Chemotactic Activity of Site-Specific Multivalent Conjugates of Stromal Cell-Derived Factor 1α on Branched Nanoparticles

Stromal cell-derived factor 1α (SDF1α) is a potent chemokine for the recruitment of stem cells. A challenge is to maintain its activity and control its release. In this study, we engineered a recombinant cysteine-SDF1α (cysSDF1α) protein, and performed multivalent conjugation of cysSDF1 through the maleimide functional group to two forms of branched nanoparticles: multi-arm poly (ethylene glycol) (MA-PEG) and hyaluronic acid (HA). We characterized the chemotactic activity of the conjugates, and determined how the molecular weight (MW) of MA-PEG and HA affected the chemotactic activity. CysSDF1α had similar efficiency to wild-type SDF1α in cell recruitment. Multivalent conjugation of cysSDF1α to low MW MA-PEG (~18 nm) did not significantly affect the chemotactic activity, while the conjugation of cysSDF1α to high MW MA-PEG (~72 nm) lowered the efficiency, possibly due to the larger spacing between conjugated SDF1α molecules. HA has a linear backbone and a high density of multivalent binding sites; however, the chemotactic activity of HA-linked cys-SDF1α was much lower, which further decreased with the increase of HA MW from 200 kDa (~0.78 μm) to 700 kDa (~2.7 μm). Digestion of HA into smaller fragments using hyaluronidase partially recovered the chemotactic activity of cysSDF1α, suggesting that high MW HA might exert steric hindrance for SDF1α binding to its receptors on cell surface and that HA could be used as a depot for SDF1α storage and release. These results demonstrate that multivalent conjugates of SDF1α to nanoparticles may be used to engineer SDF1α delivery for cell recruitment and tissue regeneration.


Introduction
Stromal cell-derived factor 1α (SDF1α) is a chemokine released by bone marrow stromal cells for homing hematopoietic stem cells (HSCs) via a receptor-ligand CXCR4-CXCL12 interaction [1] [2] [3]. In addition, SDF1α can be released upon tissue injury and recruit bone marrow cells, stem cells and progenitor cells for tissue regeneration [4]- [11]. However, the transient release of SDF1α by native tissues may not be sufficient to induce sustained responses for tissue repair. Therefore, drug delivery approaches have been developed to release SDF1α to promote tissue regeneration [12] following myocardial infarction [13] [14], muscle [15], liver [16], and neural regeneration [17]. Several methods have been developed to immobilize SDF1α onto delivery vehicles, mainly relying on heparin-mediated binding, in which the heparin covalently links to the vehicles via an NH 2 -PEG-NH 2 linker or hydrogel linker, and SDF1α is then immobilized to the conjugated heparin through its heparin-binding domain [18]- [24]. It has been shown that clustering of vascular endothelial growth factors [25], fibroblast growth factor-2 [26], or epidermal growth factors synergistically increases the activity by inducing receptor dimerization [27]. However, it is not clear whether multivalent conjugation of SDF1α onto nanoparticles affects its activity and potency.
Non-specific conjugation of proteins can occur at every amine group and carboxyl group, including lysine, N terminus, glutamic acid, and aspartic acid, and thus increases heterogeneity of molecular orientation and bioactivity. To preserve protein bioactivity and yield during ligation, several site-specific methods have been previously created to ligate proteins such as intein-mediated ligation, basic expressed protein ligation, sortase-tag expressed protein ligation, Staudinger ligation and ligation via the thiol group and maleimides [28]- [35]. To engineer SDF1α, we encoded an extra cysteine residue as a linking site at the C terminus since the N terminus of SDF1α is closed to its binding domain to its receptor.
This engineered cysSDF1α was linked to the maleimide functional group in multi-arm poly (ethylene glycol) (MA-PEG) or the backbone of hyaluronic acid (HA).
We then investigated the effects of multivalent conjugations on the chemotactic activity of SDF1α.

Cloning of Wild-Type (WT)-SDF1α and Cysteine (cys)SDF1α
DNA plasmids for WT-SDF1α and cysSDF1α were designed and constructed. To express WT-SDF1α and cysSDF1α, Rosetta 2(DE3)pLysS competent cells were transformed with 2AT-WT-SDF1α and 2AT-cysSDF1α plasmids, respectively, and subsequently cultured on agar plates containing carbenicillin (100 μg/ml). A single colony was picked and incubated overnight in 25 mL of 2YT media with carbenicillin (100 μg/ml) on a shaker at 37˚C. One liter of 2YT media with carbenicillin (100 μg/ml) was inoculated with 10 mL of the overnight starter culture and incubated with shaking at 37˚C. When OD 600 nm of the culture reached 0.6, Y.-F. Hsieh et al. Journal of Biomaterials and Nanobiotechnology IPTG was added to achieve a final concentration of 1 mM to induce protein expression. Induced cells were allowed to grow for 4 hours and subsequently were harvested by centrifuging at 4000 rpm for 15 minutes. The cell pellets were re-suspended in 20 mL of PBS with 10% glycerol and homogenized. The cell lysate was centrifuged at 15,000 rpm for 30 minutes at 4˚C. WT-SDF1α and cysSDF1α were found in the inclusion bodies.

Isolation and Refolding of WT-SDF1α and cysSDF1α
WT-SDF1α and cysSDF1α were extracted and isolated following previously published protocol [3]. In short, inclusion bodies were washed 3 times with buffer A (100 mM Tris-HCl, 5 mM EDTA, 5 mM dithiothreitol, 2 M urea, 2% Triton X-100, pH 8.0) and once with buffer B (100 mM Tris-HCl, 5 mM EDTA, 5 mM dithiothreitol, pH 8.0). During each wash, the supernatant was discarded, and inclusion bodies were re-suspended in the buffer thoroughly and sonicated for 5 minutes. Re-suspended samples were subsequently centrifuged at 13,000 rpm at 4˚C for 20 minutes. After the last wash with buffer B, inclusion body pellet was solubilized in 6 M guanidine-HCl (5 mL guanidine-HCl per 1 L bacterial culture), and diluted 1:100 into refolding buffer (100 mM Tris-HCl, 5 mM EDTA, 0.2 mM oxidized glutathione, 1 mM reduced glutathione, pH 8.0) and stirred at 4˚C overnight.

Purification of WT-SDF1α and cysSDF1α
Two different molecular weight cut-off-(MWCO) sized protein concentrators were used to desalt and purify WT-SDF1α and cysSDF1α. Refolded WT and cysSDF1α were filtered through a 0.22 μm filter system to remove cell debris and other large insoluble compounds. Protein samples were subsequently concentrated with 5 K MWCO UF centrifugal concentrators (Corning). This concentration step simultaneously removed contaminants that were smaller than 5 K MWCO in average size.
Concentrated protein samples were diluted in storage buffer (10 mM HEPES, 150 mM NaCl, pH 7.9) and concentrated again, using the same 5 K MWCO concentrators. This dilution followed by a concentration step was repeated a few times using storage buffer to remove refolding buffer at the greatest extent. Lastly, the protein samples were filtered through 30 K MWCO centrifugal concentrators (EMD Millipore) to remove contaminant proteins bigger than 30 K MWCO in average size. SDS-PAGE was performed on final purified protein followed by Simply Blue staining (Thermo Fisher) and Western blotting procedures to determine the protein purity. The commercially available SDF1α ELISA kit (R&D Systems) was used to quantify the SDF1α concentration.

Prepare MA-and HA-SDF1α
Multivalent SDF1α (HA and MA) conjugation and buffer exchange EMCH-HA ligation. Before conjugating cysSDF1α to HA, the HA was modified with a maleimide linker, following the method published by the Healy group. Two sizes of HA sodium salt were purchased from Lifecore Biomedical with molecular weights

Chemotaxis Assay
The chemotaxis assay was performed by using the Boyden chamber system in 96-well-transwell plates (Corning HTS-Transwell-96-well Plate, #3389) with 5-μm pore size polycarbonate membranes. Jurkat T-lymphocytic cells (Scientific Facilities, UC Berkeley) were maintained in RPMI media with 10% fetal bovine serum (FBS). For the chemotaxis assay, Jurkat cells were trypsinized and re-suspended in RPMI media with 1% FBS, and 100 μL of cell suspension (3,000,000 -5,000,000 cells/mL) was added to each well. SDF1α variants and conjugates were added to the lower chamber at the desirable concentration, and incubated at 37˚C for 4 hours. Then the cells suspended in the bottom wells were counted with a hemocytometer.

Release of Conjugated SDF1α from HA by Hyaluronidase
100 ng/mL of HA complexes and 4 mg/mL of hyaluronidase (Sigma Aldrich, #H3506) was dissolved in RPMI media. The solution was incubated at 37˚C for 1 hour or overnight digestion. After cooling to room temperature, the solution was used for chemotaxis assay.

Reduced SDS-PAGE Gel
The samples of all SDF1α constructs were lyophilized into powder and then reconstituted in 15 μL di-water. The 15 μL sample was thoroughly mixed with 1 μL β-mercaptoethanol and 4 μL sample buffer, and then heated for 5 minutes at 85˚C. The gradient SDS-PAGE gel (Bio-Rad) was used for electrophoresis to separate conjugates. The sample mixture and ladder dye were loaded in the well respectively. The gel was run at 100 V for 1.5 hours, then stained with a 1X Coomassie blue staining solution for 3 hours, and washed with di-water overnight.

Non-Specific-Linked SDF1α Does Not Retain Chemotactic Activity
First, we determined whether clustering of SDF1α via non-specific covalent crosslinking retains its chemotactic activity. SDF1α molecules were non-specifically crosslinked using the EDC-NHS coupling method. The EDC-NHS coupling reagent randomly links any carboxyl group to any amine group of the proteins resulting in protein clustering. In principle, linking SDF1α could increase the bioactivity, but on the other hand, clusters have a high probability of hiding the cell-sensing domains thus leading to lower chemotaxis bioactivity. As shown in

Verification of Chemotactic Activity of cysSDF1α
To enable site-specific-ligation of SDF1α we added a cysteine residue at the N-terminus of SDF1α, expressed cysSDF1α proteins in E. coli culture, and then purified and re-folded the recombinant protein. Cys-SDF expression and purification were verified via SDS-PAGE. In Figure 1(b), the SDS-PAGE gel showed that both reduced and non-reduced forms of SDF1α could be purified. To verify that cysSDF1α folded into the active conformation and retained bioactivity, we tested its chemotactic bioactivity. The results showed that both WT-SDF1α and cysSDF1α had similar bioactivity to cSDF1α at 10 ng/ml (Figure 1(c)). Dose-titration experiments showed that cysSDF1α activity exhibited dose dependent increase in the range 0.1 -10 ng/mL and that cysSDF1α was significantly more potent than cSDF1α at 1 ng/ml concentration (Figure 1(d)). However, cysSDF1α 100 ng/mL did not show further increase of chemotactic activity, which might be attributed

Chemotactic Study of Multivalent SDF1α Conjugated to MA-PEG
We generated multivalent SDF1α by ligating cysSDF1α molecules to 10 kDa and 40 kDa maleimide-modified, star-shaped 4-arm PEG via Michael addition reaction (Figure 2(a)). The sizes of the individual arms were around 18 nm and 72 nm for 10MA-PEG and 40MA-PEG molecule, respectively (Figure 2(b)).
To determine the efficiency of ligation, 10MA-PEG were concentrated 50-fold by using a lyophilizer after buffer exchanging in a dialysis tube, and then ran in the reduced SDS-PAGE gel. In Figure 3(a), the 10MA-PEG shows 5 different bands around 10 -50 kDa, corresponding to cysSDF1α monomer, and 1-, 2-, 3-, and 4-cysSDF1α-linked PEG.
Each band was shifted by roughly 8 kDa (the molecular weight of one SDF1α molecule). This small difference between these bands made it difficult to separate by size exclusion columns; therefore, we investigated whether this conjugation affected the chemotactic activity by using the heterogeneous mixture of multivalent SDF1α. In this case, the role of the PEG backbone served as an SDF1α carrier to 1) to help maintain the bioactivity of SDF1α, and 2) to bring SDF1α molecules together through multivalent conjugation.
The bioactivity of these 4-arm PEG-SDF1α were tested (Figure 3(b)). The results showed that both 10MA-PEG-SDF1α and 40MA-PEG-SDF1α had dose dependency between 1 and 100 ng/mL. In addition, 10MA-PEG-SDF1α had significantly higher chemotactic bioactivity than 40MA-PEG-SDF1α. The reduced activity of 40MA-PEG as compared to 10MA-PEG could be due to greater spacing between cysSDF1α on 40MA-PEG or greater steric hindrance from the 40MA-PEG backbone. Additionally, not only 10 ng/mL of 10MA-PEG displayed similar potency to 100 ng/mL of cSDF1α and cysSDF1α, but also 100 ng/mL of 10MA-PEG successfully increase potency than 10 ng/mL of 10MA-PEG, suggesting 10MA-PEG has potential to avoid passive dimerization. Therefore, 10MA-PEG could be potentially used to maintain SDF1α activity for delivery at high concentration. This mainly because the MA backbone structure linked to the back side of SDF1α where SDF1α used to dimerized to each other.

Chemotactic Activity of SDF1α Conjugated to HA
Compared to MA-PEG, HA has far more potential conjugation sites and is bigger in size. The conjugation scheme is shown in Figure 4. Interestingly, HAconjugated SDF1α had very low chemotactic bioactivity ( Figure 5(a)). However, after we digested HAs with hyaluronidase for 1 hour or overnight, HA-SDF1α fragments regained chemotactic activity ( Figure 5(b)). These results suggest that HA might sterically hinder SDF1α binding to its receptors on the cell surface. Presumably, once HA was fragmented, the conjugated proteins were dispersed and exposed so that the bioactivity was recovered. Therefore, HA could be used as a car-

Conclusions
Since non-specific ligation of SDF1α reduced the bioactivity, ligation of SDF1α with site-specific bonding is necessary. Ligating SDF1α with MA-PEG and HA Journal of Biomaterials and Nanobiotechnology respectively, we found that: 1) The smaller MA-PEG backbone performed better than the larger sized MA-PEG backbone. This may be explained by way of the longer arms of higher MW PEG keeping SDF1α too far apart; 2) By ligating to the MA-PEG backbone, SDF1α can increase potency at higher concentration, that is, the MA-PEG backbone can avoid SDF1α passive dimerization; 3) While HA ligation greatly reduced SDF1α bioactivity, it can be used as a strategy to load SDF1α in HA for sustained release by hyaluronidase.
On balance, even though we have found useful information about carrier backbone selection, this work has suffered from a low yield of cysSDF1α protein synthesis. We tried three different refolding buffer protocols for cysteine protein, but cysSDF1α protein still showed high aggregation in refolding. Computational biology calculations and use of a different mutation site may possibly increase the stability of cysSDF1α and by extension increase the protein yield. Once the protein yield is increased, our results could have more practical usage and they could also improve detection methods for 10 MA-PEG.