A hydroxamic acid-methacrylated collagen conjugate for the modulation of inflammation-related MMP upregulation

Medical devices with matrix metalloproteinase (MMP) modulating functionality are highly desirable to restore tissue homeostasis in critical inflammation states, such as chronic wounds, rotator cuff tears and cancer. The introduction of MMP-modulating functionality in such devices is typically achieved via loading of either rapidly-diffusing chelating factors, e.g. EDTA, or MMP-cleavable substrates, raising issues in terms of non-controllable pharmacokinetics and enzymatic degradability, respectively. Aiming to accomplish inherent, long-term, device-induced MMP regulation, this study investigated the synthesis of a hydroxamic acid (HA)-methacrylated collagen conjugate as the building block of a soluble factor-free MMP-modulating hydrogel network with controlled enzymatic degradability. This was realised via a two-step synthetic route: (i) type I collagen was functionalised with photonetwork-inducing methacrylic anhydride (MA) adducts; (ii) this methacrylated product was activated with a water-soluble carbodiimide prior to reaction with hydroxylamine, resulting in MMP-chelating HA functions. Nearly-quantitative methacrylation of collagen amines was observed via 2,4,6-trinitrobenzenesulfonic acid (TNBS) assay. The molar content of HA adducts was indirectly quantified via conversion of remaining carboxyl functions into ethylenediamine (EDA), so that 12-16 mol.% HA was revealed in the conjugate. Resulting UV cured, HA-bearing collagen hydrogels proved to induce up to ~13 and ~32 RFU% activity reduction of MMP-9 and MMP-3, respectively, following 4-day incubation in vitro. No hydrogel-induced toxic response was observed following 4-day culture of G292 cells. The novel synthetic strategies described in this work provide a new insight into the systematic chemical manipulation of collagen materials aiming at the design of biomimetic, inflammation-responsive medical devices.


Introduction
Matrix metalloproteinases (MMPs) are a family of zinc and calcium dependent proteinases, which plays a key role in breaking down extracellular matrix (ECM) proteins. The levels of MMPs are precisely regulated and are responsible for key physiological events such as homeostasis, tissue remodelling, wound healing and angiogenesis [1][2][3][4][5][6][7]. Together with their beneficial role, numerous clinical studies have indicated that the overexpression of MMPs is associated with several inflammatory states, such as the ones found in chronic wounds, rotator cuff tears and cancer [8][9][10][11][12][13].
For example, elevated levels of MMPs are found in synovial fluid samples from the glenohumeral joints of patients with massive rotator cuff tears [14,15]. Likewise, 5-to 10-fold increases in MMP levels were found in chronic leg ulcers compared to acute healing wounds; these elevated levels of MMPs explain the turnover in chronic tissue and failed wound closure [16][17][18][19]. Here, the imbalance between MMPs and tissue inhibitors of MMPs (TIMPs) promote constant degradation of neo-tissue and regionspecific growth factors, so that the typical healing process is impaired. In light of the critical role MMPs play in maintaining the integrity of tissue ECM, growing research has therefore focused on the control of MMP expression aiming to manage inflammation and restore physiological homeostasis.
MMPs, and especially collagenases, cleave all three α chains of native type I, II and III collagens at a covalent bond between the glycine residue and the leucine or isoleucine residue, located approximately three-fourths of the way down the molecules from the N-terminus [21,22]. Despite their classification, all MMPs present a catalytic zinc-binding domain, as well as a cysteine containing pro-peptide domain at the molecular level [23]. The cysteine residue is bound to the zinc atom when the enzyme is secreted in its latent conformation (pro-MMP). When the cysteine residue is dissociated from the zinc, the active site of the enzyme is exposed, triggering proteolytic activity. In vivo, proteolytic removal of the pro-domain for MMP activation is mediated by other proteinases, including MMPs, whilst MMP regulation can be further achieved by complexation with TIMPs, which block access to the active site [24]. clinical translation has been hampered by limited specificity and dose-limiting side effects (e.g. musculoskeletal syndrome) following systemic administration [25][26][27].
(ii) Application of chelating compounds aiming to revert MMP configuration to the native, deactivated state. Chelating agents, such as ethylenediaminetetraacetic acid (EDTA), have been proposed as zinc-chelating species to induce complexation with the exposed active site and to possess broad range MMP deactivation. Although this strategy has been successfully employed in some commercial wound dressings made from either hydrolysed collagen (Biostep ® ), oxidized regenerated cellulose and hydrolysed collagen (Promogran ® ), or a collagen-sodium alginatecarboxylmethylcellulose composite (ColActive ® ) [28,29], challenges with respect to the controlled release of the chelating agents and long-term device functionality have to date been only partially addressed.
To avoid unwanted side and temporal effects and to achieve long-lasting MMP regulation in situ, the design of multifunctional polymer architectures has received growing attention aiming at soluble factor-free, device-based therapies [30,31].
Synthetic polymers, i.e. polyacrylates [32] and respective hydroxamic acid-bearing derivatives [26], have been proposed as proof-of-concept anti-MMP substrates, due to the presence of ionically-charged and metal-chelating groups, respectively, along the polymeric backbone.  [37][38][39]. UV-cured covalent networks made of functionalised collagen triple helices have also been presented with integrated MMP-chelating functions, whose swelling and compression properties can be controlled in a relatively independent fashion, via variation of e.g. type and extent of collagen functionalisation [40,41]. On the other hand, current synthetic routes can only target a small content of easily-accessible functional groups, i.e. primary amines, along the collagen backbone, resulting in restricted customisation of material functionalities, e.g. with regards to MMP-modulation capability. In light of their chemical reactivity and low molar content, primary amino terminations of collagen are mainly employed for the generation of covalent crosslinks, so that the introduction of MMP-modulating chemical functions in resulting covalent networks is challenging to accomplish and control independently of the network crosslink density. Furthermore, currently-available crosslinking strategies of collagen still raises concerns with regards to the occurrence of irreversible, non-controllable crosslinking side reactions [43]. Consequently, flexible synthetic routes are still to be developed to fully realise multifunctional collagen systems with independently-customisable format, properties and functions.
In this study, we aimed to explore this challenge by building a hydroxamic acidmethacrylated collagen conjugate as a biomimetic building block for the synthesis of UV-cured hydrogel networks with MMP-modulating capability and independentlyregulated crosslink density. Hydroxamic acid (HA) adducts were introduced via reaction with hydroxylamine, whereby HA well-known chelating activity was hypothesised to serve as means for MMP complexation and deactivation.
Methacrylate functions were introduced via derivatisation of collagen amino groups to prompt the synthesis of a UV-cured crosslinked network in order to achieve enhanced enzymatic stability in the resulting materials. In addition to mediating the synthesis of the crosslinked network, the derivatisation of collagen with methacrylate functions was also exploited to protect the highly reactive, terminal amino groups of (hydroxy-)lysine residues to allow for the selective modification of carboxylic functions avoiding either irreversible and non-controllable side reactions or time-consuming de-/protection work-up.
Further to the quantification of the degree of collagen methacrylation via 2,4,6-trinitrobenzenesulfonic acid (TNBS), a new method is proposed to assess HA-induced consumption of carboxylic groups via indirect carboxylic acid amination in respective methacrylated and HA-conjugated collagen.

Materials
Rat tails were provided by the School of Dentistry, University of Leeds, UK. Type I collagen was isolated in-house from the rat tail tendons with acidic treatment [44]. Sigma-Aldrich.

(2,4,6)-Trinitrobenzenesulfonic acid (TNBS) and Ninhydrin assays
The TNBS assay was used to directly measure derivatisation of amino groups and to indirectly measure derivatisation of carboxyl groups into either methacrylate, EDA or HA groups, and also respective collagen functionalisation. Briefly, 11 mg of dry samples were mixed with 1 mL of 4 wt.% NaHCO3 (pH 8. where Abs (346nm) is the UV absorbance value recorded at 346 nm, 2 is the dilution factor, 0.02 is the volume of sample solution (in litres), 1.46×10 4 is the molar absorption x is the dry sample weight, whilst mol(Lys)coll and mol(Lys)funct. represent the total molar content of free amino groups in native and functionalised collagen, respectively. (Lys) is hereby used to recognise that lysines make the highest contribution to the molar content of collagen free amino groups, although contributions from hydroxylysines and amino termini are also taken into account.
The Ninhydrin assay was further used to confirm the degree of derivatisation of the collagen amino groups previously-obtained via TNBS. 11 mg of dry sample were mixed with 4 mL distilled water and 1 mL of 8 wt.% Ninhydrin solution in acetone. The mixture was then incubated at 100 °C for 15 min. To terminate the reaction, the mixture was cooled on ice and 1 mL of 50 % (w/v) ethanol was added. The amount of free amino groups was determined by reading the absorbance at 570 nm against a blank sample (UV-Vis, Model 6305, Jenway). A standard calibration curve was prepared by carrying out above-mentioned assay with known mass of collagen. Three replicates were used for both TNBS and Ninhydrin measurements. Data are presented as mean±SD.

UV-induced hydrogel formation
Methacrylated collagen and respective HA-conjugated products were dissolved (1.

Quantification of gel content
The gel content was measured to investigate the overall portion of the covalent hydrogel network insoluble in 17.4 mM acetic acid solution [47]. Dry collagen networks where m1 is the dry mass after the incubation. Three replicates were used and the data are presented as mean±SD.

Compression tests
Air-dried UV-cured samples were equilibrated in PBS via overnight incubation at room temperature. Resulting PBS-equilibrated hydrogel discs (Ø: 18 mm; h: 5-6 mm) were compressed at room temperature with a compression rate of 3 mm·min -1 (Instron ElectroPuls E3000). A 250 N load cell was operated up to complete sample compression. Stress-strain curves were recorded and the compression modulus quantified as the slope of the plot linear region at 25-30% strain. Three replicates were employed for each collagen network composition. Data are presented as mean±SD. where θλ is the observed molar ellipticity (degrees) at wavelength λ, d is the path length and c is the concentration (0.2 mg·mL -1 ). MRW is the mean residue weight and equals to 91 g·mol -1 for amino acids [48].

Statistical analysis
Statistical analysis was carried out using OriginPro 8.5.1. Significance of difference was determined by one-way ANOVA and post-hoc Tukey test. A p value of less than 0.05 was considered to be significant different. Data are presented as mean ± SD.

Results and discussion
In the following, the design of a HA-methacrylated collagen conjugate will be

Synthesis of HA-methacrylated collagen conjugate
The TEA-catalysed reaction of MA with collagen proceeds via an amine-initiated nucleophilic addition/elimination mechanism and was therefore carried out prior to coupling with HA. This reaction with MA was exploited for two purposes (Scheme 1): (i) to introduce photoactive functions on to free amino terminations of collagen, e.g.
(hydroxy-)lysines and amino termini, responsible for the subsequent formation of a hydrogel network [36,40]; (ii) to protect highly-reactive amino terminations prior to selective derivatisation of carboxyl into MMP-chelating functional groups, so that intracrosslinking side reactions could be minimised.
MA-mediated functionalisation of collagen was confirmed via TNBS and the content of MA adducts proved to be controlled depending on MA/Lys molar ration selected during the reaction ( To achieve the HA-methacrylated collagen conjugate, carbodiimide-induced activation of sample MA25 was pursued in PBS, due to the limited solubility of collagen triple helices in organic solvents. Introduction of EDC and NHS in collagen solutions is known to induce crosslinking reaction between collagen amino groups and NHSactivated collagen carboxylic groups, resulting in the formation of a gel [51]. In agreement with previously-discussed TNBS results, solution gelation was not observed in our case, providing evidence that minimal unwanted intra-crosslinking reaction occurred between residual amino groups and activated carboxyl groups of collagen. At the same time, solubilisation of retrieved, HA-reacted collagen product in an aqueous solution containing I2959 promptly led to the formation of a UV-cured hydrogel, as expected in light of the activation of covalently-coupled MA functions. Consequently, assessment of HA content in HA-reacted collagen products can only be qualitative when using the above-mentioned methods.

Quantification
In order to overcome this challenge, we aimed to indirectly quantify the molar content of HA by assessing the molar content of collagen carboxyl groups prior to and following reaction with HA, since the carboxyl groups in collagen were directly involved in the coupling reaction. To reach this goal, derivatisation of collagen carboxyl groups into TNBS-and Ninhydrin-detectable terminal amino groups was pursued in both methacrylated and HA-reacted collagen products via reaction with EDA (Scheme 2).
To avoid unwanted crosslinking reaction and promote selective grafting of EDA molecules, samples MA25, HA2 and HA4 were reacted with a 50-molar excess of EDA following carboxyl function activation with EDC/NHS. No gel formation was observed during the EDC-mediated activation of collagen and following reaction of EDCactivated collagen with EDA. These observations provided evidence that an EDAgrafted rather than crosslinked product was obtained, supporting the validity of the proposed approach in selectively derivitising carboxylic acids into terminal amines. Table 2 reports values related to the molar content of EDA-and HA-related adducts, as well as respective degrees of HA-mediated collagen functionalisation (FHA). These results confirm the successful coupling of HA adducts on to collagen carboxyl functions. Considering an overall amount of carboxylic acids of 1×10 -3 mol·g -1 in rat tail collagen [43,44], above-mentioned results corresponded to an averaged degree of HA-mediated functionalisation of up to ~16 mol.% (Table 2). Together with TNBS, Ninhydrin assay was also carried out to further confirm the above findings. Also in this case, a decreased molar content of amino groups was recorded in HA-conjugated compared to methacrylated collagen samples. An averaged FHA value of up to ~14 mol.% was obtained, well in line with previous TNBS data. Results obtained via collagen amination and colorimetric assays therefore provided supporting evidence that covalent coupling of both methacrylate and HA adducts could be selectively accomplished by independently targeting collagen amino and carboxyl groups, respectively. The nearly-quantitative lysine functionalisation with MA adducts proved key to selectively target free carboxyl functions of methacrylated collagen in a controlled fashion. The presented synthetic approach proved therefore proved reliable to enable the quantification of free carboxylic groups avoiding either the occurrence of well-reported, undesired intramolecular crosslinking reactions between collagen amino and carboxyl functions, or the employment of time-consuming de-/protection work-up [53,[55][56][57]. Although low variation in HA coupling was measured by both TNBS and Ninhydrin assays, a significant effect on the MMP-modulating capability of resulting UV-cured hydrogels was observed (section 3.4). It is expected that a wider range of functionalisation can be obtained by further adjusting the molar ratio of hydroxylamine with respect to carboxylic acid groups of methacrylated collagen.

Analysis of collagen conformation
SDS-PAGE was used to both elucidate the chemical composition of reacted products and to explore the triple helix organisation of methacrylated and HA-conjugated products. Retention of the triple helix architecture of collagen is key to enable resulting materials with chemotactic functionality, enhanced mechanical properties and decreased swellability [39]. Although we previously confirmed that collagen methacrylation had minimal effect on the collagen triple helix structure [40], the reaction of collagen-based polypetides with hydroxylamine has been reported to catalyse the chemical cleavage of peptide bonds between asparagine and glycine residues [58][59][60], which is likely to induce a denaturation of collagen triple helices.
Despite the low (up to 9 mM) hydroxylamine concentrations employed in this study, it was important to confirm that no degradation product was formed in reacted species, aiming not to compromise chemical sequence and organisation of native collagen.
In-house extracted type I rat tail collagen was confirmed to display electrophoretic bands of monomeric a-chains (~100 kDa) and dimeric b-components (~200 kDa), as shown in Figure 1 (A). Each of these bands was observed in the electrophoretic patterns of samples MA25, HA2 and HA4, whilst no further band could be identified.
SDS-PAGE data therefore supported the fact that no detectable collagen degradation occurred during the reaction with hydroxylamine. The minimal pattern variation in the electrophoretic reference bands associated with reacted, with respect to, native collagen, also suggested no alteration to the native triple helix organisation following collagen functionalisation. To further elucidate this point, CD spectroscopy was carried out. Collagen presents a unique CD spectrum with a negative peak at 197 nm and a positive peak at 221 nm [36,61,62]; these characteristic peaks could be identified in all collagen samples (Figure 1 B), confirming the presence of polyproline-II and triple helices, respectively [63]. The magnitude ratio of positive to negative peak RPN) in the CD spectra provides an indication of the content of collagen triple helices [40]. The RPN values measured in the CD spectra of samples MA25, HA2 and HA4 were 0.113, 0.094 and 0.101, respectively, which were only slightly lower than the RPN value measured in the CD spectra of native collagen (RPN: 0.122). By normalising the RPN value measured from functionalised samples with respect to the one of native collagen, a degree of triple helix preservation of at least 77 RPN% is observed. These results provide supporting evidence that collagen triple helices could still be preserved in HA-conjugated collagen samples, although at lower extent with respect to the case of methacrylated samples.
The preservation of collagen triple helices in HA-conjugated collagen is an interesting finding, given that additional HA, besides MA, adducts are covalently coupled to the collagen backbone. This finding may be explained by the fact that the overall HA molar content was limited up to only 16 mol.% of collagen carboxyl functions, suggesting that increasing the degree of HA-mediated functionalisation may negatively affect the triple helix organisation of native collagen.

Hydrogel impact on MMP activities and related enzymatic degradability
Following elucidation of the molecular architecture, protein organisation, and UV-cured network formation, hydrogel proteolytic degradability and MMP-regulation capability were studied in vitro with either MMP-3 or MMP-9. Both MMPs were selected in light of their documented overexpression in inflammation states, such as in the case of rotator cuff tear [5], chronic wounds [3,9] and osteoarthritis [64]. Figure   2 (A) reports the MMP-3 activity of aqueous media following 4-day conditioning with samples MA25*, HA2* and HA4*, with respect to sample-free MMP-3 solutions. All samples proved to induce a significant reduction in MMP-3 activity (p<0.001), with samples HA2* (MMP-3: 81±4 RFU%) and HA4 (MMP-3: 70±6 RFU%) displaying significantly higher effect (p<0.01) with respect to methacrylated hydrogels (MMP-3: 87±2 RFU%). Together with analysis on supernatants, the relative mass of 4-day incubated samples was measured and compared to that of a control sample of native, non-crosslinked CRT (Figure 2 B). Hydrogel HA4* proved to show the highest relative mass (µrel: 96±2 wt.%), which was significantly higher than the ones recorded in samples HA2* (µrel: 79±4 wt.%) and MA25* (µrel: 79±3 wt.%). As expected, the lowest relative mass was found in the control sample (µrel: 44±4 wt.%), due to the absence of a crosslinked network at the molecular scale.
Similar trends were confirmed following sample incubation with MMP-9-containing aqueous media (Figure 3 A). MMP-9 activities were found to be reduced to either 88±5 or 87±4 RFU% following supernatant conditioning with either sample of HA2* or HA4*, with significant difference observed in comparison to supernatant treated with sample MA25* (MMP-9: 98±1 RFU%). As observed with MMP-3, sample CRT exhibits the lowest relative mass (µrel: 57±7 wt.%), whilst both MMP-chelating hydrogels displayed more than 90 wt.% averaged relative mass (Figure 3 B). In comparison, sample MA25* displayed a lower relative mass (µrel: 87±3 wt.%), which proved significantly different with respect to the ones of all crosslinked samples. MMP-3 and MMP-9 belong to two different classes of MMPs, namely stromelysins and gelatinases, respectively. Accordingly, MMP-3 presents broad substrate specificity and induces the activation of collagenases, e.g. MMP-1, a potent collagen-degrading enzyme. In contrast, MMP-9 is reported to preferentially degrade denatured collagen, i.e. gelatin, rather than collagen triple helices [65]. Our gravimetric data support the broader substrate specificity presented by MMP-3, whereby lower relative mass was recorded in both native collagen and hydrogel samples following incubation with MMP-3, compared to the case of the MMP-9-incubated samples. The broad specificity of MMP-3 compared to MMP-9 was also reflected by the higher hydrogelinduced activity reduction of the former compared to the latter enzyme. In light of MMP-9 specificity towards single rather than triple helices of collagen, above- reported HA-containing microspheres prepared via derivatisation of poly(methyl methacrylate-co-methacrylic acid) and respective microsphere-induced reduction of the activity of MMP-2, -3, -8 and -13 [26]. Other studies investigated the effect of ovine-based collagen dressing on a broad spectrum of MMPs reduction but led to undesired clinical outcomes [66][67][68][69]. Layer by layer SiRNA coated nylon bandage was reported to yield rapid chronic wound closure in a diabetic mice model silencing 60% of MMP-9 activity in a two week study [70]; however, the biocompatibility of nylon may need to be considered further to enable the clinical applicability of a resulting device.
We have recently reported a 4VBC-functionalised collagen hydrogel with covalently-bound electron rich aromatic adducts, such that MMP-9 activity was reduced by 50% in 4 days in vitro [38]. With respect to that study, the hereby presented HA-conjugated formulation is likely to offer additional advantages, given the possibility to control the MMP-modulating capability independently of the crosslink density of the network.

Characterisation of physical properties
Once the MMP-modulating functionality of HA-bearing hydrogels was addressed,  Interestingly, despite the lower gel content in HA-based compared to methacrylated networks, samples HA* could still display decreased mass loss and induce increased reduction of MMP activity, further confirming the effect of covalently-coupled HA adducts on to the collagen backbone. In line with the gel content data, the compression modulus (Figure 4 B) proved to be significantly affected by the hydrogel formulation, so that sample MA25* showed significantly higher compressive modulus (Ec: 128±17 kPa) with respect to samples HA2* (Ec: 81±10 Pa) and HA4* (Ec: 72±16 kPa).
HA derivatives are known to display radical scavenging functionality [71], suggesting that the presence of HA adducts in the methacrylated collagen precursor impacts on the half-life of UV-generated radicals, crosslinking reaction kinetics and network crosslink density. This hypothesis is likely to explain the trends in gel content across the different formulations, since decreased gel content and compression modulus were observed in HA-bearing compared to methacrylated samples. The indirectly-observed radical scavenging functionality of HA-bearing hydrogels could also play a major role in controlling the excessive upregulation of reactive oxygen species found in e.g. chronic wounds, providing an additional capability in controlling inflammation.

In vitro cytotoxicity
The potential use of the HA-conjugated collagen hydrogels in biological system was  Many reports suggested that soluble HA derivatives exhibit cytotoxicity against several cell lines such as human fibroblast, leukaemia and cancer cells [72][73][74][75]. This finding was not observed in this study as high cell viability was observed following 7-day culture onto HA-bearing collagen hydrogels with respect to native collagen controls. This observation further supports the use of HA-coupled rather than HA-loaded systems for the prolonged and localised regulation of upregulated MMPs.

Conclusions
The synthesis of a HA-methacrylated collagen conjugate was successfully demonstrated for the development of soluble factor-free hydrogel photonetworks with integrated MMP-regulating capability. Sequential covalent coupling of methacrylate and HA adducts to collagen was selectively accomplished by independently targeting amino and carboxyl groups, respectively. In this way, the hydrogel MMP-regulating capability could be controlled independently of the collagen network crosslink density.