Age-dependent deterioration of long-lived proteins in humans may have wide-ranging effects on health, fitness and diseases of the elderly. To a large extent, denaturation of old proteins appears to result from the intrinsic instability of certain amino acids; however, these reactions are incompletely understood. One method to investigate these reactions involves exposing peptides to elevated temperatures at physiological pH. Incubation of PFHSPSY, which corresponds to a region of human αB-crystallin that is susceptible to age-related modification, resulted in the appearance of a major product. NMR spectroscopy confirmed that this novel peptide formed via racemization of the N-terminal Pro. This phenomenon was not confined to Pro, because peptides with N-terminal Ser and Ala residues also underwent racemization. As N-terminal racemization occurred at 37 °C, a long-lived protein was examined. LC-MS/MS analysis revealed that approximately one third of aquaporin 0 polypeptides in the centre of aged human lenses were racemized at the N-terminal methionine.
- FT ICR
Fourier transform ion cyclotron resonance mass spectrometry
heteronuclear single quantum coherence
nuclear Overhauser effect spectroscopy
photodiode array detector
total correlation spectroscopy
Long-lived proteins are widespread in the human body [1, 2]. The most abundant are the collagens, which account for ~ 20% of the body mass of an adult  and have a half-life in the human body that is estimated at 95 years . Over time in a biochemical environment, collagens and other persistent proteins such as elastin [5, 6], dentin [7, 8], myelin basic protein [9, 10] and lens crystallins [11, 12] undergo numerous changes. Some of these have been characterized, for example conversion of l-Asp and l-Asn residues into isomeric d- and isoAsp forms via inter-molecular condensation . Spontaneous peptide bond cleavage at Asn residues of proteins may also occur via a succinimide intermediate . Such age-related modifications have been implicated in provoking human autoimmune disease [15, 16].
The human lens contains the highest protein concentration of any tissue in the body, but there is no protein turnover . This means that proteins in the centre of the lens are present for a lifetime. This tissue may therefore be used to examine post-translational events that occur in other long-lived proteins. Prominent modifications such as methylation [18, 19], racemization [20-22], deamidation [13, 20, 23] and truncation [24-26] have been characterized; however, it is likely that others remain to be elucidated.
Some reactions of long-lived proteins may be modelled simply by exposing proteins and peptides to elevated temperatures [27, 28]. In the present investigation, a peptide was exposed to heat at pH 7.4 and the major products were characterized. A facile racemization of the N-terminal amino acid residue was observed, and a mechanism is proposed for its formation. This novel process may apply to all proteins with unblocked N-termini, as analysis of aquaporin 0 from older human lenses showed that ~ 30% of the N-terminal methionine was present as the d- form. Conversion of the terminal amino acid to the d- form may assist in stabilizing long-lived proteins against degradation by aminopeptidases.
Some reactions of long-lived proteins in the body may be conveniently investigated by exposing proteins or peptides to elevated temperatures under physiological pH conditions. During incubation of a peptide, PFHSPSY, a major product (peptide X) was observed that eluted after the PFHSPSY peak (Fig. 1). The amount of this new component increased with duration of incubation at 60 °C, as the amount of PFHSPSY decreased (Fig. 2). In order to determine the structure of peptide X, semi-preparative HPLC was performed on a sample of PFHSPSY that had been incubated for 1 week. The collected peptide X was then examined by NMR spectroscopy.
It was apparent that peptide X was closely related in structure to the starting material. As peptide bonds preceding Pro residues (i.e. Xaa-Pro, where Xaa is any amino acid) may adopt one of two conformations (cis and trans) that may be simultaneously present in solution because of the low energy barriers of rotation about the peptidyl-Pro imide bond [29, 30], 2D NMR spectroscopy was used to investigate whether cis Xaa-Pro bond formation had occurred in peptide X.
Spectroscopic identification of a cis Xaa-Pro peptide bond by NMR traditionally relies upon observation of a strong 1Hα–1Hα NOE correlation between the two sequential residues . In contrast, a trans Xaa-Pro peptide bond is expected to show a 1Hα–1Hδ NOE correlation between the two sequential residues. If peptide X is a cis Xaa-Pro isomer of PFHSPSY, there should be a 1Hα–1Hα NOE correlation between Ser4 and Pro5. Analysis by 2D ROESY did not reveal any such correlation, indicating it is unlikely that a cis Pro bond is present.
As NOE correlations are prone to chemical shift degeneration and artefacts originating from insufficient water suppression, a 13C-HSQC spectrum was also obtained. 13C chemical shifts are very sensitive to local chemical environments, with 13Cβ and 13Cγ resonances in Pro residues typically shifting up-field by 2–3 ppm as the Xaa-Pro bond changes from trans to cis . There was a negligible change in the 13Cβ and 13Cγ chemical shifts of the two samples; therefore, 13C chemical shift analysis also suggests that peptide X is unlikely to be a cis Pro variant of PFHSPSY (data not shown).
All NMR signals observed in the peptide X and PFHSPSY spectra were unambiguously assigned to atoms belonging to each amino acid residue in the peptide sequence. Close examination revealed that the NMR signals corresponding to one of the Pro residues were shifted in peptide X compared with the starting peptide (Fig. 3). The chemical shift values corresponding to Pro1 in PFHSPSY and peptide X differed significantly, whereas the signals corresponding to Pro5 in both peptides remained essentially the same.
This suggests that the N-terminal Pro (Pro1) is in a different chemical environment in peptide X and is likely to have undergone modification. However, the fact that the change in chemical shifts was relatively small in magnitude, and that neighbouring residues were little affected, suggested that the modification was localized.
As cis/trans isomerization of Pro did not appear to explain the behaviour of the peptide, we suspected that the N-terminal Pro may be racemized in peptide X. To confirm this, acid digests of both PFHSPSY and peptide X were examined by chiral HPLC to test for the presence of d-Pro. Peptide X was found to contain ~ 50% d-Pro (Fig. S1). A (d-Pro)FHSPSY standard was synthesized, and its 1H-NMR spectrum matched that of peptide X (Fig. 4). In addition, the synthetic d-Pro peptide co-eluted with peptide X by HPLC, as indicated in Fig. 1.
Having established that N-terminal racemization occurs in this peptide, other peptides were examined. Two homologous heptapeptides with Ser and Ala as the N-terminal residues were incubated and also found to racemize at the N-terminus, although the rates of racemization were lower than those of PFHSPSY (Fig. 5A). Racemization should be an equilibrium reaction, and this was established by incubation of (d-Pro)FHSPSY under the same conditions as PFHSPSY. Interestingly, the rate of the reverse reaction, i.e. (d-Pro)FHSPSY to PFHSPSY, was lower than that of conversion of the l-peptide to the d-Pro form (Fig. 5B). A similar phenomenon was observed with another peptide (MWELR) (Fig. S2).
In order to determine whether racemization occurs at physiological temperature, the incubation of PFHSPSY was repeated at 37 °C. As expected, the rate of racemization of the N-terminal Pro in PFHSPSY was lower (0.04 nmol·h−1 at 37 °C compared with 2.83 nmol·h−1 at 60 °C) (Fig. 6), but still reached levels of ~ 5% after 10 weeks. Racemization of the N-terminal Pro residue occurred at a similar rate in HEPES (100 mm, pH 7.4) and TES buffers (100 mm, pH 7.4), thus ruling out a major buffer effect. Decreasing the pH using MES (pH 5.4) resulted in an approximately sixfold reduction in the rate of N-terminal Pro racemization at 60 °C (data not shown). A possible role for metal ions was investigated by repeating the incubation of PFHSPSY in phosphate buffer (pH 7.4) with addition of 1 mm EDTA; however, N-terminal racemization occurred at a similar rate to the control (data not shown).
Having established that N-terminal racemization of peptides occurs under physiological conditions, a long-lived protein was examined to determine whether racemization was detected. Aquaporin 0, an abundant integral membrane protein in the lens [33, 34] was chosen. Aquaporin 0 has a free N-terminal methionine residue that is located within the cytosol. To investigate N-terminal racemization of aquaporin 0, fibre cell membranes from human lenses were isolated as described previously  and then treated with trypsin. The sequence of the N-terminal tryptic peptide of human lens aquaporin 0 is MWELR . Tryptic digests were then analysed by two methods. First, samples of the digest were examined by semi-preparative HPLC using conditions under which the peptide standards (l-Met)WELR and (d-Met)WELR were well resolved [(l-Met)WELR at 47.7 min and (d-Met)WELR at 50.9 min].
Peaks from the lens digests eluting at the retention times of (l-Met)WELR and (d-Met)WELR were collected and examined by MALDI mass spectrometry. Both the cortex (the outer, most recently synthesized, part of the lens) and the nucleus (which contains proteins synthesized pre-natally) of the human lenses were analysed. For both lens regions, peaks at the elution positions of (l-Met)WELR and (d-Met)WELR were collected. The two HPLC peaks from the cortex and the two HPLC peaks from the nuclear extracts were collected and found to contain an ion at m/z 734.3. This ion corresponds to the molecular ion of (l-Met)WELR [or (d-Met)WELR]. MALDI MS/MS analysis of each m/z 734 ion confirmed its identity by comparison with the MS/MS spectrum of synthetic (l-Met)WELR (data not shown).
In the second method, ESI LC-MS/MS of unfractionated tryptic digests from pooled cortex and nuclear fractions of five pairs of human lenses from subjects aged 75–85 years was used to confirm the identification, and to gauge the extent of racemization more accurately. A selected ion monitoring trace of the doubly charged molecular ions corresponding to MWELR (368.6 [M+2H]2+) isolated from the lens nuclear extract is shown in Fig. 7. The m/z 368.6 ion from each of the four peaks (i.e. the l- and d-Met peptides from cortex and nucleus) was subjected to MS/MS fragmentation and compared with the MS/MS spectra of (l-Met)WELR/(d-Met)WELR standards (Fig. S3). In each case, MS/MS analysis confirmed the identity of the peptide as (l-Met)WELR/(d-Met)WELR. Using the areas under the curves, it was estimated that ~ 28% of the total aquaporin 0 present in the nucleus of aged human lenses is present as the d-Met form, whereas ~ 13% of aquaporin 0 in the outer part of the lens had been racemized.
Long-lived proteins in the body undergo a number of modifications during extended exposure to physiological conditions [1, 2, 36]; however, details of several reactions remain to be elucidated. In this study, it was demonstrated that incubation of peptides at pH 7 results in facile racemization of the N-terminal amino acid. This was observed at 37 °C, and the rate was increased at higher temperatures. In reactions using model peptides, racemization was confined to the N-terminal amino acid residue. One potential mechanism that may account for this is shown in Fig. 8. This involves a central Schiff base intermediate, which may decompose via one of two pathways: one yielding a diketopiperazine and one producing a racemized N-terminus.
Evidence in support of this mechanism comes from the fact that two of the three peptide standards (SFHSPSY and AFHSPSY) incubated at 60 °C showed significant racemization of the N-terminal residue, with the racemization rate of N-terminal Ser approaching that of Pro (Fig. 5A), and each also displayed some loss of an N-terminal dipeptide (data not shown). Cleavage of the two N-terminal amino acids of a peptide, linked together as a diketopiperazine, has been described in detail by Steinberg and Bada .
It is proposed that racemization of the N-terminal residue and loss of a cyclic diketopiperazine are competing reaction pathways following formation of the Schiff base, with steric factors determining the relative formation of each. The case of PFHSPSY, for which diketopiperazine formation was not detected, may be considered an extreme example of this reaction. In this case, N-terminal racemization is favoured almost exclusively. It is also possible that other processes contribute to the overall reaction pathway leading to loss of the N-terminal amino acid, for example.
Therefore, racemization is not the only spontaneous reaction that involves the N-terminal amino group. The mechanism as outlined provides a pathway for formation of a racemized N-terminal amino acid, as well as cleavage of the combined N-terminal and penultimate residues as a diketopiperazine. A similar mechanism was proposed by Sepetov et al. . Although not all amino acids were investigated as N-termini in this study, those that were showed significant racemization. Further evidence in support of this mechanism is shown in Fig. 5B, where PFHSPSY with an l- form N-terminal amino acid was found to racemize at a faster rate than the corresponding d- form N-terminal amino acid. If racemization occurred via simple abstraction and re-addition of the α proton to the N-terminal amino acid, these rates would be similar.
In addition, a stereo-specific difference in the rate of inter-conversion was noted for another peptide, which contained an N-terminal Met (Fig. S2). For both the Pro and Met peptides, conversion rates to the other isomer were higher when starting with the l-peptides than the corresponding d-peptides (Fig. 5B). The reason for this is not known, but presumably reflects conformational restrictions involved in formation of the Schiff base intermediate and the degree to which it can isomerize.
To determine whether N-terminal racemization is a feature of long-lived proteins, aquaporin 0 from human lenses was investigated. Lens proteins do not turn over after their incorporation into mature fibre cells, and are thus present for life [17, 39]. Aquaporin 0 is an integral membrane protein that acts as a water channel and may also facilitate cell–cell interactions within the lens . To study whether the N-terminus of aquaporin 0 from adult lenses was racemized, we took advantage of a technique used by Schey et al.  that involves tryptic digestion of intact lens membranes.
As the N-terminus of aquaporin 0, as well as the C-terminus, are cytosolic , they are exposed in purified lens membrane fractions and may be selectively proteolysed by enzymes such as trypsin. Analysis of the nuclear region of aged lenses by LC-MS/MS revealed a high level (28%) of racemization of the N-terminal amino acid. A reduced amount of N-terminal racemization was detected in the cortex (13%) (data not shown), in accordance with the fact that nuclear proteins are, on average, older than the cortical proteins .
This study demonstrated a novel modification whereby the N-terminal residues of three peptides that terminated in Pro, Ser and Ala all underwent spontaneous racemization when incubated at pH 7.4. Incubation of PFHSPSY at 37 °C was performed to demonstrate that significant N-terminal racemization (~ 5% over 10 weeks) occurred under physiological conditions. Given the decades that some long-lived proteins are exposed to in a similar environment, it is proposed that this is a new post-translational modification to which long-lived proteins with free amino terminals may be subject. A biological example of this modification (aquaporin 0) is provided, and work is now underway to investigate whether this modification is present in other long-lived proteins.
Racemization of internal amino acid residues such as Asp/Asn [22, 43] and Ser [44, 45] in old proteins is well known. Recent data suggest that racemization rates at these sites are governed not only by the adjacent amino acid residue  but also by the structure of the protein at the particular site . Asp and Asn residues in unstructured parts of crystallins are much more susceptible to racemization than those present in β-sheets .
The ramifications of N-terminal racemization for protein structure and function are as yet unknown, and its impact will probably vary depending on the particular protein. In terms of susceptibility to enzymatic proteolysis, it is very likely that having a d-amino acid at the N-terminus will inhibit cleavage of a long-lived protein by exopeptidases, as they are generally inactive on the d-isomers . This was illustrated using the N-terminal peptide of aquaporin 0. When exposed to leucine aminopeptidase, a protease known to be widely distributed and present in the lens , the l(Met) peptide was rapidly degraded whereas the corresponding d(Met) peptide was stable (Fig. 9).
Some aminopeptidases are active on intact proteins . Leucine aminopeptidase appears to use peptides as its major substrates, but may also have a low activity with intact proteins. It is not known whether aminopeptidase degradation of long-lived proteins is a physiologically relevant process.
In summary, racemization of the N-terminal amino acid residue is a novel reaction that may be observed in polypeptides that contain free α amino groups. It joins an expanding array of reactions to which long-lived proteins in the body are susceptible.
All peptides were synthesized by GLS Biochem (Shanghai, China). TFA (Sigma, St Louis, MO, USA) was of spectrophotometric grade. Na2HPO4 and NaH2PO4 (high purity) were purchased from Amresco (Solon, OH, USA). Sequence-grade leucine aminopeptidase from porcine kidney was purchased from Sigma. Trypsin (sequence grade) was purchased from Promega (Madison, WI, USA). All solutions were prepared in MilliQ water (Waters, Billerica, MA, USA).
Peptides PFHSPSY, SFHSPSY, AFHSPSY and MWELR were dissolved in triplicate in 100 mm phosphate buffer pH 7.4 (1 mg·mL−1) and incubated at 60 °C. PFHSPSY, SFHSPSY and AFHSPSY are based on a region of αB-crystallin (16–21) sequence but with varied N-termini and with Tyr as the C-terminus. MWELR corresponds to the N-terminal tryptic peptide of human aquaporin 0. A drop of chloroform was added to each tube to prevent microbial growth. Aliquots (20 μL) were taken at regular time points and analysed by HPLC.
Peptides (l-Met)WELR and (d-Met)WELT were incubated (1 mg·mL−1) with 1 μg leucine aminopeptidase (1 : 000 enzyme : substrate) in 100 mm Tris buffer, pH 8.5, at 30 °C.
HPLC analysis and quantification
An Agilent 1100 HPLC system (Agilent Technologies, Santa Clara, CA, USA) controlled using Chemstation software and equipped with a PDA detector was used. Incubations were monitored at 280 and 216 nm. Separation of the peptides was achieved using a Jupiter Proteo 4 μm 90 Å column (150 mm length, 4.6 mm internal diameter) at 40 °C. The gradient was 0% B (0.1% TFA in acetonitrile) to 60% B (0.1% TFA in acetonitrile) over 25 min. A standard curve was generated for each peptide. The degree of racemization was calculated based on the moles of each peptide formed as a percentage of moles of peptide present at the start of the incubation. The error bars refer to the standard deviation of the replicates.
Semi-preparative HPLC purification
A Shimadzu Prominence HPLC system (Shimadzu, Kyoto, Japan) controlled by Shimadzu Class VP software equipped with a UV-vis detector (SPD-20A) and a fraction collector (FRC-10A) was used. Purification of the peptides was achieved using a Phenomonex Kinetex (Torrance, CA, USA; 100 mm length × 4.6 mm internal diameter) 2.6 μm 100 Å column at ambient temperature and was monitored at 280 and 216 nm. The gradient was 0% B (0.1% TFA in acetonitrile) to 60% B (0.1% TFA in acetonitrile) over 110 min.
Chiral amino acid analysis
Chiral amino acid analysis by HPLC was performed as described by Goodlett et al. . Peptides were hydrolysed in 6 m HCl at 110 °C for 6 h, and then lyophilized.
Nano-liquid chromatography (nano-LC) was performed using an Ultimate 3000 HPLC and autosampler system (Dionex, Amsterdam, The Netherlands). Samples were injected into a fritless nanoLC column (75 μm × 10 cm) containing Bruker (Billerica, MA, USA) C18 medium (3 μm, 200 Å; Michrom) produced as described by Gatlin . Peptides were eluted using a linear gradient from 2% B to 90% B over 37 min at a flow rate of 0.25 μL·min−1. Mobile phase A consisted of 0.1% formic acid in H2O, while mobile phase B consisted of acetonitrile : H2O (8 : 2) with 0.1% formic acid. High voltage (1800 V) was applied to a low-volume tee (Upchurch Scientific, Oak Harbor, WA, USA), and the column tip was positioned ~ 0.5 cm from the heated capillary (T = 250 °C) of a LTQ FT Ultra mass spectrometer (Thermo Electron, Bremen, Germany).
Positive ions were generated by electrospray, and the LTQ FT Ultra was operated in data-dependent acquisition mode. A survey scan of m/z 350–1750 was acquired in the FT ICR cell (resolution = 100 000 at m/z 400, with an accumulation target value of 1 000 000 ions). Up to six of the most abundant ions (> 3000 counts) with charge states > +2 were sequentially isolated and fragmented within the linear ion trap using collisionally induced dissociation with an activation q = 0.25 and activation time of 30 ms at a target value of 30 000 ions. The m/z ratios selected for MS/MS were dynamically excluded for 30 s.
Samples were prepared in 50 mm phosphate buffer, pH 7.4, containing 1 mm 4,4-dimethyl-4-silapentane-1-sulfonic acid and 10% D2O. Spectra were acquired at 25 °C on an Avance III 800 MHz NMR spectrometer (Bruker, Karlsruhe, Germany) equipped with a triple-resonance TCI cryoprobe. 1D 1H, 2D 1H-TOCSY, 2D 1H-ROESY and 2D 13C-1H HSQC experiments were acquired using standard Bruker pulse sequences.
Lens membrane digestion
Five lens pairs (from subjects aged 75–85 years) were dissected into nuclear and cortical regions using a 6 mm trephine. Tissues from the cortical regions were combined as were the nuclear regions. Lens membranes were isolated as described by Schey et al. . In brief, each lens region was homogenized in Tris buffer, pH 8.0, containing 6 M guanidine HCl to remove soluble and water-insoluble proteins. Extracts were centrifuged at 16 000 g (30 min, 4 °C) and the supernatants were removed. This was repeated five times, then the pellets were homogenized with water and centrifuged. This washing was repeated five times, and the pellets were freeze dried. The pellets were suspended in 50 mm ammonium carbonate, pH 8.0, and 20 μg trypsin was added. Samples were incubated at 37 °C for 16 h, and then freeze dried.
This study was supported by a grant from the National Health and Medical Research Council (NHMRC) (grant number 512334).