Chemical synthesis of a masked analogue of the fi sh antifreeze potentiating protein ( AFPP ) †

A recently identified Antarctic fish protein termed antifreeze potentiating protein (AFPP) is thought to act as an adjunct to the previously characterised antifreeze glycoproteins (AFGPs), the two acting together to inhibit ice crystal growth in vivo. Elucidating the functional properties of the new AFPP requires access to large amounts of pure product, but the paucity of natural material necessitates alternative approaches. We therefore embarked on the total chemical synthesis of the AFPP, through a convergent ligation strategy. After many challenges, mostly due to the solubility issues of the peptide fragments, and several revisions of the original synthetic strategy, we have successfully synthesized a masked analogue of AFPP. The key to the successful synthesis was the use of a solubilising tag attached through a hydrolysable linker.


Introduction
Antarctic fishes survive in freezing seawater (ca.−1.9 °C) through the production of protective antifreeze proteins. 1In previous work the DeVries laboratory isolated an Antarctic fish protein 2,3 termed antifreeze potentiating protein (AFPP) which can be considered a novel adjunct to the previously characterized antifreeze glycoproteins (AFGPs) circulating in the blood. 3ike AFGP, AFPP isolated from Antarctic fish serum binds to ice crystals, inhibiting their growth.Almost all Antarctic fish exposed to freezing seawater that we have analyzed to date, have internalized ice crystals sequestered in macrophages within the spleen. 4,5Since such crystals will contain adsorbed AFPP and AFGP, these observations raise the possibility that AFPP, in conjunction with AFGP, not only inhibits ice crystal growth in the blood but might also promote removal of potentially lethal circulating ice crystals by facilitating uptake of the ice/antifreeze complexes into spleen macrophages.One possibility is that this is a receptor-driven process (Fig. 1), the experimental evaluation of which will require significant amounts of pure AFPP.
To investigate the role of AFPP in this hypothesis, access to sufficient amounts of pure protein is necessary.Additionally, modified AFPP analogues labelled with specific spectroscopic probes are required to facilitate further studies for functional analysis using biophysical methods (e.g.autoradiography; fluorescence microscopy; receptor characterization with photoactive cross-linkers).However, generation of functionally modified AFPP cannot be achieved by recombinant methods, and the paucity of natural material necessitates alternative approaches.Fortunately, the atom-by-atom precision achievable through chemical synthesis enables the site-specific incorporation of labels, 6 allowing us to examine adsorption of AFPP to ice and determine the nature of the postulated molecular interaction with macrophages in the spleen.We therefore embarked upon the total chemical synthesis of the new antifreeze potentiating protein.

A. First ligation strategy
Available evidence suggests that AFPP is a member of the C1qlike protein family. 7A C1q-like domain was identified in the Antarctic naked dragonfish Gymnodraco acuticeps using molecular cloning techniques and its expression product was found to cross-react with an anti-AFPP. 8An analogue of the derived C1q-like domain sequence (lacking a glycine at position 129, and with three additional amino acids at the N-terminus likely present in the native product, Fig. 2) was used in this study.This C1q-like analogue (referred to as C1q-AFPP) is 134 amino acids in length, well within what can be achieved 9 by chemical synthesis through native chemical ligation (NCL).NCL is the chemoselective reaction between an N-terminal cysteine and a C-terminal thioester to give a native peptide bond. 10Unfortunately, the C1q-like domain of AFPP does not contain any natural cysteines, hence a ligation/desulfurization strategy was considered.First introduced by Yan and Dawson in 2001, 11 this method has extended the protein junctions accessible to NCL.In their work, native alanine-containing proteins were accessed through post-ligation desulfurization of cysteine, and more recently a variety of other thiolated building blocks that serve as precursors to natural amino acids have successfully been used in this strategy. 12CL of multiple peptide segments is only suitable for sequential ligations from the C-to the N-terminus.Yet because a convergent synthesis is more efficient, 13 we decided to use kinetically controlled ligation 14 (KCL) as part of a fully convergent ligation strategy.We aimed to split the protein sequence into peptides of approximately equal length.The N-terminal half of the sequence (see Fig. 2) was ideally suited for a KCL, as it had alanines at position 36 and 68.Substituting these for cysteines would allow a kinetically controlled ligation between fragment 1 (Pro 1 -Gln 35 thioaryl ester) and fragment 2 (Cys 36 -Thr 67 thioalkyl ester).The relative difference between the rate of ligation between a thioaryl ester and a thioalkyl ester 14 (the basis of the KCL reaction) is further supported by the difference in reactivity between glutamine and threonine thioesters, 15 thereby allowing the desired ligation to occur.
Unfortunately the C-terminal half of the sequence did not have a suitably placed alanine.The alanine at position 107 was most central, yet this ligation position would give a valine thioester for fragment 3. Valine thioesters are notoriously nonreactive, their ligation often requiring multiple days to go to completion. 15Using one of the other alanines present (Ala93/ 94) in the sequence as a ligation site would leave a very long (>40 aa) fragment 4, which is approaching the limit for synthesis by solid phase methods.Given the recent reports of ligation at valine (using the valine surrogate dimethylcysteine or penicillamine (Pen)) 16,17 we decided on a ligation site at Val 106 instead.
Synthesis of peptide fragments.We have previously observed that even short AFPP sequences are prone to irreversible aggregation in solution, making their purification and further manipulation extremely difficult.Previous work has shown that addition of charged amino acids such as arginine or lysine to the C-or N-terminus can dramatically reduce aggregation and improve solubility of peptides and proteins. 18his strategy was successfully applied by Johnson and Kent to the synthesis of hydrophobic peptide thioesters, where the charged amino acids were used as part of the thioester leaving group. 19We decided on a similar approach for the three thioester-containing fragments, using multiple lysine residues.
Examination of the amino acid composition of C1q-AFPP (Fig. 2) shows that six methionine residues are present, whose methylthioether side chain is easily oxidized under acidic conditions that occur during Boc solid-phase peptide synthesis (SPPS) protocols.This can result in mixtures of reduced and oxidised methionine, which complicates analysis and purification by HPLC methods, and hence we incorporated methionine in its methionine sulfoxide form.Additionally, using Met(O) would increase the overall polarity of the peptides, which may increase their solubility.The sulfoxide would then be reduced [20][21][22] at the end of the synthesis to give the native methionine.
N-terminal half of C1q-AFPP.Fragment 1 (Pro 1 -Gln 35 -COS-CH 2 CH 2 -Ala) (1) and fragment 2 (Cys 36 -Thr 67 -COS-CH 2 CH 2 -Lys 4 ) (2) were synthesized using in situ neutralisation Boc-SPPS under standard reaction conditions (Scheme 1). 23The synthesis of fragment 1 (1) gave a single product in high purity (see Fig. S1 †).The peptide was quite insoluble in aqueous acetonitrile mixtures, but was fully soluble in the chaotrope 6 M guanidinium chloride (Gu•HCl).Subsequent exchange of the thioalkyl ester to its thiophenyl ester using the water soluble thiol 4-mercaptophenylacetic acid (MPAA) 24 (1 mM peptide concentration, 100 mM MPAA, 20 mM tris(2-carboxyethyl)phosphine hydrochloride (TCEP•HCl), pH 6.5) on the crude material proceeded smoothly in 2-3 h, giving Pro 1 -Gln 35 -COS-MPAA (3).The synthesis of fragment 2 (2) also gave the desired peptide as the major product (see Fig. S2 †), and both peptide 2 and peptide 3 were purified under standard conditions.With the two fragments in hand, their ligation through the KCL reaction was attempted (Scheme 1).Unfortunately, the KCL was hampered by low solubility of both fragments, even with the presence of the Lys-solubilising tag on fragment 2 (2).Recent work has shown that ligation reactions can proceed in organic solvents, 25 so the KCL was attempted in DMF.This was once again unsuccessful, as peptide 2 exhibited very low solubility in this system.Furthermore, any addition of Na 2 HPO 4 buffer resulted in precipitation of the fragments, and the reaction could not be buffered to optimal pH for KCL.Eventually, dimethyl sulfoxide (DMSO) was found to be the optimal solvent.Dissolution of fragment 1-MPAA (3) and fragment 2 (2) in DMSO followed by the addition of 20 mM TCEP in Na 2 HPO 4 ( pH adjusted to 6.9, final ratio 9 : 1) resulted in near quantitative ligation after 70 min, to give the first half of the AFPP (Pro 1 -Thr 67 -COS-CH 2 CH 2 -Lys 4 ) (4) (see Fig. S3 †).
As fragment 4B (Pen 98 -Gln 134 COOH) (7) does not contain a thioester, it was conveniently synthesized using Fmoc chemistry aided by a microwave synthesizer.Fmoc-Glu-OtBu was attached via its side-chain carboxyl group to a Rink amide linker to give the desired C-terminal glutamine upon acid-mediated cleavage at the end of the synthesis. 27,28enicillamine ligation issues.The penicillamine ligation was attempted under standard NCL conditions (1 mM peptide concentration, 100 mM MPAA, 20 mM TCEP•HCl, pH 6.5).Fragment 3B (6) rapidly underwent transthioesterification to the more active MPAA thioester intermediate, which could be observed by LC-MS.Unfortunately the subsequent steptransthioesterification with the penicillamine followed by intramolecular rearrangement to the desired ligated peptideproved to be rate-limiting.The very slow ligation reaction was left overnight, during which time a significant portion of fragment 3B (6) underwent hydrolysis.While some of the desired ligation product was also observed, the long reaction time in the presence of the reducing agent TCEP caused some of the Met(O) residues in the peptides to be reduced to methionine, further complicating the LC-MS chromatogram.Attempts to optimize the ligation conditions were unsuccessful, and the hydrolysis of fragment 3B (6) continued to be the dominant reaction, leading to very poor yields of the desired product.
Fragment 4C solubility issues.We therefore needed to revise our ligation strategy once more, and hence moved the ligation site to Ala 94 ; consequently fragment 4C was now 41 amino acids long (see Scheme 2c).While the synthesis and purification of the shorter fragment 3C (Thz 68 -Ala 93 -COS-CH 2 CH 2 -Lys 5 ) (8) proceeded smoothly (see Fig. S6 †), fragment 4C (Cys 94 -Gln 134 COOH) (9) was more difficult.As expected, the length of this peptide resulted in significant solubility issues, and standard chromatography conditions were not successful at resolving the peptide from the by-products.After many trials, fragment 4C (9) was eventually successfully purified (see Fig. S7 †) on a diphenyl reversed-phase column eluted with a 1 : 1 mixture of CH 3 CN and i PrOH containing 0.1% TFA for solvent B, at 55 °C.Preparation of fragment 4C (9) was also attempted using the isoacyl dipeptide strategy (using Boc-Ser-(Fmoc-Ser(O t Bu))-OH), under microwave-mediated SPPS.This strategy has been used previously to improve both the synthesis and solubility of otherwise intractable peptides. 29owever, LC-MS analysis showed that a peptide with a single Ser deletion was the major product formed.Hence this strategy was not pursued further.
NCL of fragments 3C and 4C.The first attempted NCL of fragments 3C (8) and 4C (9) at a conc. 2 mM revealed that both of the fragments exhibit low solubility in standard NCL reaction buffer solution.A slight improvement in solubility was observed when further diluted to 0.5 mM, but still, the peptides were not fully soluble at this concentration, even with sonication and an elevated temperature of 40 °C.Solubility tests on the peptides revealed that fragment 3C ( 8) is sparingly soluble in the NCL buffer while very poor solubility was observed for fragment 4C (9).
In order to improve the solubility of both fragments in the ligation reaction, a 1 : 1 mixture of TFE and 0.2 M Na 2 HPO 4 was used.While this indeed helped drive the NCL reaction to completion, a significant amount (50%) of the TFE ester of fragment 3C was also observed.This is not surprising, however, as fragment 3C (8) was used in excess and presumably unreacted thioester 8 undergoes hydrolysis by trifluoroethanol following completion of the ligation reaction.Furthermore, even upon short-term storage in the freezer, the purified fragment 4C (9) unexpectedly underwent dimerization, presumably via an intramolecular reaction through the sulfhydryl side chains of cysteine.The dimer could be readily reduced by treatment with TCEP in solution prior to the addition of the NCL partner, fragment 3C thioester (8).
Given that the ligation was now being performed in aqueous organic media, we investigated whether neutral organic phosphines would be more suitable and could be used in place of TCEP, which is typically supplied as the hydrochloride salt.Reduction of dimeric fragment 4 (9) was tested using three readily available phosphinestributylphosphine, triphenylphosphine, and tricyclohexylphosphinein 1 : 1 TFE-0.2M Na 2 HPO 4 or neat TFE.No significant solvent differences were found, and all phosphines performed in a similar manner.
Tributylphosphine (chosen for its convenience, being a liquid) was thus added to the solution of dimeric fragment 4C in a neat TFE solution, and was left standing at room temperature for 30 min.Thz-fragment 3C-thioester (8) was then added into the above reaction mixture followed by addition of equivolume of pH adjusted (approx.7.0) 100 mM MPAA in 0.2 M Na 2 HPO 4 .The reaction mixture was left overnight at room temperature, with subsequent LC-MS analysis indicating that the ligation went to completion, giving the desired C-terminal half of C1q-AFPP (Thz 68 -Gln 134 COOH) (10) (see Fig. S8 †).Unreacted excess thioester was converted to its TFE ester under these conditions, but did not affect the yield of the desired product.
Unmasking of Thz to Cys of fragment 3,4.We next attempted a one-pot conversion 30  Failure of final NCL reaction.The final NCL between fragment 1, 2 (4) and Cys-fragment 3-4 (11) (Scheme 3) was performed using the same reaction conditions used in the KCL reaction.Unfortunately, no desired product could be detected in the reaction mixture.Due to the ongoing solubility problems of the fragments, a new synthetic strategy was therefore sought.

B. Second, revised ligation strategy using a solubilising tag
The majority of the issues faced during this synthesis could be traced back to the poor solubility of the peptide fragments, especially of the terminal fragment 4 (9).Examination of the peptide sequence revealed that each fragment has very little overall charge and therefore was expected to exhibit poor water solubility.The polycationic solubility tags attached to the thioester of fragment 2 and 3 did improve their ease of handling, but unfortunately these are lost during the ligation reaction, exacerbating the solubility problem with each subsequent ligation step.We therefore decided to place a temporary solubilising tag 31 on the C-terminus of fragment 4, attached to a base-labile hydroxymethylbenzoic acid (HMBA) linker.This linker is stable to all the ligation conditions and both Fmoc and Boc SPPS, and being non-thioester based, would not interfere with any of the reactions.It would therefore remain on the peptide throughout the various synthetic and chromatographic steps, to be removed at the very end of the synthesis using aq.buffer at pH 8. 32  also moved the ligation point between fragment 2 and 3 to Phe 71 /Ala 72 (Fig. 3), as this would give the more reactive Phe 71thioester for fragment 2. The new ligation point also has consequences for the fragment 1 and 2 KCL reaction.According to Hackeng et al., a phenylalanine thioester is slightly more reactive than a glutamine thioester. 15However, according to the relative reactivities of aryl vs. alkylthioesters, 14 the KCL reaction would still be the favoured one.The new, revised ligation strategy is presented in Scheme 4. Synthesis of peptide fragments.View Article Online were synthesized using in situ neutralisation Boc-SPPS under standard reaction conditions.Fragment 4 (Cys 94 -Phe 132 -HMBA-Lys 5 ) ‡ ( 15), being a non-thioester, was synthesized using Fmoc SPPS on a Rink amide resin with the aid of a microwave peptide synthesizer.Pleasingly, purification of all peptide fragments could be achieved using standard chromatographic conditions (solvent A = water + 0.1% TFA and solvent B = CH 3 CN + 0.1% TFA) (see Fig. S10 †).N-terminal half of C1q-AFPP.The MPAA exchange of ment 1 (12) was carried out as described for the previous strategy, to give the MPAA thioester 3 in 2-3 hours (see Fig. S11 †).The KCL of fragment 1 MPAA thioester 3 and the Cys-fragment 2 (13), using 50 mM TCEP and a 1 mM final peptide concentration, proceeded smoothly and was complete after approximately 4 hours (see Fig. 4a).The N-terminal half of C1q-AFPP (Pro 1 -Phe 71 -COS-CH 2 CH 2 -Lys 5 ) ( 16) was obtained after standard chromatographic purification (see Fig. 4b).
Unmasking of Thz in fragment 3,4 to Cys.Unfortunately, once again a one-pot Thz to Cys conversion of 17 proved to be problematic.Addition of 0.2 M MeONH 2 •HCl ( pH adjusted to approx.4) after the completion of the NCL led to the decomposition of the Thz 72 -Phe 132 -HMBA-Lys 5 -NH 2 -fragment (17), and the appearance of a new, unknown peak on the LC-MS chromatogram.Use of 2 M MeONH 2 •HCl, or hydrazine, led to the generation of an unknown by-product or the hydrazide ester, respectively.Various attempts were made to optimize this reaction, including changes to total peptide concentration, MeONH 2 •HCl concentration, and other aqueous conditions.Our results indicated that a low final concentration of peptide and higher concentration of MeONH 2 •HCl was required for this reaction to proceed without the formation of by-products.
Activity testing.The final NCL product (19) was tested for antifreeze activity using a Clifton nanoliter osmometer as described elsewhere. 1,33The defining characteristic of biological antifreeze is the presence of hysteresis activity, measurable as a difference in the freezing and melting points of a small ice crystal in a solution of antifreeze.Soluble, native AFPP has a hysteresis activity of approximately 2 °C at 4 mg mL −1 in 50 mM NH 4 HCO 3 buffer ( pH 7.8), when determined with a small seed crystal, slow cooling and annealing for 1 h at 0.1 °C above the hysteresis freezing point.Despite the presence of a solubilizing tag to aid synthesis, attempts to dissolve synthetic C1q-AFPP at concentrations up to 2 mg mL −1 in the same bicarbonate buffer were unsuccessful.This necessitated solubilizing the product first in 1 : 1 acetonitrile-50 mM NH 4 HCO 3 with subsequent removal of the acetonitrile by dialysis (2000 MW cut-off ) against the bicarbonate buffer and reduction to a final concentration of approximately 4 mg mL −1 .Further concentration led to increasing precipitation.Hysteresis determinations were highly variable with a maximum of 0.76 °C and a minimum of 0.22 °C.The maximum hysteresis value was observed when the seed ice crystal was very small (<10 µm), the cooling rate was 0.01 °C min −1 and the sample was annealed for 1 h at slightly above its hysteresis freezing point.Significantly, the morphology of ice growth at the freezing point was in the form of expanding irregular hexagons (a-axis growth), which is the same as the growth direction with native AFPP.
A second distinguishing characteristic of AFPP is that it potentiates the hysteresis activity of AFGP, especially the larger AFGP1-5 isoforms. 2When synthetic C1q-AFPP was tested in the presence of AFGP1-5, however, no potentiation was observed.Several explanations are possible, one being that the synthetic analogue may not have folded fully into its native conformation, as attested by its poor solubility.

Conclusions
We have demonstrated that a temporary solubilising tag can be successfully used in a convergent native chemical ligation synthesis of a poorly water soluble protein.The solubilising tag proved to be essential to the synthesis of a masked analogue of C1q-AFPP, allowing for improved handling and purification of protein fragments.The 132 amino acid protein with a pentalysine tag was successfully synthesised through a convergent strategy using both kinetically controlled ligation and native chemical ligation to join four separate peptide fragments.

Fig. 1
Fig. 1 Possible involvement of AFPP and AFGP in the removal of circulating ice crystals by macrophages in the spleen.

Fig. 2
Fig.2The derived amino acid sequence of the C1q-AFPP from Gymnodraco acuticeps with proposed ligation sites (first ligation strategy) marked with arrows.
of the N-terminal thiazolidine group to cysteine of peptide 10.Unfortunately, decomposition of the ligated Thz 68 -Gln 134 -OH (10) was observed overnight when 0.2 M MeONH 2 •HCl was added directly to the NCL reaction mixture.It was suspected that the presence of tributylphosphine may impede the unmasking of Thz to Cys.Indeed, with the presence of tributylphosphine in the reaction mixture, addition of MeONH 2 •HCl led to the formation of an unknown side product.By comparison, clean formation of desired Cys-fragment 3,4-OH (11) was observed (see Fig. S9 †) when the tributylphosphine was completely removed by solidphase extraction (SPE) prior to the addition of MeONH 2 •HCl.Interestingly, the extraction step was not required when TCEP was used, allowing the addition of MeONH 2 •HCl directly to the crude NCL reaction mixture.
Given the failure of the final NCL ligation, we Paper Organic & Biomolecular Chemistry 4938 | Org.Biomol.Chem., 2013, 11, 4935-4942 This journal is © The Royal Society of Chemistry 2013