Protein Engineering of Dual-Cys Cyanobacteriochrome AM1_1186g2 for Biliverdin Incorporation and Far-Red/Blue Reversible Photoconversion

Cyanobacteria have cyanobacteriochromes (CBCRs), which are photoreceptors that bind to a linear tetrapyrrole chromophore and sense UV-to-visible light. A recent study revealed that the dual-Cys CBCR AM1_1186g2 covalently attaches to phycocyanobilin and exhibits unique photoconversion between a Pr form (red-absorbing dark state, λmax = 641 nm) and Pb form (blue-absorbing photoproduct, λmax = 416 nm). This wavelength separation is larger than those of the other CBCRs, which is advantageous for optical tools. Nowadays, bioimaging and optogenetics technologies are powerful tools for biological research. In particular, the utilization of far-red and near-infrared light sources is required for noninvasive applications to mammals because of their high potential to penetrate into deep tissues. Biliverdin (BV) is an intrinsic chromophore and absorbs the longest wavelength among natural linear tetrapyrrole chromophores. Although the BV-binding photoreceptors are promising platforms for developing optical tools, AM1_1186g2 cannot efficiently attach BV. Herein, by rationally introducing several replacements, we developed a BV-binding AM1_1186g2 variant, KCAP_QV, that exhibited reversible photoconversion between a Pfr form (far-red-absorbing dark state, λmax = 691 nm) and Pb form (λmax = 398 nm). This wavelength separation reached 293 nm, which is the largest among the known phytochrome and CBCR photoreceptors. In conclusion, the KCAP_QV molecule developed in this study can offer an alternative platform for the development of unique optical tools.


Introduction
Cyanobacteria, prokaryotes that perform oxygenic photosynthesis, have various photoreceptors to detect and acclimate to ambient light conditions. In cyanobacteria, linear tetrapyrrole-binding BV-binding photoreceptors are advantageous for optogenetic control in mammals because BV is an endogenous molecule in mammals and absorbs far-red light, which penetrates into the deep tissues of mammals. The large separation of the two absorbing forms of AM1_1186g2 enables us to achieve full photoconversion, which is favorable for strict optogenetic control. In the present study, we obtained an AM1_1186g2 variant called AM1_1186g2_KCAP-QV that efficiently incorporated BV and showed far-red/blue reversible photoconversion ( Figure 1B).

RCAP: Efficient BV Incorporation
The AM1_1186g2 wild-type (WT) covalently binds PCB and exhibits unique photoconversion between a Pr dark state and a Pb photoproduct ( Figure 1A) [24]. Irradiation of the Pr form with red light results in photoisomerization of the C15-16 double bond of PCB, and transient formation of an Io form, which is intermediate immediately after the photoisomerization. Thereafter, the second Cys residue (Cys 2510 ) slowly forms a thioether linkage to the C10 of PCB to generate the blue-shifted Pb form. Although the WT protein accumulated the Io form during Pr-to-Pb photoconversion, the PR protein in which the Pro 2509 was replaced with Arg showed quick Pr-to-Pb photoconversion without noticeable accumulation of the Io form [24]. Since the Pro 2509 is positioned just before the second Cys residue ( Figure 1C), this replacement may largely affect the positioning of the second Cys residue, enabling rapid covalent bond formation.
To evaluate the BV-binding capability of the WT protein, we previously expressed the WT protein in the BV-producing Escherichia coli and found that the WT protein could slightly bind BV and absorb light in the far-red region (Figure 2A) [38]. Notably, irradiation with far-red light resulted in the bleaching of the far-red light absorption, but not an increase in the blue light region, suggesting no covalent bond formation between the second Cys residue and the C10 of the chromophore, in contrast to the PCB-binding one (Figure 2A). The BV-binding pocket may be arranged in comparison with the case for PCB incorporation, which may result in the inefficient BV incorporation and inaccessibility of the second Cys residue to the C10. Since the Pro 2509 Arg replacement improved the accessibility of the second Cys residue to the C10 of the PCB, we expected similar improvement for the BV incorporation. The PR replacement, however, resulted in almost no BV binding ( Figure 2B). protein in which the Pro2509 was replaced with Arg showed quick Pr-to-Pb photoconversion without noticeable accumulation of the Io form [24]. Since the Pro2509 is positioned just before the second Cys residue ( Figure 1C), this replacement may largely affect the positioning of the second Cys residue, enabling rapid covalent bond formation.
To evaluate the BV-binding capability of the WT protein, we previously expressed the WT protein in the BV-producing Escherichia coli and found that the WT protein could slightly bind BV and absorb light in the far-red region (Figure 2A) [38]. Notably, irradiation with far-red light resulted in the bleaching of the far-red light absorption, but not an increase in the blue light region, suggesting no covalent bond formation between the second Cys residue and the C10 of the chromophore, in contrast to the PCB-binding one (Figure 2A). The BV-binding pocket may be arranged in comparison with the case for PCB incorporation, which may result in the inefficient BV incorporation and inaccessibility of the second Cys residue to the C10. Since the Pro2509Arg replacement improved the accessibility of the second Cys residue to the C10 of the PCB, we expected similar improvement for the BV incorporation. The PR replacement, however, resulted in almost no BV binding ( Figure 2B). Therefore, we considered that the PR replacement alone is not enough for the BV incorporation and covalent bond formation. In this context, we next focused on two residues just after the second Cys residue: Val2511 and Phe2512. Since covalent bond formation of the second Cys residue, Cys2510, with the C10 may cause steric hindrance between the next Val2511 residue and the ring C carbonyl, Ala possessing a smaller side chain may be favorable for BV incorporation. Further, for structural modification around the second Cys residue, we replaced Phe2512 with Pro, which has a unique ring structure, and thus is likely to promote structural modification. The obtained mutant named RCAP possessing three substitutions (Pro2509Arg, Val2511Ala, and Phe2512Pro) showed moderately high BVbinding capability ( Figure 2C). The BV-binding RCAP protein exhibited reversible photoconversion between a far-red-absorbing form (Pfr, λmax = 691 nm) and a blue-light-absorbing form (Pb, λmax = 398 Therefore, we considered that the PR replacement alone is not enough for the BV incorporation and covalent bond formation. In this context, we next focused on two residues just after the second Cys residue: Val 2511 and Phe 2512 . Since covalent bond formation of the second Cys residue, Cys 2510 , with the C10 may cause steric hindrance between the next Val 2511 residue and the ring C carbonyl, Ala possessing a smaller side chain may be favorable for BV incorporation. Further, for structural modification around the second Cys residue, we replaced Phe 2512 with Pro, which has a unique ring structure, and thus is likely to promote structural modification. The obtained mutant named RCAP possessing three substitutions (Pro 2509 Arg, Val 2511 Ala, and Phe 2512 Pro) showed moderately high BV-binding capability ( Figure 2C). The BV-binding RCAP protein exhibited reversible photoconversion between a far-red-absorbing form (Pfr, λ max = 691 nm) and a blue-light-absorbing form (Pb, λ max = 398 nm; Figure 2C). The Pfr form is 50 nm red-shifted compared with the Pr form binding PCB, which is consistent with the absorption of BV and PCB themselves. Conversely, the Pb form of the BV-binding RCAP was blue-shifted compared with that of the PCB-binding one. This blue shift might be derived from overlap between the Soret band and blue absorption. Next, to address the purity and BV-binding efficiency of the AM1_1186g2 variant proteins, these proteins were subjected to SDS-PAGE ( Figure 2D). There were two main protein bands around 20 kDa based on the Coomassie Brilliant Blue (CBB) staining. The protein bands just below 21.5 kDa (red arrowhead) emitted fluorescence in the presence of zinc ion, indicating that these protein bands corresponded to the BV-binding holoproteins. To confirm that the bands below the holoprotein bands (black arrowhead) corresponded to the apoproteins, we purified the WT proteins from both non-BV-producing and BV-producing E. coli for His-tag immunodetection. The bands around 20 kDa did not react with the His-tag antibody, and so were concluded not to be derived from the apoproteins but from the contaminant nonspecific proteins ( Figure 2E). Since the holoprotein bands of the RCAP protein were not single but smeared in comparison with the WT and PR proteins ( Figure 2D), the RCAP holoprotein should be unstable and easily degraded.

KCAP: Improvement in Protein Stability
Judging from the absorption spectra and zinc-induced fluorescence of the WT and PR proteins ( Figure 2A,B,D), Pro 2509 Arg replacement based on the WT background had an inhibitory effect on the BV-binding efficiency and/or purity. In this context, Arg 2509 of the RCAP protein may have a similar inhibitory effect. To verify this possibility, we focused on this residue for further improvement. We replaced the Arg residue with the original Pro or Lys residue, which has a positive charge like Arg, producing CAP (Val 2511 Ala, and Phe 2512 Pro) and KCAP (Pro 2509 Lys, Val 2511 Ala, and Phe 2512 Pro) variant proteins. CAP and KCAP variant proteins showed reversible far-red/blue photoconversion, like the RCAP protein ( Figure 3A,B). The specific absorbance ratio [SAR, (peak red-band absorption)/(peak 280 nm absorption)] value of the CAP protein was lower than that of the RCAP protein ( Table 1), suggesting that the Pro 2509 Arg replacement based on the CAP background did not have inhibitory effects on the BV-binding efficiency and/or purity, in contrast to the case for the WT protein. Conversely, the SAR value of the KCAP was higher than that of the RCAP protein (Table 1). Furthermore, judging from the CBB-stained and zinc-induced fluorescent gels ( Figure 3C), the purified CAP and KCAP proteins were found to be present not as smears but as single bands, indicating that these variant holoproteins are stable in comparison with the RCAP protein. In conclusion, the KCAP protein is the best variant showing high BV-binding efficiency and holoprotein stability at this time.

KCAP_QV: Improvement in Holoprotein Expression
As the next approach for improving BV-binding capability, we compared the primary sequence of the KCAP protein with those of other BV-binding CBCRs ( Figure 1C). We previously found and developed seven BV-binding CBCRs [36][37][38]. We hypothesized that some amino acid residues conserved in these CBCRs but not in the KCAP protein might be crucial for BV-binding capability. Based on multiple sequence alignment and structural information, we identified seven amino acid residues (Phe2501, Val2503, Asn2504, Asn2508, Tyr2516, Gln2517, and Leu2526) near the chromophore that are unique to the KCAP protein. Among these residues, Phe2501, Tyr2516, and Leu2526 correspond to the BV4 residues. Since the introduction of the BV4 residues into the WT protein failed to achieve any improvement in BV-binding capability [38], we excluded these three residues from further targeting. Thus, we replaced the other four residues with residues conserved among the other BV-binding CBCRs one by one: KCAP_VQ (Val2503Gln, Pro2509Lys, Val2511Ala, and Phe2512Pro), KCAP_NE

KCAP_QV: Improvement in Holoprotein Expression
As the next approach for improving BV-binding capability, we compared the primary sequence of the KCAP protein with those of other BV-binding CBCRs ( Figure 1C). We previously found and developed seven BV-binding CBCRs [36][37][38]. We hypothesized that some amino acid residues conserved in these CBCRs but not in the KCAP protein might be crucial for BV-binding capability. Based on multiple sequence alignment and structural information, we identified seven amino acid residues (Phe 2501 , Val 2503 , Asn 2504 , Asn 2508 , Tyr 2516 , Gln 2517 , and Leu 2526 ) near the chromophore that are unique to the KCAP protein. Among these residues, Phe 2501 , Tyr 2516 , and Leu 2526 correspond to the BV4 residues. Since the introduction of the BV4 residues into the WT protein failed to achieve any improvement in BV-binding capability [38], we excluded these three residues from further targeting. Thus, we replaced the other four residues with residues conserved among the other BV  Figure 4A), whereas the other three variant proteins had lower ones ( Figure S1). In particular, Asn 2504 Glu replacement (KCAP_NE) resulted in the loss of photoconversion capability and protein aggregation induced by far-red light illumination ( Figure S1B). Judging from the SAR values, the KCAP_QV protein is the best variant for BV incorporation among the variants tested in this study.
To clarify the details of the improvement by Gln 2517 Val replacement, the KCAP and KCAP_QV proteins were subjected to SDS-PAGE. The main contaminant protein band that appeared for the KCAP preparation was almost undetectable for the KCAP_QV preparation ( Figure 4B), indicating that the higher SAR value is mainly derived from higher yield of the holoprotein expression and resultant improvement of purity. To verify this assumption, we compared the E. coli cell pellets expressing the KCAP or KCAP_QV protein. As a result, the cell pellet expressing the KCAP_QV protein showed a deep green color, whereas that expressing the KCAP protein showed a pale green color ( Figure 4C). From these results, we concluded that Gln 2517 Val mutation markedly improved the expression yield of the BV-binding holoproteins. Since high purity of the KCAP_QV protein enables precise determination of the BV-binding efficiency, we calculated the BV-binding efficiency as 67% by the method described in a previous study [38]. Since this value is comparable to those of the other BV-binding molecules developed in the previous study [38], we stopped our engineering efforts and regarded the KCAP_QV protein as the final developmental product in this study.
To characterize the photochemical properties of the KCAP_QV protein, acid-denatured spectra were measured. Its Pfr form denatured by guanidinium chloride under acidic conditions exhibited an absorption peak at 708 nm, and white light illumination did not result in any spectral change ( Figure  S2A). This result showed that the Pfr form attached to the 15Z-isomer of BV. On the other hand, that of the Pb form showed an absorption peak at 637 nm, and white light illumination resulted in a red shift peaking at 708 nm ( Figure S2B). This showed that the Pb form bound the 15E-isomer of BV.
Furthermore, we measured the photoconversion kinetics of both direction and reversibility with repetitive conversions. The absorbance change at 691 nm was monitored at room temperature during repetitive far-red and blue light illuminations ( Figure S3). The Pfr-to-Pb photoconversion kinetics was extremely slow in comparison with the opposite Pb-to-Pfr kinetics. Further, the Pfr-to-Pb kinetics was found to be very slow when compared also with the Pfr-to-Po kinetics of AnPixJg2-BV4 ( Figure  S4). This slow Pfr-to-Pb photoconversion kinetics was conserved among the BV-binding AM1_1186g2 variant proteins produced in this study (data not shown). Unlike the PCB-binding AM1_1186g2 wild-type protein, we could not detect any intermediate form during the Pfr-to-Pb conversion (data not shown), indicating that the quantum yield of the Z/E isomerization may be quite low. Repetitive photoconversion resulted in a decrease in the Pfr peak absorbance ( Figure S3), indicating that prolonged irradiation with far-red light causes bleaching of the KCAP_QV protein. Although we incubated the Pb photoproduct under dark conditions at room temperature for 9 h, we could not detect obvious dark reversion to the Pfr dark state (data not shown). To clarify the details of the improvement by Gln2517Val replacement, the KCAP and KCAP_QV proteins were subjected to SDS-PAGE. The main contaminant protein band that appeared for the KCAP preparation was almost undetectable for the KCAP_QV preparation ( Figure 4B), indicating that the higher SAR value is mainly derived from higher yield of the holoprotein expression and resultant improvement of purity. To verify this assumption, we compared the E. coli cell pellets expressing the KCAP or KCAP_QV protein. As a result, the cell pellet expressing the KCAP_QV protein showed a deep green color, whereas that expressing the KCAP protein showed a pale green color ( Figure 4C). From these results, we concluded that Gln2517Val mutation markedly improved the expression yield of the BV-binding holoproteins. Since high purity of the KCAP_QV protein enables precise determination of the BV-binding efficiency, we calculated the BV-binding efficiency as 67% by the method described in a previous study [38]. Since this value is comparable to those of the other BV-binding molecules developed in the previous study [38], we stopped our engineering efforts and regarded the KCAP_QV protein as the final developmental product in this study.
To characterize the photochemical properties of the KCAP_QV protein, acid-denatured spectra were measured. Its Pfr form denatured by guanidinium chloride under acidic conditions exhibited an absorption peak at 708 nm, and white light illumination did not result in any spectral change To elucidate oligomeric states of the both forms of the KCAP_QV protein, we performed HPLC analysis using a gel filtration column. The peaks of all preparations were detected around 3.5 min of the retention time. The retention times of the both forms became a little larger as higher concentration ( Figure 4D,E). Molecular sizes relative to the theoretical monomeric form (22.4 kDa) was ranging from 1.1 to 1.3 (24-29.5 kDa) for the Pfr form and from 1.0 to 1.2 (21.5-26.5 kDa) for the Pb form. Although these data might mean that only a small population of the proteins formed a dimer at higher concentration, the majority of the proteins in both forms were present as a monomer. Slight differences in retention times of the two forms even at the same concentration might reflect some structural change depending on the photoconversion.

Discussion
In this work, we engineered dual-Cys CBCR AM1_1186g2 to accept the mammalian intrinsic chromophore biliverdin by replacing several amino acids. The engineered molecule, KCAP_QV, efficiently incorporated the BV with high holoprotein stability and expression yield. The KCAP_QV showed reversible photoconversion between the Pfr dark state and the Pb photoproduct. The wavelength separation of the two forms was 293 nm, which is the largest among the known phytochromes and CBCRs.
We previously succeeded in identifying four residues (BV4) crucial for BV incorporation of some of the XRG CBCRs [38]. Although AM1_1186g2 also belongs to the XRG lineage, the same BV4 replacement on AM1_1186g2 failed to improve BV incorporation [38]. Since AM1_1186g2 showed reversible covalent bond formation between the second Cys residue and the C10 of the chromophore, which is a unique feature of AM1_1186g2 among the XRG CBCRs, the structural arrangement near the chromophore should be quite different from those of the canonical XRG CBCRs, probably leading to the failure.
In the case of the BV4 variant proteins, the canonical Cys residue stably ligates to the C3 2 of the BV chromophore, whereas the same residue ligates to the C3 1 of the PCB chromophore [28,38]. This unique binding mode resulted in deeper BV insertion into the protein pocket. Although we could not yet reveal the structures of the PCB-and BV-binding AM1_1186g2 variant proteins, a similar situation may be expected in the case of AM1_1186g2. In this context, the mutated residues in this study may contribute to escaping the steric hindrance derived from the deeper BV insertion into the protein pocket and thus improving the BV-binding efficiency. The reversible covalent bonding site, C10, is present between the rings B and C possessing bulky propionate side chains. The replacement of residues near the second Cys residue (Pro 2509 Lys, Val 2511 Ala, and Phe 2512 Pro) may be crucial for escaping the steric hindrance with these propionate side chains. Notably, Pro 2509 Lys replacement resulted in an improvement of holoprotein stability ( Figure 3C), although replacement with a similar Arg residue conversely led to holoprotein instability. At present, it is difficult to provide a plausible interpretation of this without some structural information.
We further identified that the Gln 2517 Val replacement based on the KCAP background made a major contribution to holoprotein expression yield, resulting in the highest purity and SAR value among the variant proteins analyzed in this study (Figure 4). Gln 2517 Val position corresponds to the residue just after one of the BV4 positions, Phe 308 Thr, based on multiple alignment and structural information. Thus, this replacement may modify the binding pocket near ring C to facilitate holoprotein expression.
Thus, the KCAP_QV variant protein is the best variant based on the AM1_1186g2 protein. Among the known BV-binding CBCR proteins, the KCAP_QV protein possesses unique features not only to show extremely large wavelength separation but also to show very slow photoconversion from the dark state to the photoproduct and not to show dark reversion from the photoproduct to the dark state. The PCB-binding WT protein showed slow photoconversion from the dark state to the photoproduct to accumulate the intermediate Io form. In contrast, the BV-binding KCAP_QV protein showed slow photoconversion without noticeable accumulation of any intermediate, suggesting low quantum yield of the light-induced Z/E isomerization. Although it is often the case that low quantum yield of the isomerization may result in bright fluorescence, fluorescence quantum yield of the Pfr form was measured to be about 0.7%, as low as those of the other photoconvertible BV-binding CBCRs [37]. Additionally, prolonged irradiation with far-red light resulted in photobleaching of the Pfr form. In this context, improvement in the Pfr-to-Pb photoconversion kinetics is needed for developing stable reversible optical tools. Nonetheless, the BV-binding KCAP_QV protein with unique features should be an alternative platform for the development of BV-based optical tools such as photoacoustic imaging [39] and photochromic FRET [40].

Plasmid Construction
The Escherichia coli strain JM109 or Mach1 was used for cloning plasmid DNA. Plasmids expressing His-tagged AM1_1186g2 wild-type and P 2509 R mutant were constructed in a previous study [24]. To perform site-directed mutagenesis, PrimeSTAR MAX mutagenesis kit reagents (TaKaRa, Shiga, Japan) were used with appropriate primers (Table S1) and all AM1_1186g2 variant constructs were obtained. The sequences of all constructs were confirmed by DNA sequencing.

Expression and Purification of His-Tagged AM1_1186g2 Variants
E. coli strain C41 (Novagen, Madison, WI, USA) harboring the pKT270 plasmid to produce BV was used for the expression of BV-binding AM1_1186g2 variants. For the expression of AM1_1186g2 apoprotein, we used the C41 strain without the pKT270 plasmid. As a pre-culture, one colony was picked up and incubated at 37 • C overnight in 10 mL of LB medium with 20 µg/mL kanamycin and/or 20 µg/mL chloramphenicol. Each culture was added into 1 L of LB medium with 20 µg/mL kanamycin and/or 20 µg/mL chloramphenicol and incubated until OD 600 was between 0.4 and 0.8. Subsequently, isopropyl thio-β-d-galactopyranoside was added to a final concentration of 0.1 mM and cultured at 18 ºC overnight for the induction of protein expression. Next, each sample was harvested by centrifugation at 4 • C and frozen at −80 • C for 30 min. Each frozen cell pellet was thawed at 4 • C and suspended in 40 mL of A buffer (20 mM HEPES-NaOH (pH 7.5), 100 mM NaCl, 10% (w/v) glycerol, and 0.5 mM Tris-carboxyethylphosphine (Pierce)) and disrupted by three passages through an Emulsiflex C5 high-pressure homogenizer at 12,000 psi (Avestin Inc., Ottawa, ON, Canada, Canada). The solution was centrifuged at 165,000 g for 30 min at 4 • C, and the supernatant containing His-tagged protein was filtered by 0.8 µm Cellulose Acetate Membranes. The filtrate was passed through a nickel affinity His-trap chelating column (GE Healthcare, Piscataway, NJ, USA). After washing with the A buffer containing 30, 50, and 100 mM imidazole, the His-tagged protein was eluted using a linear imidazole gradient from 100 to 400 mM. Then, the protein solution was incubated for 1 h with 1 mM EDTA. After that, to remove imidazole and EDTA, the protein solution was dialyzed against the A buffer including 1 mM dithiothreitol.
For the zinc-dependent fluorescence assay, the gels were immersed in distilled water for 30 min, followed by immersion with 20 µM zinc acetate at room temperature for 30 min. Thereafter, zinc-dependent fluorescence was detected using WSE-6100 LuminoGraph (ATTO, Tokyo, JAPAN) through a 600-nm long path filter upon excitation with blue (λ max = 470 nm) and green light (λ max = 527 nm) through a 562-nm short path filter.

Spectrometry and Measurement of Photoconversion Kinetics
Absorption spectra of all AM1_1186g2 variants were recorded using a model UV-2600 spectrophotometer (Shimadzu, Kyoto, Japan) at room temperature. Far-red and blue monochromatic light was provided using Opt-Spectrum Generator (OSG; Hamamatsu Photonics, Hamamatsu, Japan).
For the acid denaturation of KCAP_QV, 200 µL was diluted into 1 mL of 7 M guanidinium chloride/1% HCl (v/v), and absorption spectra were recorded before and after 1 min of illumination with white light.
Pfr-to-Pb or Pfr-to-Po photoconversion kinetics of KCAP_QV and AnPixJg2-BV4 were measured by monitoring at 691 nm. Before the measurement, KCAP_QV and AnPixJg2-BV4 were irradiated with light at 430 and 593 nm, respectively, to form a 15Z dark state. Ten seconds after the start of measurement, KCAP_QV and AnPixJg2-BV4 were irradiated with far-red light at 691 nm while measuring absorption. To measure reversible photoconversion kinetics of KCAP_QV, the absorbance at 691 nm was monitored during far-red and blue light illuminations. Before the measurement, KCAP_QV was irradiated with blue light at 398 nm to form a 15Z dark state. Ten seconds after starting the measurement, KCAP_QV was irradiated with far-red light at 691 nm while measuring absorption. After the Pfr-to-Pb photoconversion, KCAP_QV was irradiated with the blue light at 398 nm for 3 min and then irradiated with far-red light again. This light illumination cycle was repeated three times.
To calculate the molecular sizes of the native protein of each form, gel filtration chromatography was performed using a Prominence HPLC system (Shimadzu, Kyoto, Japan). Each sample was injected in a volume of 10 µL and eluted by using a HPLC column (Inertsil WP300 Diol, 4.6 i.d. × 250 mm, 5 µm; GL Sciences) at 28 • C with a buffer (50 mM phosphate (pH 7.0), 300 mM NaCl), in which these chromatograms were recorded at 280 nm. The molecular sizes were calculated by standard curve constructed from retention times of marker proteins (cytochrome C, 12.4 kD; ovalbumin, 44.3 kD; alcohol dehydrase, 150 kD; β-amylase, 200 kD; blue dextran, 2000 kD).