Identification of the Major Casein Kinase I Phosphorylation Sites on Erythrocyte Band 3

MAJOR CHANGES IN erythrocyte membrane protein phosphorylation are observed during invasion of malarial parasites,1-3 interaction of erythrocytes with Sendai virus,4 exposure of red blood cells to elevated levels of low-density but not high-density lipoproteins,5 and treatment of erythrocytes with various stimulants.6-8 The kinases thought to be responsible for these phosphorylation reactions include cAMP-dependent protein kinase,9 casein kinases I and II,10,11 protein kinase C,12 p72^syk protein tyrosine kinase,13 and a Ca²⁺/calmodulin-dependent protein kinase.9 A variety of phosphatases are also present in human erythrocytes, and these undoubtedly also participate in regulating levels of membrane protein phosphorylation.9,14,15 In some instances, the functional consequences of the phosphorylation reactions have been elucidated, and where this information is available, phosphorylation generally decreases the modified protein's affinity for other membrane proteins.9,16-20 Phosphorylation of β-spectrin by casein kinase I, for example, reduces the mechanical stability of the entire membrane.20 

Band 3, the most abundant polypeptide of the erythrocyte membrane, is phosphorylated by at least two prominent protein kinases. The protein tyrosine kinase, p72^syk, transfers phosphates primarily to Tyr 8, although other sites may be less efficiently phosphorylated.13,17,21 Modification of tyrosines 8 and 21 causes a structural change in band 3²² that reduces the affinity of several glycolytic enzymes for their inhibitory binding sites at the N-terminus of the polypeptide,17 thereby enhancing glycolytic rates.23 Casein kinase I also phosphorylates band 3,16,24 but neither the modified residues nor the functional consequences of this phosphorylation has been examined. Because casein kinase I appears to primarily regulate structural properties of the red blood cell membrane16,18,20 and because band 3 is significantly involved in control of membrane structure, primarily through its association with ankyrin, protein 4.1, and protein 4.2,17,25-28 we hypothesized that the aggressive phosphorylation of band 3 by casein kinase I might also exert an impact on erythrocyte membrane structure. To begin to test this hypothesis, we have undertaken to identify the residues on band 3 phosphorylated by casein kinase I. Because many regions of band 3 have already been identified with specific functions or peripheral protein interactions,17,27,29-31 localization of a phosphorylation site within one of these regions should facilitate further investigation of the functional impact of the phosphorylation. In this report, we use computer-based sequence analyses, polypeptide fragmentation studies, and direct amino acid sequencing to locate the major casein kinase I phosphorylation site on band 3.

Chemical cleavage reagents NH₂OH (hydroxylamine), 2-nitro-5-thiocyanobenzoic acid (NTCB), and formic acid were purchased from Sigma (St Louis, MO). Endoproteinase Asp-N (sequencing grade) and guanidine hydrochloride (ultrapure) were from Boehringer Mannheim Biochemicals (Indianapolis, IN). [δ-³²P] ATP was from Amersham Co (Arlington Heights, IL), and acetonitrile and trifluoroacetic acid were products of Baxter Health Care Co (Muskegon, MI). Amino acid sequencing reagents and reverse-phase high-performance liquid chromatography (HPLC) microbore columns were supplied by Applied Biosystems (Foster City, CA). Genetics Computation Group (GCG) software packages (version 7.2) were licensed from the GCG group (Madison, WI) by the Aids Center at Purdue University.

Phosphorylation of cytoplasmic domain of band 3 (CDB3) by casein kinase I and phosphoamino acid analysis.Cloned CDB3 (residues 1-379 of band 3) was expressed and purified according to previously described methods.26 The natural CDB3 fragment from red blood cells was isolated as described by Appell and Low.32 Purification of casein kinase I and phosphorylation of band 3 were conducted as reported by Lu et al.18 The casein kinase I preparation was judged to be pure and free from contaminating casein kinase II, as shown by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and immunoblotting with anti-casein kinase II antiserum.11 CDB3 was phosphorylated at 25°C for 30 minutes in a reaction mixture containing 50 mmol/L Tris-HCl, pH 7.5, 10 mmol/L MgCl₂ , 0.2 mmol/L [δ-³²P]ATP (specific activity, 10³ cpm/pmol), 2 μg purified casein kinase I, and 10 μg/mL each of the protease inhibitors leupeptin, pepstatin A, and aprotinin. The reaction was stopped by cooling rapidly to −20°C. Before subsequent analyses, unreacted ATP was removed from the protein solution by dialysis against phosphate-buffered saline, pH 7.5. CDB3 was subjected to acid hydrolysis and phosphoamino acid analysis as described elsewhere.17 

Chemical digestion and fragmentation of CDB3.Labeled CDB3 was denatured by dialyzing against 7.5 mol/L guanidine-HCl at 4°C overnight, and aliquots (15 μg) of denatured CDB3 were transferred to digestion tubes for further processing. For partial acid digestion, 88% formic acid (pH 2.5) was directly added to denatured CDB3 to achieve a final concentration of 75% and the mixture was incubated at 37°C for 24 hours.33 NTCB digestion of CDB3 was conducted by mixing an equal volume of NTCB stock (10 mmol/L NTCB dissolved in 1 mol/L Tris-Cl, pH 8.0) with denatured CDB3. The reaction was initiated by raising the reaction mixture pH to 9.0 with 1.5 mol/L Tris-Cl (pH 11), and the digestion was allowed to proceed in the dark for 30 hours at 37°C.27 NH₂OH digestion followed methods previously described,34 except that the reaction was conducted at pH 10.5 for 4 hours at 45°C. After each proteolysis procedure, the digested CDB3 was dialyzed against H₂O in a low molecular cutoff dialysis tube (molecular weight [Mr] cutoff = 1,000) at 4°C overnight. The dialyzed sample was lyophilized, dissolved in 50 μL SDS sample buffer, and subjected to Tricine SDS-PAGE (see below).

Proteinase digestion of CDB3 and sequencing of phosphorylated fragments.Endoproteinase Asp-N digestion of denatured CDB335 was conducted in 50 μL of 10 mmol/L Tris-Cl, pH 7.5. For this purpose, 5 μL of Asp-N (enzyme:protein ratio = 1:10) was added directly to the suspension and the reaction was allowed to proceed at 4°C overnight. The resulting mixture of ³²P-labeled CDB3 fragments was directly fractionated by HPLC using a 0% to 100% gradient of B in A (A, 0.1% trifluoroacetic acid in H₂O; B, 0.1% trifluoroacetic acid in acetonitrile) and a flow rate of 50 μL/min on a C18 reverse-phase microbore column (5 μm, 250 × 1 mm). Thirty sample tubes corresponding to visually resolvable peaks in the chromatogram were manually collected. The ³²P-labeled fractions were identified by Cerenkov counting, and amino acid sequencing was performed on labeled fragments in the Purdue University Laboratory for Macromolecule Structure using a gas phase amino acid sequencer 470A with an on-line 120A PTH analyzer.

Tricine SDS-PAGE and autoradiography of digested peptides.Tricine–SDS-PAGE was used instead of the more classical SDS-PAGE procedures to improve resolution of small peptides.36 After electrophoretic separation, gel slices were enclosed in Saran wrap (Dow Chemical, Indianapolis, IN) and subjected to autoradiography by directly exposing them to Kodak X-50 film (Eastman Kodak, Rochester, NY) at room temperature for 3.5 hours.

Biocomputation with GCG sequence analysis software.All searches and computations used GCG software packages. The amino acid sequence of CDB3 was translated from the cDNA sequence of the Lux group.37 Analysis of this sequence used the programs translate, motif, peptidemap, and peptidesort, as described in the Results. The detailed commands for each program are described in the manual for the GCG package and these can also be found using on-line help commands.38,39 

Analysis of phosphorylated amino acids.Quantitative analysis of ³²P incorporation into purified recombinant CDB3 was conducted by thin-layer chromatography after acid hydrolysis of casein kinase I phosphorylated CDB3. Approximately 1 mol of threonine phosphate was observed per mole of CDB3. Lower but readily detectable levels of serine phosphate were also measured. Thus, casein kinase I likely phosphorylates CDB3 at more than a single site.

Localization of phosphorylation sites to specific peptide fragments of CDB3.To more specifically define the residues of CDB3 phosphorylated by casein kinase I, the phosphorylated protein was subjected to several fragmentation schemes, after which the labeled fragments were separated by SDS-PAGE and analyzed for phosphorylation by autoradiography. As shown in Fig 1, intact phosphorylated CDB3 migrates as a single band at ∼43 kD in a Tricine-SDS polyacrylamide gel (lane A). Cleavage at cysteine residues with NTCB generates fragments of Mr ∼23 kD, 13 kD, and 7 kD, corresponding to residues 1-201, 202-317, and 318-379, respectively.25 Incomplete cleavage at residue 201 also yields a fragment of Mr ∼36 kD. Importantly, autoradiography of NTCB-digested CDB3 shows labeled peptides of Mr 43 kD, 36 kD, and 23 kD (lane B), all of which contain residues 1-201. Because neither the 13-kD nor the 7-kD fragment is strongly labeled, we conclude that most of the attached phosphate is in the amino terminal half of CDB3.

NH₂OH cleaves proteins specifically but inefficiently at Asn-Gly bonds.34 In CDB3, Asn-Gly peptide bonds occur at Asn 91-Gly 92 and Asn 365-Gly 366. Although the N-terminal fragment generated by NH₂OH proteolysis has an Mr of ∼11 kD, because of its highly acidic nature it migrates in SDS-PAGE at an apparent Mr of ∼15 kD.27 As noted in lane C, this peptide and the intact 43-kD CDB3 are heavily phosphorylated, whereas the ∼30-kD complementary fragment (residues 92-365) is only weakly labeled. These data thus further confine the major casein kinase I phosphorylation site to residues 1-91, but they also imply the existence of a minor phosphorylation site C-terminal to this peptide.

When CDB3 is subjected to mild formic acid hydrolysis (pH 2.5), the sensitive peptide bonds between aspartic acid and proline are specifically severed.33 In CDB3, there are three such sensitive sites, ie, Asp 23-Pro 24, Asp 183-Pro 184, and Asp 370-Pro 371, generating nonresolvably small peptides of Mr ∼2.5 kD (residues 1-23) and ∼1 kD (residues 371-379), as well as major fragments of Mr ∼18 kD (residues 24-183) and ∼20 kD (residues 184-370). Figure 1, lane D shows that peptides of Mr ∼18 kD, ∼21 kD, and ∼43 kD are prominently phosphorylated. The 18-kD fragment is readily assignable to residues 24-183, whereas the 43-kD polypeptide obviously corresponds to intact CDB3. However, the 21-kD fragment could be either residues 184-370 or 1-183, the latter arising from incomplete hydrolysis of the Asp 23-Pro 24 bond. Based on the NH₂OH fragmentation study locating the majority of radioactivity between residues 1 and 91, we suggest that the 21-kD peptide probably derives from residues 1-183. Finally, by considering all radioactive fragments identified in lanes B, C, and D, the most efficiently phosphorylated residues in CDB3 must reside between amino acids 24 and 91. However, a second, less avidly labeled site may reside between amino acids 92 and 379.

Computer-assisted refinement of the location of the casein kinase I phosphorylation sites in CDB3 and direct peptide sequencing.Examination of the CDB3 sequence between residues 24 and 91 shows 6 threonine and 2 serine residues, all but one of which lie between positions 39 and 54. Because peptide sequencing from the closest cleavage site identified above (ie, Pro 24) was unsuccessful, a more precise localization of the phosphorylation site was required before further sequencing could be initiated. For this purpose, the computer program motif was used to search the CDB3 sequence for consensus casein kinase I phosphorylation sites. Because no consensus sequence had yet been entered into the motif program for casein kinase I, we consulted several studies aimed at defining the kinase's substrate preference10,40,41 and constructed our own search strategy. More specifically, because casein kinase I has been reported to prefer Ser/Thr residues separated by one or two residues from a more N-terminal acidic amino acid,10,40,41 we searched for the motif D/E-(X)_{1 or 2} -S/T. Although 11 candidate sites were identified in intact CDB3, only 2 sites (Thr 42 and Thr 48) were found between residues 24 and 91.

To distinguish which of the two computer-identified sites was actually recognized by casein kinase I, it was desirable to cleave between the candidate threonine residues and evaluate which fragment retained the radioactive phosphate. For this purpose, the GCG program peptidemap was used to screen the CDB3 sequence for proteases that might cleave between the aforementioned residues. Asp-N endoprotease was chosen because of its availability and selectivity for digestion on the NH₂ -terminal side of aspartate residues,35 ie, in our case Asp 45. To develop a separation protocol for isolation of the derived proteolysis products, the hypothetical peptides were then examined by the program peptidesort, where each peptide's probable retention time on various reverse-phase HPLC columns was estimated. As seen in Table 1, the highly hydrophilic peptide 7 containing residues 38-44 was projected to elute near the solvent front at pH 2.1, whereas the very hydrophobic peptide 8 comprising residues 45-66 was expected to exhibit a retention time of ∼40.6 minutes. Importantly, the experimental HPLC profile of Asp-N endoprotease-digested CDB3 roughly confirmed the computational predictions. Most of the radioactivity was isolated in peptide A near the solvent front, with a much lower level of phosphate in a nearby peptide B (Fig 2). Both peptides were submitted for amino acid sequence analysis, and peptide A was found to contain the sequence Asp-Thr-Glu-Ala-Thr-Ala-Thr corresponding to residues 38-44, as indicated in peptide 7. Also, as predicted, Thr 42 was observed to be the most intensely labeled amino acid, although, due to the width of the elution peak, the possible phosphorylation of Thr 44 could not be excluded. Peptide B, in contrast, yielded the sequence Asp-Ala-Tyr-Met-Ala-Glu-Ser-Arg-Gly-Glu-Leu-Leu and serine 303 was identified as the phosphorylated amino acid. Curiously, the sequence of peptide B does not conform to the consensus sequence of a casein kinase I phosphorylation site, but rather matches the description of a casein kinase II phosphorylation site. Although immunoblotting analysis of the purity of our casein kinase I preparation showed no evidence of casein kinase II, it is nevertheless not possible to exclude that a minor contaminant might be responsible for this phosphorylation.

We have used peptide mapping, a combination of computer methods, and ultimately amino acid sequencing to localize the major casein kinase I phosphorylation site on the cytoplasmic domain of band 3. As noted above, the predominant site was identified as Thr 42, a residue within a classical casein kinase I consensus sequence. In contrast, the minor phosphorylation site was determined to be Ser 303, a residue surrounded by a sequence more characteristic of casein kinase II phosphorylation sites. However, because casein kinase I is clearly promiscuous in selecting phosphorylation sites10,40,41 and because casein kinase II does not even phosphorylate band 3,11 it is not unreasonable to assume that both Ser 303 and Thr 42 were phosphorylated by the same casein kinase I.

Analysis of the amino acids surrounding Thr 42 shows a sequence that could hypothetically participate in a hierarchical phosphorylation mechanism.40,41 Thus, casein kinase I prefers to phosphorylate a Ser/Thr one or two residues towards the C-terminus from an acidic amino acid. Phosphorylation of Thr 42 in the sequence, 40-EATATDYHT-48, could conceivably create a new substrate site in Thr 44. Furthermore, the highly acidic stretch generated by phosphorylation of Thr 42 and Thr 44 together with Asp 45 would generate a potential phosphorylation site for p72^syk at Tyr 46. Indeed, Tyr 46 has been shown to be phosphorylated in situ.22 Finally, Thr 48 or 49 could then respond to this catenation of phosphorylation reactions and become a third substrate site for casein kinase I. Although this hierarchical phosphorylation scheme is admittedly highly speculative, it should be noted that multiple phosphorylations in this region have already been reported.22 

Because analysis of the casein kinase I sites on CDB3 was motivated by an interest in identifying functional consequences of phosphorylation, the question naturally arises as to whether the newly acquired data provide any clues regarding the physiologic sequelae of casein kinase I phosphorylation. Whereas Ser 303 is well removed from any known center of CDB3 function, Thr 42 is clearly proximal to several important peripheral protein interaction sites. Not only do ankyrin, protein 4.1, and the glycolytic enzymes associate at the N-terminus of band 3,17,27,29-31 but protein 4.2 may dock near this region also. Thus, mutation of Glu 40 of CDB3 to a Lys residue has been linked to a deficiency in protein 4.2 and an accompanying hereditary spherocytosis.42 Although the above-noted mutation could directly alter the association between protein 4.2 and CDB3, it is also conceivable that the mutation could indirectly affect the interaction through its influence on casein kinase I phosphorylation, ie, mutation of Glu 40 should eliminate Thr 42 as a casein kinase I site. Further studies will obviously be required to confirm or refute these speculations.