Metal‐HisTag coordination for remote loading of very small quantities of biomacromolecules into PLGA microspheres

Abstract Challenges to discovery and preclinical development of long‐acting release systems for protein therapeutics include protein instability, use of organic solvents during encapsulation, specialized equipment and personnel, and high costs of proteins. We sought to overcome these issues by combining remote‐loading self‐healing encapsulation with binding HisTag protein to transition metal ions. Porous, drug‐free self‐healing microspheres of copolymers of lactic and glycolic acids with high molecular weight dextran sulfate and immobilized divalent transition metal (M2+) ions were placed in the presence of proteins with or without HisTags to bind the protein in the pores of the polymer before healing the surface pores with modest temperature. Using human serum albumin, insulin‐like growth factor 1, and granulocyte‐macrophage colony‐stimulating factor (GM‐CSF), encapsulated efficiencies of immunoreactive protein relative to nonencapsulation protein solutions increased from ~41%, ~23%, and ~9%, respectively, without Zn2+ and HisTags to ~100%, ~83%, and ~75% with Zn2+ and HisTags. These three proteins were continuously released in immunoreactive form over seven to ten weeks to 73%–100% complete release, and GM‐CSF showed bioactivity >95% relative to immunoreactive protein throughout the release interval. Increased encapsulation efficiencies were also found with other divalent transition metals ions (Co2+, Cu2+, Ni2+, and Zn2+), but not with Ca2+. Ethylenediaminetetraacetic acid was found to interfere with this process, reverting encapsulation efficiency back to Zn2+‐free levels. These results indicate that M2+‐immobilized self‐healing microspheres can be prepared for simple and efficient encapsulation by simple mixing in aqueous solutions. These formulations provide slow and continuous release of immunoreactive proteins of diverse types by using a amount of protein (e.g., <10 μg), which may be highly useful in the discovery and early preclinical development phase of new protein active pharmaceutical ingredients, allowing for improved translation to further development of potent proteins for local delivery.


| INTRODUCTION
Over the last several decades, the landscape of pharmaceutical drug products has been transformed from a near monolith of small molecules to a diverse space with biologics gaining more and more dominance. In 1982, the first genetically engineered form of insulin was approved. 1 By 2017, half of the top 10 best-selling drug products over the previous 15 years were biologics. 2 Recombinant proteins, fusion proteins, antibodies, and others biologics have led to therapeutic breakthroughs in a number of treatment areas. They are also costlier and often more complicated to discover, develop, formulate, and manufacture. Another challenge of biologics is that they must be injected, rather than taken orally like most small-molecule drug products, which is a significant impediment to patient compliance. 3 To reduce the number of injections and increase patient compliance, controlledrelease formulations have been developed, which require weekly, biweekly, or monthly injections rather than daily injections for noncontrolled-release formulations. Particularly useful for proteins, controlled release can also be helpful for local delivery to hard-toreach areas, like the brain, 4,5 joints, 6,7 and posterior segment of the eye. 8,9 One difficulty, however, in the evaluation and development of new protein APIs, which require slow release to evaluate drug efficacy, is that large quantities of the biomacromolecule are generally required during formulation of controlled release dosage forms. Moreover, common encapsulation procedures require trained personnel with the use of organic solvent-based unit operations. 3 These combined factors can significantly impede the early drug development process when producing and using large amounts of the proteins of interest can be financially infeasible. 10 Here, we aim to address these cases by creating a simple, general, and low-cost paradigm for preparation of local controlled-release dosage forms for protein drug discovery that could be simple enough for most any bench scientist to use.
Copolymers of lactic (or lactide) and glycolic (or glycolide) acids (PLGAs) have become a desired delivery vehicle for a wide variety of therapeutics, including peptides, proteins, antibodies, vaccine antigens, and nucleic acids. Advantages of PLGA microspheres include biocompatibility and biodegradability, injectability of PLGA microspheres through a syringe needle with minimal discomfort, and tunable and long-term complete release of the therapeutics, including peptides and proteins. [11][12][13][14][15] PLGA is used in at least 19 FDA-approved controlledrelease products on the market in the US, 3,16,17 and therefore usually is the first biodegradable polymer considered for such applications.
While many hurdles of PLGA drug product formulation have been overcome, one long-standing issue is protein stability during encapsulation. 18 Traditional methods of encapsulation in PLGA microspheres require exposing biomacromolecules to micronization, organic/aqueous interfaces, air/water interfaces, high shear stress, organic solvents, and high temperatures, all of which can result in instability or aggregation of the biomacromolecule and low encapsulation efficiencies. 16,[19][20][21][22][23][24][25] To avoid these stressors and the resulting damage to protein and low encapsulation efficiency, our group previously devised organic solvent-free self-healing microencapsulation, in which porous PLGA microspheres are mixed with an aqueous solution of biomacromolecule. 22 The temperature is then raised above the PLGA glass transition temperature (T g ), causing the pores in the surface of the microspheres to heal, encapsulating the biomacromolecule within the microspheres. 26,27 In active remote loading, a trapping agent is contained in the drug-free self-healing microspheres before exposure to the biomacromolecule to dramatically increase encapsulation efficiency. 18,28 The charge interaction between cationic peptides and the negatively charged carboxylic end-group of PLGA chains has also been targeted as an active remote loading strategy for smaller net cationic peptides that do not require preservation of tertiary structure. 29 Using active remote loading, encapsulation efficiencies greater than 95% have been achieved with elevated drug loading (>7% wt/wt). 18,28 Examples of trapping agents include aluminum-and calcium-based adjuvants 22,[30][31][32][33] for vaccines and glycosaminoglycan-like biopolymers 18 that often bind growth factors. One drawback of these methods is that the trapping agent must be paired with specific biomacromolecules that have binding affinity for said trapping agent, and therefore, the methods are not universal. Here, we aim to take advantage of the coordination binding between divalent transition metals and poly-histidine tags (HisTags) to create an active remote loading method that is more applicable to a broader spectrum of recombinant proteins.
Immobilized metal affinity chromatography (IMAC) was first developed in 1975 as a method of separating and purifying proteins based on their cysteine and histidine content. 34 These amino acids form coordination bonds with transition metals like Ni 2+ , Cu 2+ , and Zn 2+ . 35 Thus, proteins rich in cysteine and histidine can be purified by flowing them through a column with immobilized divalent transition metals. As the capability to express recombinant proteins expanded, so did strategies for IMAC. Today, HisTags (typically His 6 or His 10 ) can be expressed at the C-or N-terminus end of peptides and proteins, allowing them to be easily purified via IMAC. 36 To elute purified proteins or peptides of interest from the column, the pH can be lowered to protonate the histidine, interrupting the coordination bond, or imidazole, glycine, or a chelating agent like ethylenediaminetetraacetic acid (EDTA) can be added to the column buffer solution, displacing the molecule of interest. 37 In the active remote loading and self-encapsulation platform described in Figure 1, our approach is to directly encapsulate highmolecular-weight dextran sulfate (HDS), a negatively charged branched polysaccharide, 18 in drug-free and porous PLGA microspheres to serve as a metal-immobilizing scaffold. A divalent metal cation is then bound to the HDS to serve as a trapping agent for HisTag proteins before self-healing encapsulation. Our goal is to create a remote-loading controlled-release platform that (1) is virtually universal for any recombinant peptide or protein; (2) is highly efficient; (3) uses very small quantities of the protein or peptide; (4) slowly and continuously releases active protein; and (5) can be performed by scientists without training in microencapsulation and without specialized mixing or drying equipment. Meeting these goals would make this platform attractive during the discovery and early development of biologics, when producing and using large amounts of the molecules of interest could be very costly or infeasible and controlled-release efficacy data are desired. Due to the low quantities encapsulated here, local delivery of potent proteins is the targeted application of this platform. For example, a single injection of only 0.01% proteinloaded PLGA implants for controlled release of basic fibroblast growth factor was sufficient to rescue limbs and restore perfusion in a murine hindlimb ischemic model. 38 For decades, unpredictable or inadequate pharmacokinetics, resulting in undesirable toxicology and poor efficacy, has plagued drug candidates in clinical trials. 39 To combat this, drug delivery and formulation scientists have been introduced earlier in the development process. Allowing drug researchers with or without formulation expertise to easily and cost-effectively test early-stage drug candidates with a controlled-release formulation in vivo could allow for much better translation from the bench to the patient. 40 We offer this platform as a potential solution to this critical translational need.

| Preparation of microspheres
Porous PLGA microspheres with HDS as a metal immobilizer, MgCO 3 as a pH-modulator and porosigen, 18,19 and trehalose as a porosigen were prepared by double water-oil-water (w/o/w) emulsion and solvent evaporation. The first emulsion was created by homogenizing 1 ml of 250 mg/ml PLGA and 6% wt/wt MgCO 3 in methylene chloride with an inner-water phase of 200 μl of 4% wt/vol HDS and 3% wt/vol trehalose in a glass cell culture tube at 18,000 rpm for 60 s over an ice bath, using the Tempest IQ 2 . The second emulsion was created by adding 2 ml of 5% PVA to the primary emulsion and vortexing for 60 s. The w/o/w double emulsion was added to 100 ml of 0.5% PVA and stirred for 3 h at room temperature in a 150-ml beaker to allow for hardening and evaporation of methylene chloride. The 20-60 μm fraction of microspheres was collected using sieves and the microspheres were washed with double-distilled water and lyophilized.

| Assessment of microsphere morphology by scanning electron microscopy
The surface morphology of microspheres was examined via a Tescan F I G U R E 1 Schematic of remote loading mechanism into porous PLGA microspheres as related to immobilized metal affinity chromatography (IMAC). PLGA acts as the support structure. HDS acts as the chelating agent, immobilizing the metal ion, which binds a HisTag protein out of the loading solution through the porous network within the microsphere 2.4.2 | Inhibition of remote loading with EDTA The effect of EDTA on the capacity of Zn-loaded microspheres to remotely load HisTag HSA was determined by incubating Zn-loaded microspheres (or water-incubated microspheres as a control) in 50% saturated EDTA solution (or water as a control) for 24 h rotating at 30 rpm at room temperature. Microspheres then underwent HSA loading as described above and loading and encapsulation were determined using Coomassie Plus protein assay as described below.

| Effect of divalent metal cation on remote loading
To examine the effect of different divalent metal cations on the remote loading and encapsulation of HisTag IGF-1, zinc acetate, copper acetate, cobalt acetate, nickel acetate, or calcium acetate were used in the metal loading step as described above. Metal loading percentage and IGF-1 encapsulation efficiency were quantified as described below.

| Determination of divalent metal cation loaded
The amount of divalent metal cation remotely loaded into microspheres was determined by dissolving several mg of microspheres in acetone, centrifuging for 5 min at 8000 rpm, and removing the supernatant for three cycles. The pellet was then reconstituted in water and analyzed using a Perkin-Elmer Nexion 2000 ICP-MS using appropriate standards and scandium as an internal standard. Metal cation loading percentage was calculated as (mass of metal cation in microspheres/total mass of microspheres) Â 100.

| Determination of immunoreactive protein by ELISA
HSA and IGF-1 ELISA kits were purchased from Raybiotech and performed according to kit instructions to determine immunoreactive protein concentrations. GM-CSF ELISA kits were purchased from Raybiotech and PeproTech and were similarly applied. In all ELISAs, NoTag and HisTag proteins used for remote loading encapsulation were also included as reference standards.

| Determination of total protein by Coomassie Plus protein assay
Total protein content for No Tag and HisTag HSA in loading solutions was measured by Coomassie Plus protein assay using a 1:1 sample-toreagent ratio. BSA standards were used, with the HSA proteins included as reference standards, and absorbance was read at 595 nm in accordance with the protocol.

| Estimation of protein loading and encapsulation efficiency
Protein loading (l) in microspheres was estimated by ELISA and Coomassie Plus protein assay by comparing the final concentrations of protein in the loading solution to a control loading solution, which underwent the same conditions without microspheres as follows: where V (=0.1 ml), C C , and C MS are the volume of loading solution, concentration of protein in control loading solution, and the concentration of protein in the loading solution with microspheres, respectively.
Encapsulation efficiency of the available active protein (i.e., relative to unencapsulation control) was calculated as: where C C and C MS are the concentration of protein in control loading solution and the concentration of protein in the loading solution with microspheres quantified by ELISA, respectively. Encapsulation efficiency of available total protein was calculated similarly with concentrations of protein quantified by Coomassie Plus protein assay used in place of those measured by ELISA.
Actual active encapsulation efficiency of active protein was calculated by:

| Statistics
All significance testing was conducted using one-tailed Student's t tests. Statistical significance was considered p < 0.5.

| RESULTS AND DISCUSSION
To test our approach, HDS and MgCO 3 were co-encapsulated in the porous PLGA 50/50 microspheres, as described previously. 18 These microspheres were originally designed to microencapsulate growth factors that are known to bind to extracellular matrix. In these formulations, HDS binds the growth factor and MgCO 3 is present to both inhibit acid drop caused by PLGA hydrolysis and provide continuous release by production of salt when reacting to low-molecular-weight degradation products. 18,42 We demonstrated high loading and encapsulation efficiency, and slow release of vascular endothelial growth factor without significant loss of immunoreactivity or heparin-binding affinity for weeks during slow and continuous release. Basic proteins, bFGF20 and lysozyme, were similarly encapsulated. 18 To expand the capability of these microspheres to deliver a wider spectrum of pro- After metal-ion uptake in HDS/PLGA microspheres, exchanging the solution to a 100 μl solution of 1 μg of HisTag granulocytemacrophage colony-stimulating factor (HisTag GM-CSF), which has local delivery applications, [43][44][45] and raising the temp for 42 h at 43 C resulted in an estimated self-healing microencapsulation of~75% protein available in the loading solution ( Figure 2a; Table 1). If the Zn 2+ was not added before loading, the encapsulation efficiency (EE) dropped to~14%. If GM-CSF was added without a HisTag, the EE Although promising, the above remote loading example used a very low concentration of protein and only 49.5 ± 1% of protein remained immunoreactive in the control solution. Proteins at these low concentrations commonly bind to vessel walls even if coated with low-proteinbinding materials. 46 We then increased the protein concentration in the loading media to 50 μg/ml HisTag GM-CSF. In this case, the loading efficiency decreased slightly to~55%, and the differential advantage relative to the Zn 2+ -free/HisTag or Zn 2+ /NoTag controls also decreased slightly (Figure 3a). Moreover, encapsulation of active and total protein by employing ELISA (55 ± 7%) and Coomassie Plus protein assay (49 ± 1%), respectively, were performed and shown to yield consistent values. Here, 68 ± 6% of the protein in the control loading solution remained immunoreactive after incubation at loading conditions, which is much higher than at the lower concentration. Once again, we observed slow and continuous release of immunoreactive HisTag GM-CSF for 49 days (Figure 3b) The bioactivity of the protein was also monitored according to a CSF2RA-CSF2RB dimerization cell-based  While promising for GM-CSF, we further tested our approach with a second protein, IGF-1, which also has local delivery utility, 47-51 at the higher 50 μg/ml level. As expected, the HisTag IGF-1 was loaded in the self-healing Zn 2+ /HDS/PLGA microspheres at about 80% efficiency, as measured by ELISA (Figure 4a). All other controls displayed encapsulation efficiencies of~20% or less. Here, 59% ± 13% of the protein in the control loading solution remained immunoreactive after incubation at loading conditions. Release of the protein was again continuous and nearly complete over 56 days, although with a larger burst release than for GM-CSF (Figure 4b).
To examine the effect of the divalent cation on our approach, self-healing HDS/PLGA microspheres were exposed to acetate salts of transition metals (Co 2+ , Cu 2+ , Ni 2+ , and Zn 2+ ), an alkaline earth metal (Ca 2+ ), or no salt control. Compared to the no salt control, HisTag IGF-1 was encapsulated with higher efficiency in the microspheres exposed to divalent transition metals. Accounting for the amount of metal loaded into the microspheres ( Figure S2) and the amount of protein encapsulated in the metal-free microspheres, the encapsulated protein above control per mole of metal ion followed Cu 2+ >Zn 2+ >Ni 2+ >Co 2+ >>>Ca 2+ , as expected by the relative affinity of these cations for HisTag ( Figure 5). 34 Microspheres exposed to Ca 2+ showed no substantial difference in EE as compared to the no salt control, as Ca 2+ is known to possess much less affinity than transition metals for HisTags. Hence, these data further strongly support the HisTag-to-transition metal binding occurring during loading of the HisTag protein before self-healing encapsulation.
We then applied a third protein, HSA, as a model protein to examine various phenomena at lower cost and to further support the generality of the approach. As shown in Figure 6a, again the HisTag HSA bound preferentially to the Zn 2+ /HDS/PLGA microspheres, with an EE of >95%. Controls without HisTag, without Zn 2+ , and without HisTag and Zn 2+ were higher for HSA than for the previously studied proteins, but nonetheless all below~41% as measured by ELISA.
Here, 70% ± 1% of the protein in the control loading solution remained immunoreactive after incubation at loading conditions.
HisTag HSA release from the standard formulation was complete, and slow and continuous after a modest initial burst by ELISA (Figure 6b).
To further probe the Zn 2+ -HisTag coordination, we tested whether EDTA, a strong chelating agent often used to elute HisTag proteins of IMAC columns, interfered with the encapsulation of HisTag proteins. To do this, after exposing HDS/PLGA microspheres to Zn 2+ but before exposing the microspheres to HisTag HSA, we exposed the microspheres to EDTA. The EDTA/Zn 2+ /HDS/PLGA F I G U R E 5 The immobilization of divalent transition metals improves the total protein encapsulation efficiency of HisTag IGF-1 into PLGA microspheres. Encapsulation efficiency of HisTag IGF-1 into Ca 2+ -, Co 2+ -, Cu 2+ -, Ni 2+ -, and Zn 2+ -immobilized, and M 2+ -free PLGA microspheres. Ca 2+ EE avail by total protein assay not significantly greater than water control; p > 0.05. All transition metal EE avail by total protein assay significantly greater than water and Ca 2+ controls; p <0.05 microspheres showed nearly the same EE as microspheres that had been exposed to neither Zn 2+ nor EDTA, and far lower efficiency than the standard formulation ( Figure 7). These data again support the HisTag-to-transition metal binding and are consistent with EDTA entirely inhibiting the coordination.

| Protein stability considerations
As previously discussed, the stability of proteins is a major obstacle in

| Limitations and future work
Here we have demonstrated the basic concept to utilize M 2+ -HisTag binding to remotely load small quantities of proteins for controlled release. One potential limitation of this technique is that the HisTag must not impair the biological activity of the encapsulated protein. It is likely that this can be mitigated by the option of his-tagging either the amino-or carboxy-termini of the proteins. 36 Cleavable HisTags are often used in molecular biology 59 and it is possible that a physiologically cleavable HisTag could be utilized. There are multiple ways in which the approach could be improved and expanded in the future.
To improve the stability of proteins during encapsulation, the temperature should be decreased. We have previously shown that the addition of plasticizer to the polymer can cause healing at lower temperatures (e.g., 37 C) as the hydrated T g of PLGA is lowered. 22 Plasticization could also reduce the healing time, which further simplify the encapsulation method. Second, the initial burst in some of the above examples is higher than usually desired. We anticipate there are potential ways to mitigate this issue, such as controlling the distribution of M 2+ in the polymer matrix and the matrix porosity/microstructure.
Likewise, replacement of MgCO 3 with the less soluble base, ZnCO 3 , has been shown to reduce the initial burst of albumin from PLGA. 60 Third, protein loading will need to be increased in order to expand utility for systemic delivery or for local delivery of much less potent proteins (e.g., monoclonal antibodies). Fourth, it will be important to demonstrate in important animal models the utility of this approach for novel proteins. Due to the very low quantities of protein used in these experiments, direct measurement of loading is difficult. While LECO nitrogen analysis would require larger quantities, it is possible amino acid analysis or other colorimetric or fluorometric digestive assays could be used to directly measure the loaded protein. [61][62][63] Alternatively, the proteins may be quantified utilizing HisTag specific dyes or antibodies. 64

CONFLICT OF INTERESTS
The authors declare that there are no conflict of interests.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request.