Compatible solutes (CS) are small organic osmolytes including sugars, polyols, amino acids and their derivatives. They are compatible with cellular metabolism even at molar concentrations. (See figure 1 for a few examples). As reviewed extensively elsewhere 123, CS are found in microorganisms from all three domains: Archaea, Bacteria and Eucarya, but also in higher organisms and are used in a wide range of applications 4. A complete list of disciplines interested in compatible solutes would start with halophilic/osmophilic bacteria 56 and yeasts 7, their bioenergetics 8 and their relevance for bio-remediation 9. But the list would further extend to medical disciplines dealing with for example cancer research 10 or dermatology 1112. Even food science takes an interest in CS research, very recent findings demonstrate that CS can be found as a natural component of food traditionally processed by microorganisms 13. Therefore it is not surprising, that research on solute effects on macromolecules is widely spread. Most of it has been performed in the field of proteins. Beneficiary effects of compatible solutes on proteins in vitro have been extensively studied (e.g. 14151617) as have been effects on protein expression 18 and stabilization of whole cells 1920. Research on protein stability and protein stabilization by compatible solutes has led to the development of some theories (and variations thereof) concerning solute/protein interactions. The four most outstanding among them discuss preferential interaction 21, water replacement 22, water density fractions 23 and osmophobic effects 24 as the mechanisms of solute/protein interactions. However, this short review can not serve as a comprehensive review of these theories and their applications. Therefore I will present and discuss only the underlying ideas and their application to nucleic acids. More attention will be given to recent data relevant for solute/nucleic acid interactions 25 and on the background of these findings.

Beneficial effects of compatible solutes on nucleic acids and nucleic acid/protein complexes are mainly known from improvements in yield and specificity of polymerase chain reaction (PCR), see 26272829303132 for examples. But effects also extend to nucleic acid stabilization 33, improvement of protein/nucleic acid complex formation 34, nucleic acid purification 35 and cell free transcription 3637 as well as modulation of restriction enzyme function 3839. Contrary to other well known effector molecules like polyamines which stabilize negatively charged macromolecules due to their cationic nature 40, the mode of interaction of zwitterionic, anionic and uncharged low molecular weight compounds with nucleic acids is barely understood. There are some obvious possibilities how they might influence nucleic acids. Among them are changes in the electrostatic environment 41, intercalation 42 and a role as anti-intercalators 43.

In this review I will start with an overview of known effects of compatible solutes on nucleic acids, focusing on double stranded DNA since a wealth of data can be retrieved from this area of work. Still "DNA comes in many forms" 44, a fact we are aware of since 1957 45, only four years after Watson and Crick presented their theory of the double stranded DNA helix 46. Therefore I am also going to present more complex structures like triple- and quadruple helices. Considering the importance of riboswitches 47 and the recent advances which have been made in this field 48 interference of small metabolites with RNA has become of prime importance. Therefore RNA interactions with small osmolytes, be they direct 49 or indirect, might play an essential role in regulation of compatible solute biosynthesis and adaptation. After discussing potential models for molecular mechanics of interactions, and first steps towards applying those models to nucleic acids, I will address the questions still unanswered. A few thoughts will be given towards computational methods before I draw my conclusions and give some future prospects.

Inorganic ions are not meant to be a major subject of this review. Their influence on nucleic acids, especially RNA, and their structure has been reviewed thoroughly by Draper 50 and newer findings, including CS effects, have been presented recently 25. Nevertheless, we have to consider effects of inorganic ions when interpreting experimental data since they are an indispensable component of buffers and the native environment of nucleic acids. Most obvious is their influence on DNA stability (see next section), which is reflected in those melting temperature (T_M) calculations that consider ion concentrations or solvent ionic strength. The approach of Frank-Kamenetskii 51 may serve here as an example:

T_M (°C) = 87.16 + 0.345 × (%GC) + log [Na+] × (20.17 - 0.066 × (%GC)), [Na+] given in mM, (%GC) is GC content given in a range from 0 to 1.

Therefore cations are of special interest, among them physiologically important species like bivalent magnesium and monovalent sodium and potassium. They act as counterions to the phosphate backbone 52 of nucleic acids and – due to their charge screening effect – reduce repulsive force between the two stands, hence the increase in DNA melting temperature. Cations are able to counteract effects of compatible solutes on nucleic acids and vice versa 2553. This phenomenon is also linked to the counterion atmosphere 5455 and therefore to the counterion condensation theory, originally introduced by Manning 56 and Oosawa 57 and later refined by Shaughnessy 58.

In accordance with their native function in nucleic acid stability and functionality of nucleic acid processing enzymes, Mg²⁺ and K⁺ are able to stabilize nucleic acids in in vitro assays 36. But we also already know that, depending on concentrations, bivalent ion species can compact DNA and induce its aggregation 59. This compaction only occurs in a narrow range of concentrations. Apparently hydration of the cations, especially the size of the hydration shell, plays an important role: Li⁺ was reported to behave very differently from other monovalent alkali ions 60. Ions are also known to influence the stability of more exotic DNA structures 6162 and to be involved in halophilic adaptation of protein DNA interactions 636465. This is of importance when comparing and contrasting organisms which accumulate salt for osmoadaptation (salt-in strategy) with those using compatible solutes (CS strategy). K⁺ is not only used by salt-in organisms (which might have equally high Cl^- concentrations inside 66) but also by compatible solute producers in their early up-shift adaptation phase 6768. Organisms using the CS strategy usually employ potassium as a transient solute and glutamate as its counterion. All this shows us that an in vivo model of nucleic acid/compatible solute interactions definitely has to include their ionic environment.

A first approach to compatible solute nucleic acid interactions should aim at the effect of compatible solutes on the melting of double stranded DNA (dsDNA).

Many formulas for calculation of dsDNA melting temperature T_M exist. The Wallace rule 69 is the most simple one, but it is limited to DNA oligomers with 14 to 20 nucleotides and a highly specialized formula which requires the use of certain buffer concentrations. As already mentioned in the previous section, other methods for T_M calculation include terms for counterion concentrations or ionic strength of the buffer, e.g. 51. Even more sophisticated models include thermodynamic lattice models and nearest neighbour calculations 707172.

Since 1993 we know that compatible solutes have an effect on thermal melting. Rees and co-workers 53 showed that glycine betaine (or simply betaine) in high concentrations eliminates the GC-dependency of dsDNA melting, a phenomenon counteracting and being counteracted by the influence of inorganic ions. Similar effects were reported from several other groups, for example, DNA helix destabilization by proline and its possible role in osmoadaptation 73, effects of phenoxazine derivatives 43 or lowering of dsDNA T_M by trehalose 30. Data of compatible solute influence on T_M are presented in figure 2 and compared to those by sodium chloride and SYBR^®-green, an intercalator. More complex structures like DNA triplices are stabilized by water structure forming solutes 74 but also by Hofmeister salts 62. Even quadruplex/duplex equilibria are influenced by low molecular weight osmolytes 7576. While molecular crowding 77 is one important factor for quadruplex stability, the quadruplex/duplex transition is also induced by ions 61 and also effected by putrescin and polyethyleneglycol 78. Finally, high molecular weight dextrans stabilize nonviral vectors during lyophilization at low osmolalities 79. The above observations were obtained by the most simple scenario possible: a two component system with only solutes and nucleic acids. But to be able to understand how compatible solutes might act on nucleic acid stability, we need to clarify the fundamental principles first: The basics of nucleic acid stability and how this is influenced by physical parameters and other substances.

When thinking about nucleic acid stability the first thing which comes to mind is the DNA melting curve or, more precisely, thermal stability and thermal melting of dsDNA, which seems like a simple enough experiment. However, the simplicity behind this is deceiving. One is most likely tempted to forget about the influence of counterions, general ionic strength, solvent dielectricity and pH on dsDNA stability. In addition, other DNA structures and RNA are neglected completely. But even with a focus on thermal dsDNA melting the topic is quite complex 808182 and, besides the melting point, we are able to gain thermodynamic data from a melting curve. Since even small PCR products can have complex melting profiles 83, the exact determination of melting temperatures might prove to be an arduous task with researchers being eager to find new efficient methods or to improve existing ones 71.

Thus, looking at the principles of nucleic acids stability does not really help us to find answers but points out even more factors which we have to keep in mind, as for example: the mechanics of base stacking and pairing 84, the main factors in DNA stability 85, intermediate states in DNA melting including bubble nucleation, cooperativity 86 and their fluctuation 87, local cooperativity in DNA melting 88 or sensitivity of T_M to the presence of counterions 52. All of these might or might not be affected by compatible solutes.

What other factors, besides compatible solutes, are known to influence nucleic acid stability? In 1999 Spink and co-workers demonstrated that the stability of DNA duplex and triplex structures not only depends on molecular forces such as base pairing or tripling or electrostatic interactions but also on its aqueous environment 89. Thermal melting of RNA for example was shown to depend on pH and solvent 90. Recent investigations point out the influence of mixed solvents and their different dielectric constants on DNA denaturation 91. Such solvent-dependent effects indicate indirect effects on nucleic acid stability, possibly an influence of the solute or co-solvent on the structure of the solvent (water) or its electrochemical properties.

On the other hand, we are presented with reports on the proportionality of effects of intercalating substances on DNA duplex melting 42, which are quite similar to the linear dependency of solute concentrations and DNA melting points reported by Rees and co-workers 53. Surfactants are known to interact with DNA intercalators and, depending on their surface charge, to intercalate or to affect T_M to varying degrees 92. SYBR^®-green, commonly used in realtime PCR applications, even binds preferentially to certain DNA structures 93.

Again we have a huge amount of data and cannot really decide whether solutes have an indirect effect, interact directly with nucleic acids or both. To make the uncertainty complete, different compatible solutes might have different modes of action. But we are not completely left in the dark. Spink and co-workers have recently published calorimetric data and consider the importance of enthalpies of DNA melting in the presence of osmolytes 94.

In a polymerase chain reaction we have to take care of the proteins and their interactions with DNA in addition to thermal denaturation of the DNA template and formation of new DNA duplexes. To complicate matters the enzymatic reaction itself might be influenced by the addition of solutes or co-solvents, but we cannot ignore this aspect, especially since a lot of data concerning the influence of solutes are presently available. Additives which improve PCR reactions are highly sought after since even a simple "standard" application to amplify a certain region of DNA might not work properly with a particular sequence. Possible reasons might range from the simple, like GC-content or secondary structure formation of the template, to the obscure, like the complexity of compounds in a diagnostic PCR performed on clinical samples 95. Cations, mainly K⁺ and Na⁺ are among the main inhibitors of a PCR and Mg²⁺ – which is needed by the DNA polymerase – has a large, concentration dependent impact on PCR specificity 96.

It was the group of Weissensteiner which presented glycine betaine in 1996 as the first substance to counteract the effect of NaCl 27. They introduced the term cosolute for such additives. Ever since betaine has been used as a PCR facilitator: as a single compound 299798, in combination with DMSO 263299100101 or together with BSA 102. It has since been used on templates with varying GC content, to enhance formation of long PCR products 103104 in low temperature PCR with heat labile polymerases 105 and in diagnostic PCR 106. Sugars like trehalose or sucrose 107108109 have also been used with similar success as have been low molecular weight sulfones 110, amides 111 and sulfoxides 112. In addition, recent data demonstrate that even synthetic derivatives of ectoines can act as powerful PCR enhancers 31.

Most of those reports lack a reasonable explanation as to how these compounds work. Some, for example sucrose and trehalose 108 or sarcosine 113, are supposed to stabilize the DNA polymerase. Betaine and again trehalose are believed to facilitate PCR by lowering T_M or eliminating its dependency on base composition 30, which would be consistent with the early observation that DNA regions with high T_M prevent amplification 114. Furthermore, certain DNA sequences can cause DNA polymerase to pause, a phenomenon which is again counteracted by betaine 115. More exotic substances like 7-deaza-2'-deoxyguanosine compete with dNTPs for the active site and slow down PCR 116. This in turn gives a proofreading DNA polymerase more time to detect mismatched bases. But this is not to be confused with a true compatible solute, it is rather a PCR enhancer working on a different level.

Similar to the effects reported for nucleic acid stability and DNA melting we do not get a clear answer with respect to a possible mode of action. In spite of the large amount of reports on compatible solute effects in PCR we only get a vague picture as to what aspects of a PCR might be affected. And again we learn that different solutes react differently.

Already in 1998 Record 117 included DNA protein interactions into his considerations on biophysical aspects of bacterial adaptation to osmolarity. With insufficient data to get a more detailed model, low water activity as the cause for molecular crowding and reduced biopolymer diffusion and interactions was proposed as the reason behind reduced growth rates. However, observing a whole cell might not be a good idea. Nucleic acid/protein structures without catalytic function would be more ideal targets for observations of compatible solutes effects. Unfortunately, this was only done in detail for the reconstitution of functional 50S ribosomes 34. Though the model itself is not really simple and does not yield mechanistic information some impressive data were obtained: The authors demonstrated that, when using in vitro transcripts of Escherichia coli 23S rRNA, ribosome reconstitution was stimulated by a factor of up to 100 in the presence of trimethyleaminoxide in combination with an antibiotic. More recent but less detailed data are available for a complex formed by leucine-responsive regulatory protein (LRP) with ribosomal DNA (rDNA) 118. Here the CS ectoine stabilized the complex while the amino acid leucine had a destabilizing effect.

Returning to nucleic acid-protein interactions with catalytic activity I would like to draw your attention to restriction endonuclease complex formation and activity. Recent data show, that the dissociation of the EcoRI DNA complex is slowed down drastically by the neutral osmolytes glycine, glycerol, triethylene glycol and sucrose 119, an effect highly dependent on the resulting osmolarity and subsequent low water activity 120121. In another study glycine betaine was reported to improve restriction of DNA resistant to digestion despite the presence of appropriate recognition sites 39. This was compared to the positive effect of betaine on PCR and contributed to the same – unknown – mechanisms. The opposite effect, protection to the point of complete inhibition of restriction, was observed for tetrahydropyrimidine derivatives and a range of type II restriction endonucleases 38. The latter observation can be traced back to a patent application by the group of Lapidot 122123. In their work we find some NMR data which imply that proximity or binding of compatible solutes to guanine might play a role in this process. These findings are consistent with the general observation that CS have more pronounced effects on GC-rich sequences, which was made as early as 1993 53. Unfortunately the Lapidot group had published only one more project concerned with the influence of osmolytes on nucleic acid/protein interaction 124 and, apparently, did not continue NMR studies on the topic.

The findings of Lee and Gralla have a slightly more complex background: potassium glutamate takes part in the regulation of promoters for genes involved in osmoadaptation 125 and can even act as a global inhibitor of housekeeping genes under osmotic stress 126. These contribute to changes in the DNA double helix structure, probably alteration of DNA bends, and depend on glutamate concentration. Related to this observation are data from the regulation of the betaine uptake system Bus in Lactococcus lactis 127. The in vitro stability of a complex of regulator BusR and promotor busA strongly depended on the ionic strength of the buffer. Unfortunately no attempt was made to explain the molecular basis.

Naturally we would like to derive a general underlying concept for the influence of solutes on promoters. In this context it might be of interest to learn that the promoter region of a vast array of Human genes was reported to have a conserved set of distinct flexible and rigid regions independent of the consensus sequence 128. Therefore, it would be challenging to test the influence of compatible solutes on mechanical properties of promoter regions, an aspect which so far has been neglected. A possible area worthy of investigation deals with solvent property changes, such as the effect of solvent dielectric constant on DNA torsional properties 129. Again dielectric constants have not been reported for CS solutions.

So far there is only one serious approach towards describing the molecular mechanisms of compatible solute action on nucleic acid properties, a fact which can be largely ascribed to the lack of basic mechanistic data. As already mentioned above, research on protein-solute interactions is more advanced and has led to a number of models. These might prove more or less useful for nucleic acids, depending on the general mode of interaction and the solute in question.

Prior to analysing those models against the background of nucleic acid research, we have to ask ourselves whether and up to which level we can treat nucleic acids and proteins alike. With a view to nucleic acid aptamers 130131, the complexity of structures and surfaces of nucleic acids, in many ways similar to the antigen recognition site of the antibody F_ab chain, seems to be comparable to that of proteins. Therefore, one might be tempted to assume similar macromolecular properties. But what about a comparison in detail?

In both cases we have a backbone/sidechain structure. However a phosphate/sugar backbone will behave differently in comparison to a polypeptide backbone, and only four hydrophobic bases offer less sequence flexibility than the 20 proteinogenic sidechains (derivatives of both species not counted). Similarly, it would be questionable to compare double-stranded helical DNA or RNA to alpha-helical or beta-strand like structures and other conformations to omega-loops, random coils and the like. Returning to the phosphates of the nucleic acids backbone, the reason for the most striking difference becomes apparent: nucleic acids always have a strong negative net charge, even when compared to acidic proteins, and thus are surrounded by an atmosphere of positively charged countercations.

In view of all those differences, what do nucleic acids and proteins have in common? The answer lies in the basic physical principles behind interactions, which include the surrounding water as solvent and which are for both biopolymers the driving force to fold. Therefore in general all models concerning those physical principles should be applicable to proteins and nucleic acids alike.

So far there are neither publications proposing a general concept for thermodynamic calculations of compatible solute effects on nucleic acids nor molecular dynamics simulations thereof. A recent publication by Rösgen 162 discusses the basics of solute protein and solute protein metabolite thermodynamics and might prove to be a good starting point for similar considerations towards solute nucleic acid interactions. Closest to calculations concerning nucleic acids and solutes come a thermodynamic model including the effects of salt concentration on nucleic acid duplex-simplex transitions 163 and molecular dynamics studies of ion distributions for DNA duplexes and clusters 164. But as discussed above, the influence of solutes also arises from indirect effects, mainly as a consequence of a modulation of the properties of the solvent water. The thermodynamics of force-induced melting of DNA double helices, for example, including indirect effects of solvents and solutes 165166 are well known. In addition the thermodynamics of tRNA microhairpin stability and its solvent dependence have been investigated 90167. Some earlier publications already discussed solvent effects on nucleic acid base associations and simulations thereof 168169 as well as general DNA dynamics in aqueous solutions 81. Modes of direct interaction of compatible solutes which display structural similarities to DNA, like the tetrahydropyrimidines, might be derived from considerations on DNA aggregates and their melting behaviour 170. More general models for DNA melting, which consider base pairing and stacking 84, bubble nucleation and cooperativity 8688 or bubble relaxation 87 and inhomogeneity of DNA melting 171 are permanently under improvement. And a closer investigation into the secondary structure of nucleic acids 172 and their involvement in nucleic acid/protein interactions 173 might help us to understand how compatible solutes affect nucleic acid/protein complexes and their formation.

Compatible solute effects on nucleic acid properties and nucleic acid/protein complexes have been known for some time now. Besides their exploitation for in vitro applications they are also of great significance in vivo and have a large impact on vital functions. In contrast to the large amount of application data, we know only little about the molecular background of the observed effects. What is missing is a broad systematic approach to basic data, especially such simple things like thermal melting and stability. Nevertheless, we do know more than one might suspect from a quick glance at present day research.

One major problem is the comparability of data, which even starts at simple thermal melting experiments. In a system where buffer composition strongly affects the properties of the subject under investigation we have to expect three- (or more-) way interactions between target (nucleic acid), buffer and effector (compatible solute). Therefore, a broader coherent set of basic data is definitely needed before we can start to interpret the results of more sophisticated experiments in a proper way, which is especially true for anionic organic solutes.

Building a model to describe interactions between compatible solutes and nucleic acids is a difficult task, because the topic is far more complex than interactions between solutes and proteins. As summed up in figure 4, we possibly have to deal with a multitude of effects:

a) direct interactions with nucleic acid bases, probably due to cation-π-interactions similar to the situation in the solute binding proteins

b) direct interactions with the negatively charged phosphate backbone

c) interactions with the positively charged counterions and

d) indirect interactions via changes in solvent properties.

When investigating interactions with nucleic acid/protein-structures things get even more complex, since we have to consider not only the effects on nucleic acids, but also interactions of solutes with the protein(s) and the influence of osmolytes on the interactions between nucleic acids and proteins. In vivo of course, all other metabolites add to the whole picture.

We can use the concept that the effects of osmolytes are based on the exclusion or accumulation at surface elements for calculation of thermodynamics as demonstrated by Lambert 25, and all four of the models presented here have their – however limited – use for compatible solute/nucleic acid interactions. But – since those models are developed for protein/solute interactions – direct interactions are to a large extent excluded. Another point left out is the negative charge of the phosphate backbone and its counterion atmosphere. And lastly, in analogy to the osmophobic effect of proteins we might want to mint a new term here: The potential osmophilic effect of nucleic acid bases.