Cuypers et al in 2008.216 They applied ITC to investigate on the thermodynamics of hydrogen bonding com- plexations of various phenolics with tributyl phosphate (TBP) and trioctyl phosphine oxide (TOPO) extractants. Importantly, they inves- tigated complexation of TBP and TOPO in neat diluents, but also in water-saturated diluents. The impact of water was obvious, and inter- pretation of the role of water in the complexation thermodynamics, as well as the impact that this has on the liquid extraction performance was studied. Especially, considering that the ITC was used to under- stand liquid-liquid extraction to remove phenols from aqueous solu- tions, it is key to understand the impact of water that is being co- extracted, even when it is only present in the organic solution in small concentrations. They found that water interfered in hydrogen bonding between the extractant and the solute, the fitted complexation con- stant was significantly lower than in the neat solvent. The constant obtained with ITC corresponded well with the observed complexation constant based on fitting models to data from liquid-liquid extraction experiments.

In liquid-liquid separations, simple 1:1 interactions often seen in ITC biological applications are atypical. ITC data is ideal for determin- ing the stoichiometry of interactions.183 The work of Sprakel and Schuur183 investigated the solute concentration range in the molar range, which is appropriate for bulk chemical separations. This con- centration range differentiates this work from Cuypers et al216 and ITC experiments undertaken by protein chemists or synthetic chem-
ists. In combination with allowing for complex stoichiometry, this research applied a sequential reaction according to A + B ⇆ AB and A + AB ⇆ A2B and found that obtaining equilibrium constants for the sequential reaction model is possible in the molar concentration range.

Having established and validated the application of ITC to the area of fluid separations,183 Sprakel and Schuur have used the tech- nique to investigate on the impact of diluents on the interactions between extractants and solutes.186 For decades, extraction studies have appeared claiming both hydrogen bonding complexes and proton transfer from the acid to the base216 and often it was unclear which of the mechanisms should be applied. Using a combination of ITC and molecular modeling,186 for each of the diluents applied it could be very clearly seen when ion exchange did occur and when it did not. Active diluents capable of supporting hydrogen bonds such as 1-octanol clearly showed 1:1 stoichiometry and proton transfer, resulting in the larger values of ΔH1,1 observed by ITC, was confirmed by molecular modeling. Such understanding of extractant – solute interactions can also be extended to study the impact of variations in the extractant molecular structure. By investigating on the impact of the molecular structure on the binding of the solute,185 insights were obtained on the temperature dependency of liquid-liquid extraction, which aids process design. Another example where ITC helped under- standing phenomena observed in liquid-liquid extraction systems involved a phosphonium phosphinate ionic liquid.180 In order to understand why it was so hard to recover acetic acid from the ionic liquid, a multitude of techniques including 1H-NMR, 31P-NMR, FT-IR, and ITC were applied. ITC confirmed earlier suggested multi-acid to phosphinate interactions,218 as well as a clear trend in interaction energies, showing strong binding of the first acid.

Aiming at understanding also the impact of entrainers (solvents) in extractive distillation, Sprakel et al184 performed ITC studies on a wide range of entrainers for a variety of binary distillation systems. Although good insights were obtained on the energies required to yield interactions that were sufficiently strong to modify the relative volatility in the extractive distillation, but sufficiently weak to allow regeneration of the complexes, quantification of complexation con- stants was mostly not possible, because the typical S-shaped curves in the ITC data were not obtained. Not being able to quantify complexes is primarily due to the absence of strong complexes when a single molecular solvent is applied. As no complex constants could be fitted, also information on the Gibbs energy was lacking, and as a result, it was not possible to determine a quantitative relation to the VLE with ITC results only.

In conclusion, for the standard fluid separation operations in the chemical industry, using ITC, possibly in combination with other tech- niques such as molecular modeling or spectroscopic techniques, can yield very valuable insights in intermolecular interactions, whether interactions are exothermic or endothermic, and the magnitude of the overall heat effect. Quantification of models describing phase equilib- ria is not always directly possible, and depends on the presence of the characteristic S-shaped ITC curve. For the specific situation that affin- ity separations are addressed, with clear attractive interactions involv- ing the solutes to separate (and typically exothermic heat effects), the typical S-shaped curve allows one to fit one or more complexation constants. From these, the Gibbs energies are easily calculated all- owing quantitative comparisons with phase behavior. If the S-curve is not clear enough, quantification of the Gibbs energy is difficult, and thus far, only qualitative insights have been obtained.

5.3 | Liquid-liquid extraction for mineral processing

To recover metal ions from leachates, liquid-liquid extraction is com- monly applied. Mechanistically, liquid-liquid metal extractions differ from liquid-liquid extraction of most organics, because in contrary to organics, the physical solubility of metal ions organic solvents is typi- cally negligible. Because of the negligible solubility in the organic dilu- ent, doing homogeneous organic phase ITC is thus not an option. Perhaps this limitation to biphasic systems and the corresponding additional degree of complexity is the reason that only a very limited number of studies have appeared to date. As early as 1973, Marcus and Kolarik219 published an article on the use of an ITC for characteri- zation of lanthanide extractions, the machine was based on an 800 mL measurement cell. In subsequent years, their equipment reduced in size so that approximately 100 mL was sufficient.220 Although this early work showed that determination of thermody- namic information including the enthalpy of the complexation reaction in the extraction, the Gibbs energy, and corresponding entropy is pos- sible, most other researchers continued to use other approaches, including several spectroscopic techniques. Around the same time that Cuypers et al216 reported the coupling of ITC with liquid-liquid extraction of phenolics, the topic was picked up again, now with a micro-ITC described in a paper by Zalupski and Nash, who applied a biphasic ITC analysis.221 One key aspect for carrying out ITC in liquid- liquid biphasic systems is to pre-saturate the solvent with the appro- priate aqueous solution (eg, aqueous nitrate solution if the metal ion is leached using nitric acid) to avoid additional heat effects interfering with the measurement.

An important mechanism in metal extractions is chelation, because to overcome the strong endothermic dehydration of the metal ions in their aqueous solution,221 a strong interaction in the organic phase is necessary. A recent study showed that also in the aqueous phase, chelation of metal ions can take place, and the ther- modynamics including enthalpy, entropy, and equilibrium constant for rare earth ion complexations with linear poly(ethylenimine methylenephosophonate) were determined.13 The work by Archer and Schulz13 very nicely displayed that the aqueous phase complexa- tions are entropically driven, and depending on the degree of func- tionalization with phosphonate groups, a stoichiometry of between two and five phosphonate groups per rare earth element ion was observed, the maximum value being reported for Eu(III) and the lowest for Dy(III). Although in this work no extraction was studied, it clearly demonstrated that ITC can be of great value in characterization of the nature and strength of complexes formed with rare earth elements and other metals, and the work may be of use for extending these findings to aqueous two-phase systems. Among other applications, aqueous two-phase systems comprising a polymer that binds metals may be useful for extraction of rare earth elements. This is certainly of interest for the process engineering community related to metals extraction and recycling, and is likely to become a subject of studies in the coming years.

5.4 | Adsorption processing for toxin removal

A totally different application field is adsorbing toxins from aqueous streams, an important field in process engineering. Kato et al177 reported an ITC study on the adsorption of uremic toxins (potassium cresyl sulfate and potassium indoxyl sulfate) on Zr-based metal organic frameworks (Zr-MOFs). It was found by ITC that the complexation of the Zr-MOFs with the uremic toxins was enthalpic in nature while a positive entropic term TΔS was observed. The ITC study was supported by density func- tional theory (DFT) calculations, identifying π-π interactions between the pyridine functionality in the Zr-MOF and the indoxyl groups of the cresyl moiety. This study clearly shows how powerful ITC can be also for char- acterization of adsorbent – adsorbate interactions.

6 | CLOSING REMARK

6.1 | Strengths and weaknesses in published research

Publication of the thermograms in the manuscript or in the supple- mentary data has increased over the last 5 years enabling appraisal of the quality of the data and building confidence in the accuracy of the analysis. Quality of much of the published ITC research is very good and the authors have sensibly limited the interpretation of the data. Probably the most common published error in ITC analysis is failure to let the signal equilibrate before the next injection. This mistake builds a systematic error into the calculated thermodynamic constants. The criticism by Brian Pethica of published ITC research is a cautionary warning to ITC users.222 It also can serve as guidance to help improve the quality of ITC research.15

Studying weaker interactions continue to present a challenge to some research groups. The authors refer the reader seeking assistance in this field to Joel Tellinghuisen’s 2008 paper on low c binding.223

The take home message from this paper is ITC can be used to understand interactions between a range of molecules from nanoparticles down to relatively small solutes and solvents. ITC is not only able to detect interactions but also measure their kinetics. ITC’s capacity to answer research questions is more versatile than most researchers currently utilize. It will be interesting to ACBI1 see if the next 5 years will see yet more innovative uses for ITC.