Transcriptional frameshifts contribute to protein allergenicity.

Transcription infidelity (TI) is a mechanism that increases RNA and protein diversity. We found that single-base omissions (i.e., gaps) occurred at significantly higher rates in the RNA of highly allergenic legumes. Transcripts from peanut, soybean, sesame, and mite allergens contained a higher density of gaps than those of nonallergens. Allergen transcripts translate into proteins with a cationic carboxy terminus depleted in hydrophobic residues. In mice, recombinant TI variants of the peanut allergen Ara h 2, but not the canonical allergen itself, induced, without adjuvant, the production of anaphylactogenic specific IgE (sIgE), binding to linear epitopes on both canonical and TI segments of the TI variants. The removal of cationic proteins from bovine lactoserum markedly reduced its capacity to induce sIgE. In peanut-allergic children, the sIgE reactivity was directed toward both canonical and TI segments of Ara h 2 variants. We discovered 2 peanut allergens, which we believe to be previously unreported, because of their RNA-DNA divergence gap patterns and TI peptide amino acid composition. Finally, we showed that the sIgE of children with IgE-negative milk allergy targeted cationic proteins in lactoserum. We propose that it is not the canonical allergens, but their TI variants, that initiate sIgE isotype switching, while both canonical and TI variants elicit clinical allergic reactions.

Detection of IgE-reactive proteins in hydrolysed dog foods.

BACKGROUND: Commercial hydrolysed diets are used for the diagnosis of food allergy in dogs. The cleaved parent proteins are presumed to be too small to elicit an allergic response by reacting with allergen-specific immunoglobin E (IgE). OBJECTIVES: To evaluate three commercial hydrolysed dog diets for proteins. ANIMALS: Sera were collected from dogs with suspected food allergy. METHODS: Two batches of each hydrolysed diet were examined by electrophoresis and visualized by Coomassie blue, silver nitrate staining and IgE immunoblotting. RESULTS: From two to five proteins, ranging from 21 to 67 kDa, were detected in all three diets evaluated. Circulating IgE antibodies targeting these proteins were detected by immunoblotting of dog sera. Six different carbohydrate proteins were identified by mass spectrometry; maize/potato granule-bound starch synthase-1, soybean glycinin, soybean ß-conglycinin α chain, potato aspartic protease inhibitor, rice glutelin type B1 and soybean sucrose-binding protein. Four of these proteins have been described as allergens in humans. CONCLUSIONS: Some commercial hydrolysed diets contain carbohydrate proteins. Some dogs have circulating IgE antibodies targeting these proteins. The clinical significance of these findings is unknown.

The crystal structure of the FAD/NADPH-binding domain of flavocytochrome P450 BM3.

We report the crystal structure of the FAD/NADPH-binding domain (FAD domain) of the biotechnologically important Bacillus megaterium flavocytochrome P450 BM3, the last domain of the enzyme to be structurally resolved. The structure was solved in both the absence and presence of the ligand NADP(+), identifying important protein interactions with the NADPH 2'-phosphate that helps to dictate specificity for NADPH over NADH, and involving residues Tyr974, Arg966, Lys972 and Ser965. The Trp1046 side chain shields the FAD isoalloxazine ring from NADPH, and motion of this residue is required to enable NADPH-dependent FAD reduction. Multiple binding interactions stabilize the FAD cofactor, including aromatic stacking with the adenine group from the side chains of Tyr860 and Trp854, and several interactions with FAD pyrophosphate oxygens, including bonding to tyrosines 828, 829 and 860. Mutagenesis of C773 and C999 to alanine was required for successful crystallization, with C773A predicted to disfavour intramolecular and intermolecular disulfide bonding. Multiangle laser light scattering analysis showed wild-type FAD domain to be near-exclusively dimeric, with dimer disruption achieved on treatment with the reducing agent dithiothreitol. By contrast, light scattering showed that the C773A/C999A FAD domain was monomeric. The C773A/C999A FAD domain structure confirms that Ala773 is surface exposed and in close proximity to Cys810, with this region of the enzyme's connecting domain (that links the FAD domain to the FMN-binding domain in P450 BM3) located at a crystal contact interface between FAD domains. The FAD domain crystal structure enables molecular modelling of its interactions with its cognate FMN (flavodoxin-like) domain within the BM3 reductase module.

Effect of surfactant pluronic F-68 on CHO cell growth, metabolism, production, and glycosylation of human recombinant IFN-γ in mild operating conditions.

The control of glycosylation to satisfy regulatory requirements and quality consistency of recombinant proteins produced by different processes has become an important issue. With two N-glycosylation sites, γ-interferon (IFN-γ) can be seen as a prototype of a recombinant therapeutic glycoprotein for this purpose. The effect of the nonionic surfactant Pluronic F-68 (PF-68) on cell growth and death was investigated, as well as production and glycosylation of recombinant IFN-γ produced by a CHO cell line that was maintained in a rich protein-free medium in the absence or presence of low agitation. Under these conditions, a dose-dependent effect of PF-68 (0-0.1%) was shown not only to significantly enhance growth but also to reduce cell lysis. Interestingly, supplementing the culture medium with PF-68 led to increased IFN-γ production as a result of both higher cell densities and a higher specific production rate of IFN-γ. If cells were grown with agitation, lack of PF-68 in the culture medium decreased the fraction of the fully glycosylated IFN-γ glycoform (2N) from 80% to 65-70% during the initial period. This effect appeared to be due to a lag phase in cell growth observed during this period. Finally, a global kinetic study of CHO cell metabolism indicated higher efficiency in the utilization of the two major carbon substrates when cultures were supplemented with PF-68. Therefore, these results highlight the importance of understanding how media surfactant can affect cell growth as well as cell death and the product quality of a recombinant glycoprotein expressed in CHO cell cultures.

Lipolysis stimulated lipoprotein receptor: a novel molecular link between hyperlipidemia, weight gain, and atherosclerosis in mice.

The lipolysis-stimulated lipoprotein receptor, LSR, is a multimeric protein complex in the liver that undergoes conformational changes upon binding of free fatty acids, thereby revealing a binding site (s) that recognizes both apoB and apoE. Complete inactivation of the LSR gene is embryonic lethal in mice. Here we show that removal of a single LSR allele (LSR(-/+)) caused statistically significant increases in both plasma triglyceride and cholesterol levels, a 2-fold increase in plasma triglyceride changes during the post-prandial phase, and delayed clearance of lipid emulsions or a high fat meal. The longer postprandial lipoprotein clearance time observed in LSR(-/+) mice was further increased in LSR(-/+) mice lacking functional low density lipoprotein (LDL) receptors. LSR(-/+) mice placed on a Western-type diet displayed higher plasma triglycerides and cholesterol levels, increased triglyceride-rich lipoproteins and LDL, and increased aorta lipid content, as compared with control mice on the same diet. Furthermore, a direct correlation was observed between the hyperlipidemia and weight gain but only in the LSR(-/+) mice. Knockdown of LSR expression by small interfering RNA in mouse Hepa1-6 cells led to decreased internalization of both DiI-labeled cyclohexanedione-LDL and very low density lipoprotein in the presence of oleate. These data led us to conclude that LSR contributes to the physiological clearance of atherogenic triglyceride-rich lipoproteins and LDL. We propose that LSR cooperates with the LDL receptor in the final hepatic processing of apoB-containing lipoproteins and represents a novel therapeutic target for the treatment of hyperlipidemia associated with obesity and atherosclerosis.

A single oral sensitization to peanut without adjuvant leads to anaphylaxis in mice.

BACKGROUND: A model of peanut food allergy has been developed in mice using a simple sensitization protocol leading to a quantitatively measurable allergic response. METHODS: C3H/HeJ mice received a single intragastric administration of whole peanut (80 mg) without adjuvant. Two weeks later, intraperitoneal challenge with peanut extract led to a severe anaphylaxis. RESULTS: Anaphylactic reaction was evidenced by vascular leakage, severe clinical symptoms, a drop in body temperature, a decrease in breathing rate and also by increased concentrations of serum mouse mast cell protease-1. Sensitization to peanut was demonstrated by positive skin tests (ear swelling test and intradermal skin testing) and increased peanut-specific IgE levels. CONCLUSIONS: Thus, we obtained a model of severe peanut hypersensitivity within 2 weeks following single oral exposure without adjuvant. This model may be useful for further basic and applied studies on peanut allergy.

Nonrandom variations in human cancer ESTs indicate that mRNA heterogeneity increases during carcinogenesis.

Virtually all cancer biological attributes are heterogeneous. Because of this, it is currently difficult to reconcile results of cancer transcriptome and proteome experiments. It is also established that cancer somatic mutations arise at rates higher than suspected, but yet are insufficient to explain all cancer cell heterogeneity. We have analyzed sequence variations of 17 abundantly expressed genes in a large set of human ESTs originating from either normal or cancer samples. We show that cancer ESTs have greater variations than normal ESTs for >70% of the tested genes. These variations cannot be explained by known and putative SNPs. Furthermore, cancer EST variations were not random, but were determined by the composition of the substituted base (b0) as well as that of the bases located upstream (up to b - 4) and downstream (up to b + 3) of the substitution event. The replacement base was also not randomly selected but corresponded in most cases (73%) to a repetition of b - 1 or of b + 1. Base substitutions follow a specific pattern of affected bases: A and T substitutions were preferentially observed in cancer ESTs. In contrast, cancer somatic mutations [Sjoblom T, et al. (2006) Science 314:268-274] and SNPs identified in the genes of the current study occurred preferentially with C and G. On the basis of these observations, we developed a working hypothesis that cancer EST heterogeneity results primarily from increased transcription infidelity.

Predictive value of skin prick tests using recombinant allergens for diagnosis of peanut allergy.

BACKGROUND: Current diagnosis of peanut allergy relies on natural extracts that lack standardization. Recombinant DNA technology allows production of pure biochemically characterized proteins. Their usefulness for peanut allergy diagnosis is not established. OBJECTIVE: This study aimed to evaluate the diagnostic value of the 3 major recombinant peanut allergens. METHODS: Recombinant (r) Ara h 1, rAra h 2, and rAra h 3 were produced according to the recommendations of good manufacturing practice for recombinant allergens. Skin prick tests (SPTs) and IgE ELISA assays were performed in 30 patients with peanut allergy and 30 control subjects without food allergy: 15 nonatopic and 15 sensitized to birch pollen. Disease severity was graded by clinical scoring. RESULTS: All patients with peanut allergy showed positive SPT results to rAra h 2; 40% reacted with rAra h 1 and 27% with rAra h 3. No control subjects reacted with any of the recombinant allergens. Monosensitization to rAra h 2 was observed in 53% of patients. Neither SPT size nor levels of specific IgE were correlated with the disease severity. However, patients with monosensitization to rAra h 2 had a significantly lower severity score than polysensitized subjects and a lower level of specific IgE against peanut extract and rAra h 2. CONCLUSION: Skin prick tests to individual recombinant peanut allergens appear to be a safe and effective diagnostic tool. Cosensitization to rAra h 2 and rArah 1 and/or rAra h 3 is predictive of more severe reactions. CLINICAL IMPLICATIONS: Recombinant peanut allergens can be used by SPTs for diagnosis and evaluation of allergy severity.

Misfolded forms of glyceraldehyde-3-phosphate dehydrogenase interact with GroEL and inhibit chaperonin-assisted folding of the wild-type enzyme.

We studied the interaction of chaperonin GroEL with different misfolded forms of tetrameric phosphorylating glyceraldehyde-3-phosphate dehydrogenase (GAPDH): (1) GAPDH from rabbit muscles with all SH-groups modified by 5,5'-dithiobis(2-nitrobenzoate); (2) O-R-type dimers of mutant GAPDH from Bacillus stearothermophilus with amino acid substitutions Y283V, D282G, and Y283V/W84F, and (3) O-P-type dimers of mutant GAPDH from B. stearothermophilus with amino acid substitutions Y46G/S48G and Y46G/R52G. It was shown that chemically modified GAPDH and the O-R-type mutant dimers bound to GroEL with 1:1 stoichiometry and dissociation constants K(d) of 0.4 and 0.9 muM, respectively. A striking feature of the resulting complexes with GroEL was their stability in the presence of Mg-ATP. Chemically modified GAPDH and the O-R-type mutant dimers inhibited GroEL-assisted refolding of urea-denatured wild-type GAPDH from B. stearothermophilus but did not affect its spontaneous reactivation. In contrast to the O-R-dimers, the O-P-type mutant dimers neither bound nor affected GroEL-assisted refolding of the wild-type GAPDH. Thus, we suggest that interaction of GroEL with certain types of misfolded proteins can result in the formation of stable complexes and the impairment of chaperonin activity.

Switching pyridine nucleotide specificity in P450 BM3: mechanistic analysis of the W1046H and W1046A enzymes.

Flavocytochrome P450 BM3 is a member of the diflavin reductase enzyme family. Members include cytochrome P450 reductase, nitric-oxide synthase, methionine synthase reductase, and novel oxidoreductase 1. These enzymes show a strong preference for NADPH over NADH as reducing coenzyme. An aromatic residue stacks over the FAD isoalloxazine ring in each enzyme, and in some cases it is important in controlling coenzyme specificity. In P450 BM3, the aromatic residue inferred from sequence alignments to stack over the FAD is Trp-1046. Mutation to Ala-1046 and His-1046 effected a remarkable coenzyme specificity switch. P450 BM3 W1046A/W106H FAD and reductase domains are efficient NADH-dependent ferricyanide reductases with selectivity coefficients (k(cat)/K(m)(NADPH)/k(cat)/K(m)(NADH)) of 1.5, 67, and 8571 for the W1046A, W1046H, and wild-type reductase domains, respectively. Stopped-flow photodiode array absorption studies indicated a charge-transfer intermediate accumulated in the W1046A FAD domain (and to a lesser extent in the W1046H FAD domain) and was attributed to formation of a reduced FADH(2)-NAD(P)(+) charge-transfer species, suggesting a relatively slow rate of release of NAD(P)(+) from reduced enzymes. Unlike wild-type enzymes, there was no formation of the blue semiquinone species observed during reductive titration of the W0146A/W146H FAD and reductase domains with dithionite or NAD(P)H. This was a consequence of elevation of the semiquinone/hydroquinone couple of the FAD with respect to the oxidized/semiquinone couple, and a concomitant approximately 100-mV elevation in the 2-electron redox couple for the enzyme-bound FAD (-320, -220, and -224 mV in the wild-type, W1046A, and W1046H FAD domains, respectively).

Expression, purification, and characterization of Bacillus subtilis cytochromes P450 CYP102A2 and CYP102A3: flavocytochrome homologues of P450 BM3 from Bacillus megaterium.

The cyp102A2 and cyp102A3 genes encoding the two Bacillus subtilis homologues (CYP102A2 and CYP102A3) of flavocytochrome P450 BM3 (CYP102A1) from Bacillus megaterium have been cloned, expressed in Escherichia coli, purified, and characterized spectroscopically and enzymologically. Both enzymes contain heme, flavin adenine dinucleotide (FAD) and flavin mononucleotide (FMN) cofactors and bind a variety of fatty acid molecules, as demonstrated by conversion of the low-spin resting form of the heme iron to the high-spin form induced by substrate-binding. CYP102A2 and CYP102A3 catalyze the fatty acid-dependent oxidation of reduced nicotinamide adenine dinucleotide phosphate (NADPH) and reduction of artificial electron acceptors at high rates. Binding of carbon monoxide to the reduced forms of both enzymes results in the shift of the heme Soret band to 450 nm, confirming the P450 nature of the enzymes. Reverse-phase high-performance liquid chromatography (HPLC) of products from the reaction of the enzymes with myristic acid demonstrates that both catalyze the subterminal hydroxylation of this substrate, though with different regioselectivity and catalytic rate. Both P450s 102A2 and 102A3 show kinetic and binding preferences for long-chain unsaturated and branched-chain fatty acids over saturated fatty acids, indicating that the former two molecule types may be the true substrates. P450s 102A2 and 102A3 exhibit differing substrate selectivity profiles from each other and from P450 BM3, indicating that they may fulfill subtly different cellular roles. Titration curves for binding and turnover kinetics of several fatty acid substrates with P450s 102A2 and 102A3 are better described by sigmoidal (rather than hyperbolic) functions, suggesting binding of more than one molecule of substrate to the P450s, or possibly cooperativity in substrate binding. Comparison of the amino acid sequences of the three flavocytochromes shows that several important amino acids in P450 BM3 are not conserved in the B. subtilis homologues, pointing to differences in the binding modes for the substrates that may explain the unusual sigmoidal kinetic and titration properties.

Electron transfer in flavocytochrome P450 BM3: kinetics of flavin reduction and oxidation, the role of cysteine 999, and relationships with mammalian cytochrome P450 reductase.

Cys-999 is one component of a triad (Cys-999, Ser-830, and Asp-1044) located in the FAD domain of flavocytochrome P450 BM3 that is almost entirely conserved throughout the diflavin reductase family of enzymes. The role of Cys-999 has been studied by steady-state kinetics, stopped-flow spectroscopy, and potentiometry. The C999A mutants of BM3 reductase (containing both FAD and FMN cofactors) and the isolated FAD domain are substantially compromised in their capacity to reduce artificial electron acceptors in steady-state turnover with either NADPH or NADH as electron donors. Stopped-flow studies indicate that this is due primarily to a substantially slower rate of hydride transfer from nicotinamide coenzyme to FAD cofactor in the C999A enzymes. The compromised rates of hydride transfer are not attributable to altered thermodynamic properties of the flavins. A reduced enzyme-NADP(+) charge-transfer species is populated following hydride transfer in the wild-type FAD domain, consistent with the slow release of NADP(+) from the 2-electron-reduced enzyme. This intermediate does not accumulate in the C999A FAD domain or wild-type and C999A BM3 reductases, suggesting more rapid release of NADP(+) from these enzyme forms. Rapid internal electron transfer from FAD to FMN in wild-type BM3 reductase releases NADP(+) from the nicotinamide-binding site, thus preventing the inhibition of enzyme activity through the accumulation of a stable FADH(2)-NADP(+) charge-transfer complex. Hydride transfer is reversible, and the observed rate of oxidation of the 2-electron-reduced C999A BM3 reductase and FAD domain is hyperbolically dependent on NADP(+) concentration. With the wild-type BM3 reductase and FAD domain, the rate of flavin oxidation displays an unusual dependence on NADP(+) concentration, consistent with a two-site binding model in which two coenzyme molecules bind to catalytic and regulatory regions (or sites) within a bipartite coenzyme binding site. A kinetic model is proposed in which binding of coenzyme to the regulatory site hinders sterically the release of NADPH from the catalytic site. The results are discussed in the light of kinetic and structural studies on mammalian cytochrome P450 reductase.

Determination of the redox potentials and electron transfer properties of the FAD- and FMN-binding domains of the human oxidoreductase NR1.

Human novel reductase 1 (NR1) is an NADPH dependent diflavin oxidoreductase related to cytochrome P450 reductase (CPR). The FAD/NADPH- and FMN-binding domains of NR1 have been expressed and purified and their redox properties studied by stopped-flow and steady-state kinetic methods, and by potentiometry. The midpoint reduction potentials of the oxidized/semiquinone (-315 +/- 5 mV) and semiquinone/dihydroquinone (-365 +/- 15 mV) couples of the FAD/NADPH domain are similar to those for the FAD/NADPH domain of human CPR, but the rate of hydride transfer from NADPH to the FAD/NADPH domain of NR1 is approximately 200-fold slower. Hydride transfer is rate-limiting in steady-state reactions of the FAD/NADPH domain with artificial redox acceptors. Stopped-flow studies indicate that hydride transfer from the FAD/NADPH domain of NR1 to NADP+ is faster than hydride transfer in the physiological direction (NADPH to FAD), consistent with the measured reduction potentials of the FAD couples [midpoint potential for FAD redox couples is -340 mV, cf-320 mV for NAD(P)H]. The midpoint reduction potentials for the flavin couples in the FMN domain are -146 +/- 5 mV (oxidized/semiquinone) and -305 +/- 5 mV (semiquinone/dihydroquinone). The FMN oxidized/semiquinone couple indicates stabilization of the FMN semiquinone, consistent with (a) a need to transfer electrons from the FAD/NADPH domain to the FMN domain, and (b) the thermodynamic properties of the FMN domain in CPR and nitric oxide synthase. Despite overall structural resemblance of NR1 and CPR, our studies reveal thermodynamic similarities but major kinetic differences in the electron transfer reactions catalysed by the flavin-binding domains.

P but not R-axis interface is involved in cooperative binding of NAD on tetrameric phosphorylating glyceraldehyde-3-phosphate dehydrogenase from Bacillus stearothermophilus.

Homotetrameric phosphorylating glyceraldehyde-3-phosphate dehydrogenase (GAPDH) from Bacillus stearothermophilus can be described as a dimer of dimers with three non-equivalent P, R, and Q interfaces. In our previous study, negative cooperativity in NAD binding to wild-type GAPDH was interpreted according to the induced-fit model in terms of two independent dimers with two interacting binding sites in each dimer. Two dimeric mutant GAPDHs, i.e. Y46G/S48G and D186G/E276G, were shown to exhibit positive cooperativity in NAD binding. Based on the molecular modeling of the substitutions and the fact that the most extensive inter-subunit interactions are formed across the P-axis interface of the tetramer, it was postulated that both dimeric mutant GAPDHs were of O-P type. Therefore, the P-axis interface was assumed to play a major role in causing cooperativity in NAD binding.Here, two other mutant GAPDHs, Y46G/R52G and D282G, have been studied. Using small angle X-ray scattering, the dimeric form of the D282G mutant GAPDH is shown to be of O-R type whereas both dimeric mutant GAPDHs Y46G/R52G and Y46G/S48G are of O-P type. Similarly to dimeric Y46G/S48G mutant GAPDH, the dimeric Y46G/R52G mutant GAPDH exhibits positive cooperativity in NAD binding. On the other hand, no significant cooperativity in NAD binding to the dimeric form of the D282G mutant GAPDH is observed, whereas its tetrameric counterpart exhibits negative cooperativity, similarly to the wild-type enzyme. Altogether, the results support the view that the P-axis interface is essential in causing cooperativity in NAD binding by transmitting the structural information induced upon cofactor binding from one subunit to the other one within O-P/Q-R dimers in contrast to the R-axis interface, which does not transmit structural information within O-R/Q-P dimers. The absence of activity of O-P and O-R dimer GAPDHs is the consequence of a pertubation of the conformation of the active site, at least of the nicotinamide subsite, as evidenced by the absence of an ion pair between catalytic residues C149 and H176 and the greater accessibility of C149 to a thiol kinetic probe.

Thermal unfolding used as a probe to characterize the intra- and intersubunit stabilizing interactions in phosphorylating D-glyceraldehyde-3-phosphate dehydrogenase from Bacillus stearothermophilus.

Tetrameric phosphorylating glyceraldehyde-3-phosphate dehydrogenase (GAPDH) from Bacillus stearothermophilus can be described as a dimer of dimers with three nonequivalent interfaces. To investigate the contribution of intra- and intersubunit interactions to GAPDH thermostability, 10 residues located either at the cofactor domain (amino acids 1-148 and 313-333) or at the catalytic domain (amino acids 149-312) were mutated and the thermal unfolding of the mutants was studied by differential scanning calorimetry in the absence and presence of saturating concentrations of NAD. Disruptions of intrasubunit interactions lead to a drastic decrease in thermostability of the N313T, Y283V, and W310F mutants. Moreover, for the N313T mutant, a weakening of cooperative interactions between the catalytic and the cofactor domains and an inefficient binding of NAD are observed. This is likely the consequences of modification or loss of the hydrogen bonding network associating N313 and residues 236-238 and N313 and the nicotinamide carboxyamide of NAD, respectively. For the residues Y283 and W310, which are involved in stacking hydrophobic interactions, mutating both positions does not affect the efficiency of NAD binding. This shows that the factors involved in the thermostability of the tetrameric apo GAPDH are then different from those induced by NAD binding. Disruption of intersubunit hydrogen bonds between the catalytic domain and the NAD-binding domain of a neighboring subunit also leads to a significant destabilization of the apo tetrameric form as observed for the D282G mutant. Moreover, no efficient binding of NAD is observed. Both results are likely the consequence of a loss of hydrogen bonds across the P-axis and the Q-axis between D282 and R197 and between D282 and R52, respectively. Similar results, i.e., a destabilizing effect and inefficient NAD binding, are observed with the T34Q/T39S/L43Q mutant in which steric hindrance is introduced at the S-loop of the R-axis-related subunit via mutations at the adenosine subsite. The dimeric form of the D282G mutant exhibits a single partial heat absorption peak, whereas the Y46G/R52G mutant which exists only as a dimer shows two peaks. Taking into account the recent small-angle X-ray scattering studies which suggested that the dimeric form of the D282G mutant and of the dimeric Y46G/R52G mutant are of the O-R and O-P types, respectively (Vachette, unpublished results), we propose that the presence of one or two peaks in thermal unfolding of dimers is a signature of the dimer type.