Determination of AMP-activated protein kinase phosphorylation sites in recombinant protein expressed using the pET28a vector: a cautionary tale.

AMP-activated protein kinase (AMPK) is responsible for sensing of the cell's energetic status and it phosphorylates numerous substrates involved in anabolic and catabolic processes as well as interacting with signaling cascades. Mutations in the gene encoding the gamma 2 regulatory subunit have been shown to cause hypertrophic cardiomyopathy (HCM) with conduction abnormalities. As part of a study to examine the role of AMPK in the heart, we tested whether specific domains of the thick filament component cardiac myosin binding protein-C (cMyBP-C) were good in vitro AMPK substrates. The commercially available pET28a expression vector was used to generate a recombinant form of the cMyBP-C C8 domain as a fusion protein with a hexahistidine tag. In vitro phosphorylation with activated kinase showed that the purified fusion protein was a good AMPK substrate, phosphorylated at a similar rate to the control SAMS peptide and with phosphate incorporation specifically in serine residues. However, subsequent analysis of alanine replacement mutants and thrombin digestion revealed that the strong AMPK phosphorylation site was contained within the thrombin cleavage sequence encoded by the vector. As this sequence is common to many commercial pET vectors, caution is advised in the mapping of AMPK phosphorylation sites when this sequence is present.

AMP-activated protein kinase (AMPK)1 is an enzyme found in all mammalian cells and is responsible for sensing energetic status. It is activated by 5′-AMP and phosphorylation, and itself phosphorylates numerous substrates involved in anabolic and catabolic processes as well as interacting with signaling cascades [1,2]. It is a trimer composed of a catalytic α subunit, a β subunit that includes a glycogen-binding domain and a γ subunit that binds adenine nucleotides via two pairs of cystathione β synthase (CBS) domains. Mutations in the γ2 isoform have been shown to cause hypertrophic cardiomyopathy (HCM) with conduction abnormalities [3].

In order to investigate the precise role of AMPK in the heart, we have performed yeast two-hybrid screenings of human heart cDNA libraries using AMPK subunits as bait. A screen using the short splice variant of the γ2 isoform [4,5] as bait identified the C-terminal domains of cardiac myosin binding protein-C (cMyBP-C) as a putative interactor. This protein is a component of thick myofilaments of cardiac muscle and is postulated to play a role in the regulation of cardiac muscle contractility [6]. It is a modular protein composed of 11 domains (termed C0–C10), eight of which are IgI-like, the other three being fibronectin three-like, and mutations in the gene encoding cMyBP-C are a common cause of HCM [7].

This report describes the use of the commercially available pET28a expression vector to generate recombinant cMyBP-C C8 domain as a fusion protein with a hexahistidine tag. In vitro phosphorylation of the fusion protein with purified activated kinase suggested that this domain was a good AMPK substrate. However, subsequent analysis revealed that the strong AMPK phosphorylation site is contained within the thrombin cleavage sequence encoded by the vector.

Materials and methods

Cloning and protein production

DNA encoding the C8 domain of cMyBP-C (amino acids 975–1066) was amplified by PCR and cloned into the bacterial expression vector pET28a (Novagen®, Fig. 1) using NdeI and HindIII. Alanine replacement mutants of serines within the C8 sequence were generated by our standard two-step PCR mutagenesis protocol using complementary primers. Plasmid DNA was purified according to standard procedures and the sequence of the pET28a-C8 constructs verified.

Fig. 1

Partial sequence of pET28a showing the translation start site, hexahistidine tag and thrombin recognition site. The serines encoded within the leader peptide are boxed; the serine within the thrombin cleavage site shown to be phosphorylated in this study is indicated by a shaded box.

pET28a-C8 DNA was used to transform BL21(DE3)pLysS Escherichia coli and his6-C8 protein was overexpressed in cultures of transformed cells grown in L broth containing 30 μg/ml kanamycin and 25 μg/ml chloramphenicol at 37 °C. Expression was induced by the addition of isopropyl β-d-1-thiogalactopyranoside to 0.4 mM after which the cultures were grown for a further 3 h; cells were harvested by centrifugation and the cell pellets frozen at −80 °C. For purification, cell pellets were thawed, resuspended in 50 mM sodium phosphate pH 8.0, 300 mM NaCl, 10 mM imidazole and cell lysis brought about by sonication. Cell debris was removed by centrifugation (20,000 g for 20 min at 4 °C) and his6-C8 protein isolated from the lysis supernatant using Ni2+–NTA (Qiagen) affinity chromatography according to the manufacturer’s instructions (Fig. 2A). Digestion of his6-C8 fusion protein with thrombin (1 U/μg protein) was carried out at 4 °C overnight in 20 mM Tris–HCl pH 8.4, 150 mM NaCl, 2.5 mM CaCl2.

Fig. 2

Phosphorylation of his6-C8 by AMPK. (A) One microgram of his6-C8 was phosphorylated by AMPK in the presence of γ32P-ATP as described in Materials and methods and separated by SDS–PAGE. Lane 1: Coomassie Blue-stained gel; lane 2: autoradiograph. (B) Time course of AMPK phosphorylation of 800 pmol SAMS (filled circles) and 800 pmol his6-C8 (open circles). Phosphorylation reactions were set up as described; aliquots were taken at the indicated time points and spotted onto phosphocellulose paper. The radioactivity remaining on the paper after washing with 10% TCA was determined by scintillation counting. (C) Phosphoamino acid analysis of 32P-labeled his6-C8. Total amino acids were separated by electrophoresis at pH 1.9 (first) and pH 3.5 and 32P-amino acids detected by autoradiography. Position of stained phosphoamino acid standards are marked.

Phosphorylation studies

Phosphorylation reactions were carried out at 37 °C for 1 h with 0.37 MBq γ32P-ATP in 50 mM HEPES pH 7.4, 150 mM KCl, 25 mm MgCl2, 1 mM dithiothreitol, 125 μM AMP, 1 mM ATP in the presence of activated recombinant α1β1γ1 AMPK trimer [8]. Phosphorylation was analyzed by SDS–PAGE and autoradiography or by spotting onto phosphocellulose paper and scintillation counting [9].

Phosphoamino acid analysis was carried out according to a standard protocol [10]. Briefly, 32P-labeled protein prepared as described above was precipitated with 20% trichloroacetic acid (TCA) and hydrolyzed in 5.7 M HCl at 100 °C for 30 min. After removal of HCl by evaporation, the amino acids were resuspended in pH 1.9 electrophoresis buffer with phosphoserine, phosphothreonine and phosphotyrosine standards. The mixture was then separated by electrophoresis on a cellulose plate in two dimensions, first at pH 1.9 and then at pH 3.5. Phosphoamino acid standards were visualized by ninhydrin staining and 32P-labeled amino acids by autoradiography.

Results

Our unpublished yeast two-hybrid data indicated a putative interaction between the α1 subunit isoform of AMPK and domains 8–10 of cMyBP-C, thus implying that this region of cMyBP-C may also be a substrate for this kinase. To test the latter, we used recombinant C8 domain (amino acids 975–1066) produced as a hexahistidine fusion protein using the vector pET28a (Fig. 1). Incubation of his6-C8 with activated AMPK resulted in incorporation of phosphate into the recombinant protein (Fig. 2A). Furthermore, a time course suggested that his6-C8 was a strong AMPK substrate with a rate of phosphate incorporation similar to the model SAMS peptide substrate (Fig. 2B). Analysis of the phosphoamino acid content of 32P-labeled his6-C8 by acid hydrolysis and two-dimensional thin layer chromatography revealed that the phosphate was exclusively incorporated into serine residue(s) (Fig. 2C).

The above data led us to suppose that the C8 domain of cMyBP-C is a good in vitro AMPK substrate and phosphorylated on one or more serine residues. There are three serines within the domain (residues 1020, 1024 and 1040) and a series of alanine replacement mutants were generated using the parent pET28a construct to determine which were the phosphorylatable residues. Two single alanine replacement mutants (Ser1020Ala and Ser1040Ala), a double (Ser1020Ala/Ser1024Ala) and a mutant with all three serine residues replaced (Ser1020Ala/Ser1024Ala/Ser1040Ala) were engineered and the recombinant proteins overexpressed and purified as wild type. However, each alanine replacement mutant was phosphorylated to a similar level as wild type his6-C8 (data not shown). This surprising finding led us to examine the possibility that a phosphorylatable serine lies within the N-terminal amino acids encoded by the pET28a plasmid DNA; inspection of the sequence revealed that there are five serines in this short sequence, a pair either side of the hexahistidine tag and one at the C-terminal end of the six-residue thrombin recognition sequence (Fig. 1). To test this, phosphorylated his6-C8 was digested with thrombin (Fig. 3A and B) and also thrombin-cleaved his6-C8 phosphorylated with AMPK (Fig. 3C and D). In contrast with thrombin treatment of unphosphorylated his6-C8 (Fig. 3C), digestion of phosphorylated his6-C8 was found to be incomplete as full length fusion protein (estimated 21% of total by scanning densitometry) was present along with a doublet produced by protease action (Fig. 3A). Autoradiography revealed that in the thrombin-treated sample, only full length fusion protein, not the doublet, was phosphorylated (Fig. 3B). This suggested that the phosphorylation is not within the C8 domain itself; this was supported by the finding that the doublet produced by thrombin digestion prior to AMPK phosphorylation was a very poor AMPK substrate compared to intact his6-C8 (Fig. 3D). Taken together, these data strongly suggest that the principal serine within his6-C8 that was phosphorylated in this study is that present within the thrombin recognition site. Phosphorylation at this residue clearly inhibits thrombin digestion (Fig. 3A and B). As this serine is C-terminal to the site of thrombin cleavage, it is present as the second of the four amino acids fused to C8 after digestion to remove the hexahistidine sequence (Fig. 1); however, with only one flanking N-terminal residue, the serine is no longer a good substrate for AMPK (Fig. 3D).

Fig. 3

Thrombin digestion reveals that the phosphorylatable serine of his6-C8 lies in the leader peptide not within C8 itself. (A and B) Thrombin digestion of AMPK-phosphorylated his6-C8. A: Coomassie Blue-stained gel; B: autoradiograph. Lane 1: undigested AMPK-phosphorylated his6-C8 (control); lane 2: AMPK-phosphorylated his6-C8 after thrombin digestion. (C and D) AMPK phosphorylation subsequent to thrombin digestion of his6-C8 fusion protein. A: Coomassie Blue-stained gel; B: autoradiograph. Lane 1: undigested AMPK-phosphorylated his6-C8 (control); lane 2: AMPK-phosphorylated his6-C8 after thrombin digestion.

Conclusions

This study aimed to test whether the IgI-like C8 domain of cMyBP-C is an in vitro substrate of AMPK. We used the commonly employed approach of producing the protein of interest fused to a hexahistidine tag to facilitate purification. The fusion protein was readily purified using Ni2+–NTA affinity chromatography and was shown to be efficiently phosphorylated by AMPK in vitro. However, further study showed that the phosphorylatable serine was present within the thrombin recognition sequence encoded by the vector and not within C8 itself. The primary recognition sequence for AMPK is known to be very broad and has been proposed to be ϕ(X,basic)XXSer/ThrXXXϕ where ϕ is a hydrophobic residue (met, val, leu, ile or phe) [11-13]. Clearly there are many protein sequences matching these criteria that are not AMPK phosphorylation sites, suggesting that the nature of the local secondary structure may play an important role in recognition and that successful prediction of strong AMPK sites is difficult. This is illustrated by the divergence from the notional consensus sequence on known phosphorylation sites, for example ser1200 of rat acetyl-coA carboxylase [11]. The sequence surrounding the site identified here also does not fit with the proposed sequence.

The presence of the strong AMPK site within the vector-encoded residues proved a confounding factor in this study and we urge caution in the use of fusion proteins containing the thrombin recognition site in the work aimed at identifying AMPK substrates. As this sequence occurs in many commercial pET vector series (for example, pETs 14, 15, 41, 42, 43, 47, 48, 49, 50 and 52 as well as 28) we hope our observation will prove useful in other laboratories’ experimental design.