Thermal decomposition of the amino acids glycine, cysteine, aspartic acid, asparagine, glutamic acid, glutamine, arginine and histidine

Background The pathways of thermal instability of amino acids have been unknown. New mass spectrometric data allow unequivocal quantitative identification of the decomposition products. Results Calorimetry, thermogravimetry and mass spectrometry were used to follow the thermal decomposition of the eight amino acids G, C, D, N, E, Q, R and H between 185 °C and 280 °C. Endothermic heats of decomposition between 72 and 151 kJ/mol are needed to form 12 to 70% volatile products. This process is neither melting nor sublimation. With exception of cysteine they emit mainly H2O, some NH3 and no CO2. Cysteine produces CO2 and little else. The reactions are described by polynomials, AA→a NH3+b H2O+c CO2+d H2S+e residue, with integer or half integer coefficients. The solid monomolecular residues are rich in peptide bonds. Conclusions Eight of the 20 standard amino acids decompose at well-defined, characteristic temperatures, in contrast to commonly accepted knowledge. Products of decomposition are simple. The novel quantitative results emphasize the impact of water and cyclic condensates with peptide bonds and put constraints on hypotheses of the origin, state and stability of amino acids in the range between 200 °C and 300 °C.


Background
The so-called 20 standard amino acids are fundamental building blocks of living systems [1]. They are usually obtained in solid form from aqueous solution by evaporation of the solvent [2]. Most of today's knowledge about amino acids is therefore limited to the temperature and pressure range of liquid water. A huge number of physical and chemical data were unequivocally established − at least for the 20 standard amino acids [3]. Thermal stability or instability of amino acids, however, is one of the few fields which remains speculative until today, at least to some extent. One major reason for deficiencies in that respect could be that data aquisition is usually performed without analyzing the entire system, where liquid amino acids and decomposition products, as well as their respective gas phases must be taken into account. We used a commercial thermal analysis system with a direct transfer line to a mass spectrometer for characterizing the melting or decomposition process of amino acids under inert atmosphere in the temperature range between 323 − 593 K and detection of masses between 1 − 199 Da in the vapour phase. Mass analysis was calibrated with respect to NH 3 , H 2 O and CO 2 by searching for suitable reference substances, with the goal to identify whether or not there is a common underlying principle of melting − solidification and/or sublimation − desublimation and/or irreversible decomposition for amino acids. Previous reports missed the quantitative identification of gaseous products. The broader implication of this general relationship between amino acids and their condensation products is that amino acids might have been synthesized under prebiological conditions on earth or deposited on earth from interstellar space, where they have been found [4]. Robustness of amino acids against extreme conditions is required for early occurrence, but little is known about their nonbiological thermal destruction. There is hope that one might learn something about the molecules needed in synthesis from the products found in decomposition. Our experimental approach is not biochemical, it is merely thermochemical.

DSC, TGA, QMS
Altogether 200 samples of amino acids of at least 99.99% purity from Sigma-Aldrich were tested in a Simultaneous Thermal Analysis apparatus STA 449 Jupiter (Netzsch, Selb, Germany) coupled with Mass spectrometer QMS 403C Aëolos (Netzsch). Specimens of typically 10 mg weight in Al cans were evacuated and then heated at 5 K/min in argon flow. Differential scanning calorimetry (DSC) and thermal gravimetric analysis/thermogravimetry (TGA, TG), as well as quantitative mass spectrometry (QMS) outputs were smoothed to obtain the data of "Raw data" section. The mass spectrometer scanned 290 times between 30°C and 320°C, i.e. at every single degree in 1 Da steps between 1 Da and 100 Da. Alltogether, 290 × 100 × 200 = 5.8 million data points were analyzed.

Visuals
A MPA120 EZ-Melt Automated Melting Point Apparatus (Stanford Research Systems, Sunnyvale, CA, U.S.A.) equipped with a CAMCORDER GZ-EX210 (JVC, Bad Vilbel, Germany) was used for the optical observations. The same heating rate of 5 K/min was employed, but without inert gas protection. Screen shot images were extracted from continuous videos registered from 160 to 320°C, for all amino acids significant moments are shown in "Results".

Raw data
Although we examined all 20 amino acids, we report results for those eight of them, for which the sum of the volatile gases, NH 3 , H 2 O, CO 2 and H 2 S, matched the mass loss registered by thermogravimetry (TG). Only if both, the mass and the enthalpy balance match precisely, as in our case ±5 Da (see Table 1 for details), it is possible to take these data as a proof for the correctness of the proposed reaction. This is the reason why only 8 of the 20 amino acids are reported here. Only for them we know for sure, how they decompose. For each amino acid we show the skeleton structure, the optical observations, the DSC signal in red and the TG signal in black, as well as the ion currents for important channels, quantitatively significant are only the 17 Da (NH 3 , green lines), 18 Da (H 2 O, blue lines), and 44 Da (CO 2 , grey lines) signals. The logarithmic scale overemphasizes the molecular weights. The DSC data are given in W/g, the TG data in %. The QMS data are ion currents [A] per sample. All data are summarized in Fig. 1 for glycine, Fig. 2 for cysteine, Fig. 3 for aspartic adic, Fig. 4 for asparagine, Fig. 5 for glutamic acid, Fig. 6 for glutamine, Fig. 7 for arginine, and Fig. 8  The DSC, TGA and QMS curves share one essential feature: In DSC there are peaks at a certain temperature T peak for each amino acid, at the same temperatures they are accompanied by drops in TGA and QMS peaks. The simple fact that the DSC and QMS signals coincide in bell shaped peaks with the TGA drop proves that essentially one simple decomposition process takes place, there is not a spectrum of decomposition temperatures, as there would be for proteins. Qualitatively this proves that the process observed is neither melting nor sublimation (as claimed in the literature [5]

Calibration and quantitative mass spectrometry
The DSC signals have the dimension of specific power [W/g], the QMS ones are ion currents of the order of pA. Integration over time, or, equivalently, temperature, gives the peak areas, which are specific energies [J/g] and ionic charges, of the order of pC. Reduction from specimen weights, typically 10 mg, to mol values is trivial. In absolute terms the ion currents and ionic charges are meaningless, because equipment dependent, calibration is needed. Only one reliable calibration substance was available, sodium   (Fig. 9) there appeared a 20 μC/mol signal for the reference substance The absolute ion currents of Figs. 9, 10 and 11 are equipment dependent and not significant, but the relative values are encouraging. One mol NH 3 produces 12% less and CO 2 54% more ions than one mol H 2 O. Indeed the ionization cross sections of NH 3 , H 2 O and CO 2 are reported to be in that order [6].
Figures 9, 10, 11 and 12 and Table 1 summarize the experimental data: With the exception of cysteine, thermal decomposition results in three gases, mainly H 2 O, less NH 3 and hardly any CO 2 . The weight of these three gases adds up to the weight loss registered by TGA, therefore no other gases evolve in appreciable amount − they are not seen in QMS either. The proximity of the molfractions to integer or half-integer values indicates simple decomposition chains. The process causing the peaks cannot be melting (because of the mass loss), nor sublimation (because of the QMS signals). One concludes that amino acids do not exist in liquid or gaseous form. They decompose endothermally, with heats of decomposition between −72 and −151 kJ/mol, at well defined temperatures between 185°C and 280°C.

Data analysis, amino acid by amino acid
These amino acids consist of different side chains attached to the Cα of the same backbone, NH 2 − Cα − (C * OOH), but their decomposition chains are quite different. The pyrolytic process is controlled by three balance laws: In terms of Da the masses must add up, chemically the atomic species must balance, and the enthalpy of formation must equal the enthalpies of formation of the products plus the endothermic heat of reaction. The amounts of volatile products are experimental values (TGA and QMS). For the residues only the mass is experimental, their composition is inferred. In this section we analyze possible pathways. Although the choices, restricted by compositional, mass and enthalpy considerations, are convincing, they cannot be unique beyond doubt. Alternatives to our proposals, but indistinguishable by us, are possible. Analyses of the decomposition chains are, therefore, tentative or speculative. Nevertheless, they are less speculative than those of Rodante et al. [7], who had only TGA and DSC, but no QMS at their disposal. What Acree and Chickos [5] call "sublimation enthalpies" agree more or less with our decomposition  Ionic charges in the peaks on the left, mol CO 2 /mol amino acid on the right. Only C produces 1 mol CO 2 , the level of the others is negligible enthalpies. One concludes that they must refer to decomposition, not merely sublimation without composition change.
We made use of the enthalpy values listed for standard conditions [8] without minor corrections for specific heats and entropies up to the actual reaction temperatures. Moreover, hydrogen gas escapes our attention, it is too light (2 Da) to be registered in QMS and TGA, nor does it appear in the enthalpy sum, its heat of formation being zero by definition. With exception of hydrogen, the mass balance, controlled by TGA, confirms that beyond the residue and the three gases, nothing else is formed. The real constraint is the enthalpy balance. For the enthalpy balance production of water is necessary. The expression of the formation enthalpies of the 20 amino acids The oxygens counterbalance the others with −182 kJ/mol. The obvious way of efficiently transferring enthalpy from the reactants to the products is the formation of water, Detailed analysis for each amino acid is helped by preliminary reference to a few reactions possible in principle. CO 2 production in Cys is obviously a special case. In principle one expects the N-termini to be stable, making desamination to produce NH 3 unlikely. Nitrogen in the side chains is another matter, indeed the NH 3 producing Asn and Arg have nitrogen in their side chains. The predominance of H 2 O production indicates instability of the C-terminus beyond the C * atom, where dehydration can occur by n-oligomerization, which yields (n-1)/n mol H 2 O/mol AS, from dimerization for n = 2, to 1 mol H 2 O/mol for n → ∞ in polymerization. A special case of dimerization is external cyclization in the diketopiperazine reaction, which yields 1 mol H 2 O/mol AA. These   (Fig. 13a) or its isomer 1,2-Diazetine-3,4-dione, C 2 H 2 N 2 O 2 , chemspider 11383421, 86 Da, Fig. 13b. The scarcity of hydrogen is such that not even the smallest lactam, 2-Aziridinone, C 2 H 3 NO, 57 Da, chemspider 10574050, bp 57°C (Fig. 13c) can serve as residue.
The simplest pathway seems the formation of linear Glycylglycine, chemspider 10690 (Fig. 13d), m.p. 255°C (which is T peak !) C 4 H 8 N 2 O 3 , H f (s) = − 748 kJ/mol [8] from which the central peptide bond −(C=O)-NH− is detached by cutting off the NH 2 -Cα-H 2 − group on one, and the Cα-H 2 -C * OOH group on the other side. The former makes NH 3 plus C, the latter makes 2 C plus 2 H 2 O. This process is specific for the glycine dimer, in which the C α atoms are not protected by proper sidechains, they are just −Cα-H 2 − units. This pathway to shear peptide bonds is of interest in the context of possible peptide nucleic acid (PNA) synthesis [10]  On the left −1068 kJ, on the right −1113 kJ. Pathway to the formation of C 4 H 7 N might be ejection of the carboxyl group −C * OOH and the −SH group from Cys, the remaining chain NH 2 -Cα-C * is too short for internal, but suitable for external cyclization. Two of these form the asymmetric 5-ring (3-pyrrolidinamine, chemspider 144134), Fig. 14a, from which the −NH 2 is cutt off. Indeed the 1 2 NH 3 ejected confirms such dimerization. That leaves the molecule C 4 H 7 N: 2,5-Dihydro-1Hpyrrole, chemspider 13870958, b.p. 90°C, 69 Da, H f (s)= −46.6 kJ/mol (Fig. 14b), or another pyrroline, with the double bond elsewhere in the ring. Indeed there is heavy The compound (C 4 H 3 NO 2 ) n is polysuccinimide. The two peaks prove that the reaction occurs in two steps, in the first at 230°C the condensation reaction produces polyaspartic acid, n Asp −→ H 2 O + (Asp) n , in the second at 250°C the poly-Asp degrades to polysuccinimide (PSI) by ejection of another 1 mol H 2 O/mol Asp. Such a reaction was reported by Schiff [11]. The molecule drawn in Fig. 15 is β-poly-Asp, there is an isomer, α-poly-Asp, where the next C in the ring forms a bond to its neighbour. We have no possibility to decide between the two.   Fig. 16 occur, and there is one broad peak containing both NH 3 and H 2 O. Though the formulae for PSI formed from Asp and from Asn, are the same, (C 4 H 3 NO 2 ) n , these two residues need not be identical. For kinetic reasons the oligomerization or polymerization might have proceeded to different lengths in poly-Asp and poly-Asn, therefore the degraded products PSI might have different lengths, with different stabilities and melting points. Moreover, the telomers are different, −OH for PSI from Asp and −NH 2 for PSI from Asn. Indeed PSI from Asp remains a white powder up to 289°C, while PSI from Asn starts melting at 289°C.

Glutamic acid
Glutamic acid, Glu, E, C 5 H 9 NO 4 : 147 Da, H f = −1097 kJ/mol; T peak = 200°C, H peak = −88 kJ/mol. At 200°C, 1 mol H 2 O /mol Glu is seen in QMS, the DSC area is −121 kJ/mol, mass loss in the peak is 12% (17 Da). The dehydration of Glu has been known for a long time [12]. = 433°C (Fig. 17a). This lactam is biologically important, Fig. 16 Interpretation of Asparagine data. Two pathways from asparagine (N) to polysuccinimide (PSI): either through polyasparagine (poly-N) or polyaspartic acid (poly-D). Compared with succiminide, the N-C bond in polysuccinimide economizes two hydrogen atoms but its enthalpy of formation is apparently not known.
Known is the H f (s) = −459 kJ/mol and H f (g) = −375 kJ/mol for C 4 H 5 NO 2 , succinimide, 99 Da, the five ring with O= and =O as wings (the structure is like pyroglutamic acid, but with the carboxyl group −COOH replaced by =O, shown in Fig. 17b). The additional O should add about −200 kJ/mol, which makes the −734 kJ/mol for pyroglutamic acid plausible. The TGA weight loss beyond the peak is evaporation. Pyroglutamic acid is formed by inner cyclization of E: after the −OH hanging on C δ is ejected, the C δ joins the −NH 2 hanging on C α . This was suggested by Mosqueira et al. [13], ours is the first experimental evidence for this process. Since QMS does not show any CO 2 , the reaction to C 4 H 7 NO, the lactam pyrrolidone (Fig. 17c), 85 Da, T m = 25°C, b.p. = 245°C, H f (l) = −286 kJ/mol, yellow liquid, can be ruled out, although cutting off the CO 2 is sterically tempting.

Glutamine
Glutamine, Gln, Q, C 5 H 10 N 2 O 3 : 146 Da, H f = −826 kJ/mol. The precise 1 2 mol fractions of H 2 O and NH 3 in the peak at T peak = 185°C, H peak = −77 kJ/mol, indicate that a dimer serves as intermediate step, γ -glutamylglutamine (Fig. 18a)   A small peak without mass loss at 220°C, −14 kJ/mol, and a main peak at 230°C, −52 kJ/mol, producing 1 mol NH 3 plus 1 mol H 2 O in QMS, confirmed by the weight loss of 20% of 174 Da in TGA. The precursor peak without mass loss at 220°C, −14 kJ/mol, probably comes from a rearrangement in the guanidine star. In the large peak a double internal cyclization occurs. The loss of the amino group −NH 2 in the backbone, and internal cycling joining the N next to the C δ in the side chain to C α , forms an intermediate, 1-Carbamimidoylproline, 157 Da, chemspider 478133 (Fig. 19a). It is called "..proline", because the ring is spanned between an N and C α , though the N is not from the backbone. By losing the −OH and a second inner cyclization joining the =NH or the −NH 2 to C * , one or the other tautomer of the final residue is formed. The total reaction is C 6 H 14 N 4 O 2 −→ NH 3 + H 2 O + C 6 H 9 N 3 O, drawn in Fig. 19b, not quoted in the database [9,14].
This final residue is remarkable. It contains the proline ring, the guanidine star and a peptide bond in the ring of creatinine, which is the 5-ring with the =O and =OH double bonds. Creatinine, H f = −240 kJ/mol, m.p. 300°C, C 4 H 7 N 3 O, chemspider 568, has several tautomeric forms. The end product in question might contain either of those rings. We have no way to decide between the alternatives, but a double ring structure seems likely.

Histidine
Histidine, His, H, C 6  The observed 1mol H 2 O /mol His, confirmed by the weight loss of 13% of 155 Da, could stem from the condensation reaction of polymerization, but the volatility seen optically contradicts this option. Inner cyclization seems likely. If the C * of the backbone joins the C of the imidazole ring, with =O and −NH 2 attached outside, the 5-ring formed joins the 5-ring of the imidazole. The proposed structure is shown in Fig. 20.

Entropy of decomposition
In the tables of Domalski [15], Chickos and Acree [16] and Acree and Chickos [17], at temperatures coinciding with our peak temperatures, "Heats of sublimation" of the order of our endothermic peak areas are reported. Our QMS signals prove that chemical decomposition is involved, but that should have been obvious from the DSC data alone. The average of the entropies of transformation, S peak = H peak /T peak , is 215 J/Kmol, way above the usual entropies of melting (22 J/kmol for H 2 O, 28 for NaCl, 36 for C 6 H 6 ), and higher than typical entropies of evaporation (41 J/Kmol for H 2 O, 29 for CS 2 , 23 for CO 2 , 21 for NH 3 ). The endothermic heats in the peaks are therefore neither enthalpies of fusion nor enthalpies of sublimation, they are heats of reaction accompanied by phase changes. There is transformation and decomposition, but no reversible melting. Amino acids are stable in solid form, but not as liquids or gases.

Peptide bond formation
Five of the eight amino acids have residues containing peptide bonds, −C(=O)−NH−, only Asp and Asn leave polysuccinimide (PSI), Cys leaves cyclic pyrrolines. The preponderance of water in thermal decomposition is not surprising. In natural protein formation, each participating amino acid suffers damage. In the condensation reaction, where the N-terminus of one molecule reacts with the C-terminus of its neighbour, the planar peptide bond −Cα−CO−N−Cα− is formed. The N-atoms on, and the keto-bound O−atoms off the backbones retain their position. H 2 O is ejected, but neither NH 3 nor CO 2 are produced in protein formation. Thermal decomposition of amino acids is analogous. In protein formation, the endothermic heat is provided by ATP, in amino acid decomposition it is thermal energy. Figure 21 summarizes the results.

Peak areas
Quantitatively, the parallel between protein formation and pyrolysis is confirmed on the enthalpy level. In the formation of a dipeptide, X + Y −→ H 2 O + (X − Y), the difference between the enthalpies of the reactants and the products go into the formation of the peptide bond (PB): With the tabulated value [8,15] for H f (X), H f (Y) and H f (X-Y) one calculates H PB = −67 kJ in glycylglycine, −70 kJ in alanylglycine, −43 kJ in serylserin, −78 kJ in glycylvaline, −65 kJ in leucylglycine, −91/2 kJ in triglycylglycine, −86/2 kJ in leucylglycylglycine, and −58 kJ in glycylphenylalanine. The average value is −59 ± 13 kJ per peptide bond. The narrow standard deviation indicates that the enthalpy of forming a peptide bond is insensitive to its environment, therefore the endothermic values of oligomerization or polymerization should be close to this. One concludes that the formation of a peptide bond in a linear dimer is endothermic with an enthalpy of 59 ± 13 kJ. It is tempting to compare this with the areas of the DSC peak, the observed endothermic heat of the decomposition reaction. The average of the eight amino acids is −105 ± 27 kJ/mol. One concludes that essentially the endothermic heat of decomposition, the peak area, goes into peptide bond formation.

Production of NH 3
In cases where the N-terminus, untouched by the condensation, remains attached to a cyclic product, it could be cut off as NH 3 , contributing up to 1 2 mol NH 3 /mol AA. Remarkable is the absence of methane (CH 4 , 16 Da), hydrogen cyanide (HCN, 27 Da) and formamide (CH 3 NO, 45 Da), all in mass channels where we would have seen them. These, suspected in prebiotic synthesis of amino acids, do not appear in their decomposition. Although at most only three molecules are involved, two gases and one monomolecular residue, identification of the structure of the latter is not unequivocal, there remain more or less probable other possibilities than our choices. Clearly, without QMS, data from DSC and TGA could not possibly suffice to identify decomposition chains.

Water, cyclic compounds and peptide bonds
The novel quantitative results emphasize the importance of water and cyclic condensates containing peptide bonds. All postulated residues are cyclic compounds, five of the 8 contain peptide bonds. The residues are stable at temperatures >180°C and beyond the respective peak termperatures. These facts put constraints on hypothetical origin, state and stability of amino acids in the range between 200°C and 300°C in the absence and presence of water, Fig. 21 Overview of the residues with respect to H 2 O or NH 3 contents. All residues are obtained by ejection of 0, 1 2 , 1, 1 1 2 or 2 mols of H 2 O or NH 3 , placing them on two axes. All residues contain either 0, 1 2 or 1 mol NH 3 or H 2 O, placing them on two axes. Most of them contain peptide bonds. The polysuccinimide of D and N is an exception, cysteine, for lack of oxygen, the other but literature is sparse in that respect. The history of diketopiperazine and derivatives is extensively reviewed by Prasad [18] back until 1888, the synthesis of cyclo-Gly-Gly by Curtius and Gloebel [19], but Prasad emphasized the relevance of these compounds as a class of natural products no earlier than 1922 [20][21][22]. Today, CDPs are recognized as "transkingdom signaling molecules" [23], indicating highly conserved mechanisms from earliest stages of life on earth. The potential of CDPs and related substances as novel drugs for biomedical applicatons is comprehensively reviewed by Borthwick [24], though he does not cover aqueous regimes above 150°C. Thermal formation of cyclo-Leu-Leu from the Leu-Leu dipeptide in the solid state was reported to occur at 177°C [25] − actually several month after the release of QMS data for amino acids Gly, Cys, Asp, Asn, Glu, Gln, Arg, and His in the solid state [26]. There might be differences in terms of thermal cyclization by dehydration, depending on whether or not the origin is an amino acid crystal or a dipeptide. Controlled biosynthesis of CDP by highly conserved enzymes is found in all domains of life [27]. However, the biochemistry of cyclo-dipeptides and related enzymatic pathways is a comparatively unexplored interdisciplinary field, usually based on genome analyses. Just recently, their presence in extremophilic organisms has been highlighted in more detail [28,29]. Previously reported evidences along with the first conclusive demonstration of thermal cyclization of Gly, Cys, Asp, Asn, Glu, Gln, Arg, and His by QMS, DSC and TGA as reported here put emphasis on the fact that cyclic dipeptides or cyclic compounds could represent thermally more stable precursors of prebiotic life.

Conclusions
Our comparative analysis allowed us to identify the eight of the twenty standard amino acids, for which the thermochemical equations unequivocally agree with stoichiometric release of NH 3 , H 2 O and CO 2 . The predominance of the release of H 2 O during the process of decomposition instead of melting indicates a common principle of condensation and, depending on the individual properties of the respective intermediate products, subsequent decomposition of the condensation products. Comparative data for all 20 standard amino acids, obtained by complementing DSC and TGA with quantitative mass spectrometry (QMS), have never been reported. For the eight with data closure we can say: Amino acids decompose thermally, they do not sublimate, nor do they melt. Only three gases are formed, mostly H 2 O, less so NH 3 and hardly any CO 2 . Cys forms H 2 S, but not CS 2 . In all amino acids investigated, Gly, Cys, Asn, Asp, Gln, Glu, Arg, His, the liquid or solid residues are lactams and heterocyclic compounds with 5-or 6-membered non-(or only partially) aromatic rings, containing one or two nitrogen atoms (pyrrolidines, piperidines, pyrrazolidines, piperazines ), most of them with peptide bonds present.
In summary, this work addresses an important question of amino acid thermal stability. Several processes may occur upon heating, chemical decomposition or sublimation/evaporation without decomposition. The aim of this work was to accurately determine these processes. For 8 out of the 20 standard amino acid, we demonstrated that these have a well defined temperature of decomposition. The contemporary detection of products of 17 Da, 18 Da, and 44 Da in the gas phase is the proof for decomposition, concise mass and enthalpy balances do not leave any room for speculation. Analysis and interpretation rule out the existence of any byproduct hydrocarbons which could not have been validated. The analysis for glycine, cysteine, aspartic acid, asparagine, glutamic acid, glutamine, arginine and histidine is beyond any doubt. At a heating rate of 5 K/min, neither melting nor sublimation take place. At least 8 of 20 standard amino acids do not exist in liquid form.