The polypeptide biophysics of proline/alanine‐rich sequences (PAS): Recombinant biopolymers with PEG‐like properties

Abstract PAS polypeptides comprise long repetitive sequences of the small L‐amino acids proline, alanine and/or serine that were developed to expand the hydrodynamic volume of conjugated pharmaceuticals and prolong their plasma half‐life by retarding kidney filtration. Here, we have characterized the polymer properties both of the free polypeptides and in fusion with the biopharmaceutical IL‐1Ra. Data from size exclusion chromatography, dynamic light scattering, circular dichroism spectroscopy and quantification of hydrodynamic and polar properties demonstrate that the biosynthetic PAS polypeptides exhibit random coil behavior in aqueous solution astonishingly similar to the chemical polymer poly‐ethylene glycol (PEG). The solvent‐exposed PAS peptide groups, in the absence of secondary structure, account for strong hydrophilicity, with negligible contribution by the Ser side chains. Notably, PAS polypeptides exceed PEG of comparable molecular mass in hydrophilicity and hydrodynamic volume while exhibiting lower viscosity. Their uniform monodisperse composition as genetically encoded polymers and their biological nature, offering biodegradability, render PAS polypeptides a promising PEG mimetic for biopharmaceutical applications.

The TrxA-PAS fusion proteins were produced in a soluble state in the cytoplasm of E. coli KS272. [8] Bacteria were grown at 25 °C in an 8 L bench top fermenter in M9 glucose minimal medium using a previously published protocol [9] and induced at OD 550 ≈ 30 for 4 h with 0.5 mM isopropyl-β-D-1-thiogalactopyranoside (IPTG; Formedium, Hanstanton, UK). After harvest by centrifugation, the bacterial pellet was resuspended in about 1 L 500 mM NaCl, 40 mM NaP i pH 7.5, homogenized in a PANDA Plus homogenizer (GEA Niro Soavi, Biberbach, Germany) and then heated to 70 °C for 15 min in a water bath to precipitate the majority of host cell proteins. After cooling to ~20 °C in a cold-room, the supernatant was clarified by centrifugation and stored frozen at -20 °C.
From the thawed solution, aggregated protein was removed by centrifugation and the supernatant was treated with 0.18 volumes (corresponding to 15 % saturation) 4.1 M (NH 4 ) 2 SO 4 (AppliChem, Darmstadt, Germany) on a magnetic stirrer at 4 °C to achieve fractionated protein precipitation. Initially precipitated host proteins were separated by centrifugation and, after adding another 0.13 volumes of the (NH 4 ) 2 SO 4 solution (corresponding to 25 % saturation), the PAS fusion protein was selectively precipitated and recovered by centrifugation whereas other contaminants remained in the supernatant. This protein pellet was resolubilized in 20 mM Tris/HCl pH 9.5 and dialyzed over night against the same buffer, prior to application to a 30 mL Resource 15 Q anion exchange column (GE Healthcare). Elution of pure protein was finally performed with a NaCl concentration gradient from 0 to 200 mM in the same buffer. Alternatively, cleavage of the corresponding Ser-free P/A-polypeptides from the TrxA fusion partner was achieved by treatment with BrCN. [10] To this end, about 100 mg of the fusion protein was dialyzed three times against 50 mM NH 4 HCO 3 pH 8.5 and lyophilized in a Speed Vac centrifuge (Christ, Osterode am Harz, Germany). The dried protein was dissolved at 5 mg/mL in 70 % (v/v) formic acid (Roth, Karlsruhe, Germany). Then, a 100fold molar amount of BrCN (5 M solution in DMF; Fluka, Steinheim, Germany) was added and the cleavage reaction was allowed to proceed for 1-2 days. The solution was again lyophilized and the resulting solid was resolubilized in 20 mM Tris/HCl pH 9.5 under ultrasonication for 15 min.

Preparation of PAS polymers by proteolysis or chemical cleavage
The raw product of each cleavage reaction was dialyzed against 20 mM Tris/HCl pH 9.5 overnight and, after sterile filtration, applied to AEX as described above. This time the uncharged PAS biopolymer appeared in the flow through and was collected, whereas residual uncleaved fusion protein as well as the liberated TrxA moiety remained bound to the column. The resulting PAS polypeptide solution was concentrated by ultrafiltration (10 kDa cut-off for PAS(400) and PAS(600), 3.5 kDa cut-off for PAS(200)).
For concentration measurement the PAS samples were diluted 1:1000 with spectroscopy buffer (50 mM K 2 SO 4 , 5 mM KP i pH 7.5). After that, a 350 µL aliquot of the diluted polypeptide solution was applied to a PD-10 column pre-equilibrated with spectroscopy buffer and eluted in 3.5 mL of the same buffer, leading to a final 1:10 000 dilution with respect to the starting concentration of the PAS polypeptide. UV absorption spectra were recorded in a UV/Vis Spectrophotometer (Ultrospec 2100 pro; GE Healthcare) and quantified at 205 nm: [11] (1)

Mass spectrometry of PAS polymers
The PAS-IL-1Ra fusion proteins and the isolated PAS polypeptides were dialyzed against

SDS-PAGE
Fusion proteins, PAS polypeptides and PEG polymers were analyzed by SDS-PAGE using the buffer system of Fling & Gregerson. [12] PEG can be conventionally stained in SDS polyacrylamide gels using a barium iodide/iodine solution. [13][14] We found that essentially the same procedure can be applied to detect PAS polypeptides. In our modified protocol, the gel (after electrophoresis) was first rinsed with water and then incubated in 2.5 % (w/v) BaI 2 (barium iodide dihydrate; Sigma-Aldrich) for 5 min. After rinsing with water again, the gel was transferred to Lugol solution, prepared by dissolving 5 % (w/v) I 2 (Riedel de Haen, Seelze-Hannover, Germany) in 10 % (w/v) KI (AppliChem), and incubated for 5 min. Subsequent destaining in water cleared the red-brownish background to yellow and led to dark red/brown bands. Of note, the stain faded within several minutes, thus requiring immediate photographic recording.

Reverse Phase Chromatography (RPC)
Solutions of the purified PAS polypeptides and fluorescein-labelled PEG polymers with concentrations of about 1 mg/mL were first adjusted to 2 % (v/v) acetonitrile, 0.1 % (v/v) p. 8 formic acid (buffer A). Then, a 500 µl sample was applied to a 1 mL Resource RPC column (GE Healthcare) equilibrated with buffer A. Elution was subsequently achieved with a linear gradient from 100 % buffer A to 100 % buffer B (80 % (v/v) acetonitrile, 0.1 % (v/v) formic) over 20 bed volumes at a flow rate of 2 mL/min. Sample-specific UV absorption was monitored for both types of polymer at 225 nm, thus detecting either the PAS peptide bonds or the chromophore of the fluorescein-PEG conjugate. Note that the fluorescein-specific absorption around 496 nm does not prevail in the acidic milieu of the buffers applied here.

Circular dichroism spectroscopy (CD)
Secondary structure analyses were performed with a J-810 spectropolarimeter (Jasco, Groß-Umstadt, Germany) using a 0.1 mm path length quartz cuvette (106-QS; Hellma, Müllheim, Germany) and ~10 µM protein solutions in 50 mM K 2 SO 4 , 5 mM KP i pH 7.5. Spectra were recorded at room temperature within a wavelength range from 190 to 250 nm by accumulating 15 runs (bandwidth 1 nm, scan speed 100 nm/min, response time 4 s). The spectra were corrected for buffer blank and smoothed using the instrument software. The molar ellipticity θ M was calculated according to the equation where θ obs is the measured ellipticity, c the protein concentration [mol/L] and d the path length of the quartz cuvette [cm]. The normalized data were plotted against the wavelength using KaleidaGraph (Synergy Software, Reading, PA). For the PAS-IL-1Ra fusion proteins, the θ M spectrum of the unfused IL-1Ra was subtracted to obtain the difference spectrum corresponding to the PAS moiety alone. In another experiment, a linear temperature gradient from 20 to 90 °C was applied at 1 K/min and the change in ellipticity at 195 nm was monitored.

Size exclusion chromatography (SEC)
A protein sample of 300 µL was applied to a Superdex S200 10/300 GL column (GE Healthcare) equilibrated with 500 mM NaCl, 50 mM NH 4

Viscometry
Polymer viscosity was measured at 25 °C for dilution series ranging from 100 to 1 mg/mL of the PAS and PEG samples using an m-VROC microviscometer (Rheosense, San Ramon, CA) equipped with an mVROC2.5-GA05 flow cell. Individual viscosity data represent the average from triplicate measurements of the same sample. Generally, the dependence of viscosity on solute concentration c can be mathematically expressed as a Taylor series: [15] (3) For purposes of graphical display (see, e.g., Fig. 3a), each the lowest order (i.e., from 3 to 6) was chosen to yield an acceptable curve fit.
At reasonably low polymer concentrations, higher polynomial orders become negligible and different linearization methods with regard to c can be applied, e.g. as proposed by Huggins [16] (4) or, alternatively, by Kraemer [17] (5): as derived from the measured sample viscosity η and the viscosity of the pure solvent η 0 .
The intrinsic viscosity, [η], of the polymer solute can be obtained by extrapolation of both equations to infinitesimal concentration: for each PAS or PEG polymer were evaluated according to both methods using concentrations of 10, 20, 30, 40, and 50 mg/mL and curve fit according to equations (4) and (5), respectively (see Fig. S3). Values for [η] were then calculated from the two linear graphs as the mean of both extrapolated axis intercepts, i.e. at zero sample concentration c.
To estimate the hydrodynamic radius, ! !"#$ , from this viscometric analysis the Einstein-Simha relation [18] can be applied to solute molecules under the assumption of a spherical shape (with mass M and the Avogadro number N A ): with the proportionality constant K and the so-called shape factor a, which ranges from 0.5 to 1.0 for random coil polymers.
Notably, these two characteristic parameters can also be estimated independently from the SEC data for the investigated polymers (see Fig. 3d). In SEC, the elution volume V is usually considered to correlate with M as follows: [20] (11) = − · av + A and B are constants while the K av parameter relates to the measured elution volume, V, with V 0 as the void volume and V t as the total (bed) volume of the column: In order to obtain a universal calibration curve valid for all kinds of polymers irrespective of their specific conformational preferences, [20][21] the intrinsic viscosity [η] should be included as another factor in equation (11) in a more precise treatment since, in fact, it is the logarithm of the hydrodynamic volume determined by the product [η]·M (not just M) that is proportional to K av : [21] (13) ( · ) = − · av + Usually, for molecular weight estimation of ordinary globular proteins a constant value of [η] is assumed (typically in the range of 0.71-0.75), [22] which results in the same ordinate intercept (included in the apparent constant B) for both the protein calibrants and analytes and, therefore, its influence is neglected. weight for such polymers [23] according to the conventional equation (11), as was also applied for the ordinary SEC analysis in the present study up to this point.
If the K av values for different lengths (i.e., masses M) of the same polymer are evaluated according to equation (11), the constants A and B -with the latter implicitly including -log [η] for this polymer according to eq. (13) -can be individually determined by linear regression for each polymer type.
On the other hand, rearrangement of equation (10) yields: As the expressions for log M from equations (11) and (14) are similar, both can be combined and rearranged to a new relationship between [η] and the SEC elution volumes: With the knowledge of A and B for a polymer as determined from the K av measurements by SEC as explained above, a linear plot of log [η] from the corresponding viscometry data versus K av permits independent evaluation (without explicit use of the polymer mass) of the Mark-Houwink parameters a (from the slope, -a·A) and K (from the ordinate intercept, a·B + log K; see Fig. 3d and Fig. S4).
Notably, the relationship between [η] and M according to equation (9), in combination with equation (10), can also be employed for alternative determination of parameters a and K using the correlation between [η] and the corresponding hydrodynamic radius ! !"#$ (with the compound constant = !"·!·! ! ! ): However, this equation still contains the molecular mass M, which differs for the polymers under investigation. Applying the expression of M according to equation (10) from the viscometry analysis leads to: Solving for [η] results in the final expression: Hellma) and the hydrodynamic radii, ! !"# , were measured at 25 °C in a Zetasizer Nano-S instrument (Malvern Instruments, Herrenberg, Germany) as average of triplicates. To p. 14 investigate the influence of temperature on the polypeptide shape, the sample P/A#1(600) was also measured after heating from 20 to 90 °C in steps of 10 K.
In this DLS technique the hydrodynamic radius ! !"# is automatically calculated by the instrument software from the translational friction factor f as derived from the experimentally determined diffusion coefficient D according to the Stokes-Einstein model: [18] (22) = (22) contains the experimental ! !"# value together with the viscosity η of the dilute polymer solution, which is assumed as the one of pure water (1.00 cP at 20 °C). Notably, equation (23) is only valid for a spherical particle shape, whereas different molecular shapes lead to deviations from the linear relationship between f and ! !"# .
For less anisotropic, elongated molecular shapes the rotational contribution to the friction factor can no longer be neglected, which leads to enlarged experimental friction factors and, likewise, apparent hydrodynamic radii obtained by DLS. [24] For this reason, the DLS method is even more sensitive to the molecular shape than SEC (see above). Thus, a comparison of apparent molecular masses independently estimated by both techniques can be used as an empirical measure, which we have termed "hydrodynamic quotient", for the degree of conformational elongation of a polymer chain in solution.
Perrin [25] has shown a way to estimate the shape of an ellipsoid, serving as a first approximation of an average molecular ensemble structure deviating from the ideal spherical shape in solution, by calculating the friction factor, f P , from the ratio of the corresponding hydrodynamic radii: The derivation by Perrin finally leads to two mathematical expressions for f P that can be interpreted as prolate and oblate ellipsoids. [25][26] In the present study, only the prolate ellipsoid was considered as a reasonable shape estimation for the investigated polymers since a linear chain should exhibit a single preferred dimension of propagation: With an algorithm implemented in the DLS instrument software (Zetasizer Software version 6.32; Malvern Instruments) mathematical solutions for the prolate axial ratio p, derived from the f P value, can be calculated from the measured ! !"# and the independently determined ! !"#$ parameters. Of note, instead of ! !"#$ , the specific volume was the required input for the algorithm. To this end, was calculated from [η] via the hydrodynamic volume VH for spherical molecules (see section on viscometry) by rearrangement of equation (9) using thus yielding: 5 p. 17          Table S5). (b) CD spectra of P1A1(200) and P1A3(200) polypeptides, both at 20 and 90 °C. Spectra are shown in solid lines together with the corresponding difference spectra (dashed lines) between the two temperatures. Local intensity minima and maxima are marked with black dots in order to illustrate the spectral changes upon change in temperature. The P1A1 polypeptide loses contributions from random coil while gaining additional PPII structure at elevated temperatures (see also Table S3). In contrast, P1A3 aggregated during the heating process and yielded a CD spectrum indicating a structure dominated by β-turn.