A tunable and injectable local drug delivery system for personalized periodontal application

In periodontal treatment, patient differences in disease phenotype and treatment responses are well documented. Therefore, therapy duration and dosage should be tailored to the requirements of individual patients. To facilitate such personalized medication, a tunable and controllable system is needed to deliver drugs directly into the diseased periodontal pockets. The current study established a system to achieve different drug release rates and periods by incorporating bioactive agents into poly(lactic-co-glycolic acid) (PLGA) microspheres dispersed into a novel thermo-reversible polyisocyanopeptide (PIC) hydrogel. Specifically, two drugs, i.e. doxycycline and lipoxin, were separately loaded into acid-terminated and ester-capped PLGA by electrospraying. Different formulations were developed by loading the two kinds of PLGA microspheres with different mass ratios in the PIC gels. The results demonstrated that the PIC-PLGA vehicle exhibited appropriate injectability, long-term structural stability, and no obvious in vivo inflammatory response for the desired clinical application. Furthermore, the release profiles of drugs could be manipulated by adjusting the loaded mass ratio of acidand esterterminated PLGA microspheres in the PIC gels. The more ester-capped PLGA was used, the slower the release rate and the longer the release period, and vice versa. Additionally, the released drugs still preserved their bioefficacy. This PIC-PLGA system can be further developed and tested in translational studies to demonstrate the final clinical benefit.


Introduction
Periodontitis was reported as the 11th most prevalent disease globally in 2016 [1]. Currently, the primary treatment option is to remove the pathogenic biofilm mechanically. However, rapid recolonization of the biofilm after debridement often induces the recurrence of periodontitis in susceptible patients [2,3]. In such cases, adjunctive therapy is required, e.g. direct administration of antimicrobial, antiseptic or anti-inflammatory agents into the periodontal pocket [4]. Unfortunately, there is no standard protocol for every patient with refractory periodontitis, due to considerable differences in disease phenotype and treatment response between patients [5]. A solution for this problem can be found by tailoring the therapy duration and medication dosage to the requirements of the individual patient itself does not facilitate the long-term sustained release of biomolecules [7]. A prolonged release can be obtained by incorporating another drug carrier into the hydrogel without compromising the main advantages of the hydrogel, like its injectability.
Poly(lactic-co-glycolic acid) (PLGA) spheres are commonly combined with gels, due to long-existing experience in various clinical applications and favorable degradation characteristics [10]. Importantly, drug release kinetics can be adjusted by controlling the characteristics of PLGA, mainly by modifying the molecular weight and polymer endgroup [11,12]. Our previous study revealed that the molecular end group of PLGA had more effect on the release pattern of encapsulated doxycycline than molecular weight. A significant faster release rate for a shorter period was found in acid-terminated PLGA microspheres, compared to ester-capped spheres [13]. Consequently, we hypothesized that the drug release profile of PIC gel for periodontal therapy could be manipulated by adjusting the mass ratio of acid-and ester-capped PLGA microspheres loaded in the gel to achieve the desired personalized delivery system. Some examples of injectable nanoparticles/hydrogels composite for prolonged periodontal drug delivery can be found in the literature [14,15]. However, none of the investigated systems aims for the delivery of personalized medicine for periodontal application, which has been largely ignored in dentistry [6].
Therefore, this study aimed to develop a tunable and injectable local drug delivery system for periodontal application. Doxycycline (DOX) (which is a hydrophilic drug exhibiting anti-microbial as well as antiinflammatory effects) [16,17] and lipoxin A 4 (LXA 4 ) (which is a hydrophobic drug acting as a pro-resolving inflammatory mediator) [18,19] were chosen as therapeutic model drugs in this study [20]. Both drugs were separately loaded into acid-terminated and ester-capped PLGA microspheres by using an electrospraying procedure. Subsequently, 10 different formulations, were fabricated by adjusting the loaded weight ratio of the different kinds of PLGA microspheres in the PIC gel. The in vitro injectability, rheological properties, and drug release patterns of the various formulations were investigated. Subsequently, the efficacy of the released DOX and LXA 4 were further examined by in vitro assays. Finally, the tissue response of the various formulations was tested by subcutaneous implantation in a rat model.

Electrospraying drug-loaded PLGA microspheres
P2 or P2A was dissolved in acetone to a concentration of 0.16 g/mL. Then, either a solution of 24 mg DOX in 110 μL methanol or 50 μg LXA 4 in 500 μL ethanol was added to 2 mL PLGA solution. The mixture was stirred until clear. In total, 4 solutions were prepared to prepare drug loaded microspheres, i.e., P2-DOX, P2A-DOX, P2-LXA 4 , P2A-LXA 4 . The electrospraying procedure was referred to as our previous work [13]. Briefly, the solution was ejected at a flow rate of 10 μL/min under a voltage of 20 kV. The electrosprayed microspheres were collected in a water bath filled with 1% Poloxamer 407 in Milli-Q. The collector was gently shaken to avoid particle agglomeration. The distance between the spinneret and collector was 17-20 cm. Thereafter, the microspheres were centrifuged and washed with Milli-Q three times to remove poloxamer. Finally, the microspheres were freeze-dried and stored at −20°C until further use. Blank controls -P2 and P2A microspheres without loading drugs were also prepared by the same methods above for the injectability and rheology tests.

Morphology of the microspheres
The shape and surface morphology of the electrosprayed microspheres were observed using a field emission scanning electron microscope (Sigma 300, Zeiss, Heidenheim, Germany). The size distribution of the PLGA microspheres was determined by a dynamic light scattering (DLS) technique using a Zetasizer Nano-S instrument (Malven Instruments Ltd., Worcestershire, UK) at 25°C for 60s (n = 3).

Characterization of PIC gel loaded with blank microspheres
The synthesis of PIC polymer (Mw~635 kDa) was described previously [8]. A PIC solution was prepared by dissolving PIC polymer in cold PBS at 4°C overnight to a final concentration of 0.5 wt%. For the various characterization assays, four different weight ratios (5%,10%,15%, and 20%) of blank P2 or P2A microspheres, were separately dispersed into a cold PIC solution and mixed thoroughly under vibration for injectability and rheology measurements. To visualize the spatial distribution of PLGA microspheres in PIC gels, samples were freeze-dried and cut by a surgical blade. The cross-sections of the samples were observed by scanning electron microscopy as described above.

Injectability test
To evaluate the effect of the added PLGA microspheres on the injectability of PIC solutions, an injectability test was performed using a tensile bench (Lloyd Instruments; Ametek, West Sussex, UK) in a compression mode [21]. In brief, 1 mL ice-chilled solution was loaded in a plastic syringe (Becton Dickinson, Plymouth, UK) with a 22G needle. The syringe was positioned in the dynamometer holder downwards, and the plunger end of the syringe was placed in contact with a loading cell. Testing was carried out at the crosshead speed of 1 mm/s, representative of manual delivery to a patient. The average force required to sustain the movement of the plunger to expel the content of the syringe was measured (N), and it was recorded whether the microspheres were thoroughly pushed out with the gel. As controls Milli-Q water and pristine 0.5% wt PIC gel were tested. Three samples were tested in each group.

Rheology measurements
The samples with good injectability were further tested for viscoelastic and rheological properties using an AR2000 Advanced Rheometer (TA Instruments, Asse, Belgium) with a steel Peltier plate geometry (D = 40 mm). An aliquot of 800 μL of cold PIC-PLGA solution was loaded into the rheometer (T = 4°C). To test the temperature of the solgel transition and the mechanical properties of the formed gels, a temperature sweep program was initiated at 5°C with a heating rate of 2°C/min until 37°C with a load frequency of 1 Hz and strain of 2%. The sol-gel transition temperature (T sol-gel ) was determined as the onset of the increase in storage modulus G'. The value of G' and G" of each gel at 37°C were recorded. Thereafter, the temperature sweep started by cooling down at 2°C/min to 5°C to confirm the reversibility of the PIC-PLGA system. Pristine 0.5% wt PIC gel were tested as the control. Three samples were tested in each group. of microspheres in the PIC gel was decided. Within the loading range, for the drug release test, 15 mg of drug-loaded PLGA microspheres was dispersed in 200 μL cold PIC solution with the formulations shown in Table 1. For each drug, five different weight ratios of P2/P2A microspheres, 100/0, 75/25, 50/50,25/75, and 0/100 were mixed. PIC solution mixed with free DOX or LXA 4 , but without PLGA microspheres, served as control.
The drug release test was performed at 37°C as described before [7]. Briefly, the cold samples at~4°C were injected into 1.5 mL Eppendorf tubes and then gelled at 37°C (n = 3). Subsequently, 800 μL 37°C PBS was laid over the gels. Samples were then incubated at 37°C with agitation at 200 rpm/min. At each predetermined time point, 600 μL supernatant was withdrawn carefully and 600 μL fresh PBS was refilled. The amount of released LXA 4 and DOX in the supernatant was detected by reverse phase-high performance liquid chromatography (RP-HPLC) according to previous protocols [7,13]. The release experiments were performed up to 6 weeks. The cumulative drug release percentage of each group was calculated and normalized based on its total release amount in the entire release period. The pH of supernatant was measured at the same time to monitor the degradation of PLGA microspheres. Finally, light photographs of samples were also recorded for visual observation.

Efficacy test of the released DOX and LXA 4 2.6.1. Efficacy test of DOX
The efficacy of the released DOX was assessed by the ability to inhibit the gram-negative anaerobic bacteria Porphyromonas gingivalis ATCC 33277 (Pg) [22]. The release medium from Group D3 (P2/ P2A = 50/50) collected at day 3, day 8, and day 14, was chosen for this antibacterial test (n = 3). Specifically, 30 μL release medium was absorbed into a piece of sterilized filter paper (Ø8mm, 0.6 cm 2 ) and then air-dried. Different amounts of fresh DOX solution were made and absorbed into sterilized filter papers to serve as positive controls. Aliquots of 9.3 × 10 8 colony-forming units of Pg in physiological saline were spread evenly over the entire surface of a Brucella agar plate (Becton Dickinson). The prepared filter paper samples were placed on the Brucella agar plate. Penicillin and metronidazole tablets were used as positive controls. The agar plates were then transferred into GasPak jars (Becton Dickinson). An anaerobic environment with 0.2% O 2 , 9.9% CO 2 , 5% H 2 and 84.9% N 2 was achieved by Anoxomat Mark II (Mart Microbiology B.V., Drachten, the Netherlands). The inhibition zone was measured after incubation of 72 h at 36°C. 4 The efficacy of the released LXA 4 was assessed by the ability to stimulate RAW264.7 macrophage (Gibco) phagocytosis using a fluorescent bead internalization assay [7]. LXA 4 released from group L3 (P2/ P2A =50/50) at the time point of day 3 was used as the experimental group. Freshly prepared LXA 4 was used as a positive control (PC). The lipoxin solvent -ethanol, was used as a negative control (NC). Release medium from blank microspheres (P2/P2A =50/50) without LXA 4 was collected at the same time point as a vehicle control. Three samples were tested in each group. RAW264.7 cells were seeded in 96-well plates at a density of 2 × 10 4 cells/cm 2 in α-minimum essential medium eagle (Gibco) and 10% fetal bovine serum (Gibco) and left for 24 h for attachment prior to experimentation. Thereafter, each macrophage was exposed to approximately 100 Nile-red fluorescent latex beads (D = 2 μm; Invitrogen, Carlsbad, CA). Then, 50 μL LXA 4 releasate (concentration calculated by HPLC results) or freshly prepared LXA 4 (PC) was added into the culture to reach a final LXA 4 concentration of 100 nM. The same volume of NC and vehicle control medium was added accordingly. After incubation for 4 h, the fluorescence of nonphagocytosed beads was quenched with 0.4% trypan blue. The fluorescence intensity of each well was read in a Synergy HTX multimode reader (BioTek Instruments, Winooski, VT), with excitation/emission wavelengths at 535/575 nm.

Efficacy test of LXA
Phagocytosis results were also visualized by fluorescent microscopy. RAW 264.7 cells were seeded on an 8-well μ-slide (ibidi GmbH, Gräfelfing, Germany) at a density of 15,000 cells/cm 2 . The cells were treated using LXA 4 releasate and negative control medium with the same procedure mentioned above. After 4 h of incubation, the cells were washed with PBS and then fixed in 4% paraformaldehyde for 10 min and rinsed in PBS three times. Cell nuclei were stained with 0.5 g/mL 4,6-ditureamidino-2-phenylindole (DAPI; Invitrogen) for 10 min. Fluorescence was visualized under an argon-laser-powered confocal fluorescent microscope (Olympus FV1000, Zoeterwoude, the Netherlands).

Experimental design and procedure
The animal experiment was approved by the Dutch Central Animal Experiment Committee (project license AVD10300 2015-241) and the work protocol (No. 2015-0072-004) was approved by the Animal Experiments Committee of Radboud University according to the legal regulations as stipulated in the amended Animal Testing Act in the Netherlands and Directive 2010/63/EU of the European Parliament and of the Council.
Ten 8-week-old male Wistar rats (~250 g) were used. The experimental material was injected into a tissue cage made of stainless steel wire mesh, left to gelation and then implanted subcutaneously ( Fig. 1) for 4 weeks to obtain histological and exudates samples. Three formulations, (1) P2: 1 mL PIC gel loaded with 75 mg P2 microspheres, (2) P2A: 1 mL PIC gel loaded with 75 mg P2A microspheres, and (3) P2/ P2A: 1 mL PIC gel loaded with a mixture of 37.5 mg P2 and 37.5 mg P2A (mass ratio of 50:50), were tested. Blank cages were implanted as controls.
During the surgery, the animals were anesthetized with isoflurane. After 4 subcutaneous pockets had been created, four cages from different groups were separately inserted into the pockets of each rat according to a computerized random sequence generator (www.random. org). All rats were housed under standard conditions in groups and euthanized on day 28 using CO 2 suffocation. After retrieval of the cages, exudate samples were collected from the cages of 7 randomly selected rats with 18G needles and then stored at −80°C until further cytokine analysis (n = 7). Tissue-covered specimens were retrieved from the remaining 3 rats for histology (n = 3).

Measurement of cytokines in exudates and histological evaluation
Before analysis, the collected exudates were centrifuged at 2000 g at 4°C to isolate supernatants and then tested for inflammatory cytokines with a Bio-Plex Pro™ assay (Bio-Rad, Hemel Hempstead, UK) using a rat cytokine magnetic bead assay kit (Bio-Rad). Seven cytokines were tested, including the chemokine growth-regulated oncogene-KC (GRO/ KC or CXCL1), monocyte chemotactic protein-1 (MCP-1), and cytokines tumor necrosis factor-alpha (TNF-α), interleukin-1α (IL-1α), interleukin-1β (IL-1β), interleukin-6 (IL-6), and interleukin-10 (IL-10). Cytokine concentrations (pg/mL) were determined from fluorescence intensities compared to the standard curve. The samples for histology were first fixed in 10% formalin, dehydrated in a graded series of ethanol and embedded in polymethylmethacrylate (PMMA). Thereafter, samples were split from the mid of the long axis of the cages into two halves and cross-sections of 350 μm perpendicular to the long axis of the cages were prepared using a microtome (Leica Microsystems SP 1600, Nussloch, Germany). Three sections of each specimen were stained with methylene blue and basic fuchsin. The sections were cut from at least three arbitrary chosen regions.

Statistical analysis
For drug release results, a linear mixed model analysis was used to determine the difference of the release patterns. For the other in vitro results, a one-way analysis of variance with Tukey's multiple comparisons was used. For the in vivo cytokine test, different statistical analyses were used. To detect the overall difference between the experimental groups (P2, P2A and P2/P2A) and the cage control group, a multilevel model with random intercepts was used. To detect the differences between each group, either a simple paired t-test only when the cytokine concentrations of all samples could be detected, or a Wilcoxon signedrank test was performed followed by Bonferroni correction for multiple comparisons. The significance level was set at α = 0.05 for the entire set with a Bonferroni correction value to α/n, n representing the number of comparisons. All analyses were done on SPSS 26.0 (IBM, Armonk, NY) or the lme4 library of the package R (version 3.6.1).

Morphology and size distribution of drug-loaded microspheres
Drug-loaded electrosprayed PLGA microspheres exhibited a round shape and smooth surface ( Fig. 2A-D). Drug loaded P2 and P2A microspheres had similar shape and diameter. LXA 4 loaded P2 and P2A microspheres were around 2 μm, while DOX-P2 and DOX-P2A were around 1 μm in size (Fig. 2E). The blank microspheres, which were used in the injectability and rheology test, were observed to have a round form but were bigger in diameter (4-5 μm) than drug-loaded ones (Fig.  S1). The difference in sizes was likely caused by the increased the electrical conductivity of the formulation after the incorporation of drugs: higher electrical conductivity leads to smaller particle size [24].

Injectability and rheology
When P2 or P2A microspheres were dispersed into PIC solution, all formulations appeared visually as a homogenous milky suspension. The freeze-dried gel samples (Fig. 3A) showed that the microspheres were homogenously embedded in the PIC polymer. With the addition of 5% and 10% wt% P2 microspheres, the flow of the mixture was easy and continuous, and the mixtures were squeezed completely out the syringe. The average glide force required to sustain the movement of the plunger for those samples was comparable to that for water and pure PIC solution (Fig. 3B). With the increase of the number of PLGA microspheres in PIC solutions, i.e. 15% and 20% wt% formulations, microspheres sometimes agglomerated and clogged the needle inducing a filter pressing effect. Some microspheres remained in the syringe. The Fig. 1. Implantation experiment. (A) Cages were prepared from surgical-grade stainless steel wire mesh (~1.5 cm in length, 0.8 cm in diameter, 0.25mm wire diameter and 0.8-mm opening width with 58% open area), as previously described [23]. Cages were sterilized by autoclaving. (B) The cage was filled with PIC-PLGA hydrogel. In each formulation, 1 mL 0.5% PIC gel was mixed with 75 mg PLGA microspheres. PIC gel was prepared under sterile conditions with sterile PBS. The PLGA microspheres were sterilized by 25 kGy γ-irradiation (Synergy Health, Ede, The Netherlands). (C) After anesthesia, the backs of each rat were shaved and sterilized with iodine scrub. Then, four 2 cm paravertebral incisions were made. Subcutaneous pockets were created by blunt dissection. Thereafter, cages were implanted subcutaneously. (D) After the implantation, wounds were closed using skin staples (InstruVet C.V., Cuijk, the Netherlands). Four cages were implanted per rat.
samples loaded with P2A (Supplement- Fig. S2) showed similar results as those with P2, indicating that PLGA type had no influence on the injectability of the mixture formulations.
In the rheology test, there were no significant changes on the G' and G" after loading of 10% P2 microspheres into PIC gel (Fig. 3C). The gelation temperature was not significantly affected either. The PIC-PLGA system was still thermo-reversible. The result of P2A was similar to P2 (supplement- Fig. S3).

DOX release
DOX release was assessed for all D1 to D6 formulations ( Fig. 4 A-C).
The plain PIC gel (D6) facilitated DOX release for about 4 days. The 100% P2A (D5) formulation sustained DOX release for about 1 week with less burst release compared to pristine PIC gel. In the first 4 days (Fig. 4B), initial burst release was gradually decreased from D5 (~95%) to D1 (~70%). From D5 to D1, the release rate gradually decreased and resulted in a longer release period. Eventually, D1 released DOX for 6 weeks (Fig. 4C). Comparison of the release data for all formulations indicated that the observed differences were significant (P < .05). Examination of the pH changes in the release medium (Fig. 4D) showed that after 4 days, the pH of all groups containing P2 or P2A particles started to drop, but with a faster rate for the P2A compared to P2. But plain PIC gel group constantly kept pH ≈ 7. These data corroborated with the visual observation: the volume of PIC-PLGA groups  Injectability of different ratios of P2 mixed with 0.5% PIC compared with Milli-Q water and 0.5% PIC solution. Data were expressed as Mean ± SD, n = 3. No statistical differences were found between groups. (C) The rheology measurement demonstrated thermo-reversibility of 0.5% PIC with 10% w/w P2 microspheres compared with plain PIC. The triangles represent the heating procedure and circles represent the cooling procedure. Please note that the difference of the symbol (triangles or circles) density between groups was induced by the different frequencies of data acquisition, not related to the heating/cooling rates.
(D1 to D5) decreased and the erosion rate was positively associated with the amount of P2A (Fig. 4E).

LXA 4 release
LXA 4 was released from the plain PIC gel (L6) for about 4 days, while PLGA prolonged release to 7-10 days depending on the endgroup composition (Fig. 5A, B). Both dominant acid-terminated microsphere formulations (L5 and L4) showed that release of LXA 4 was completed at day 7, and no significant differences were found compared to L6. In contrast, for the dominant ester-capped microsphere  formulations (L1 and L2), and also for the L3 group, the release of LXA 4 lasted up to 10 days, and the difference with L6 was significant. Furthermore, L2 and L3 significantly differed from L1. Notably, the initial burst release was reduced in L1, L2 and L3 compared to L4, L5 and L6. The pH changes in the release medium of LXA 4 (Fig. 5C) followed a similar trend as described above for DOX (Fig. 4D). An obvious pH drop started at day 4, and a higher percentage of acid-terminated PLGA resulted in faster pH decrease. The visual observation indicated similar results as compared to those loaded with DOX (Fig. 5D).

In vitro efficacy test of released DOX and LXA 4 3.4.1. Bioactivity of released DOX
The efficacy of released DOX from group D3 was evaluated through the antibacterial test against Pg, by comparing the diameters of the inhibition zones from freshly prepared DOX versus released DOX on day 3, day 8 and day 14 (Fig. 6A, B). After curve fitting, the diameters of the inhibition zone showed a positive correlation with DOX amount. Furthermore, the released DOX demonstrated a comparable inhibition zone compared to the fresh DOX (Fig. 6C).

Bioactivity of released LXA 4
The efficacy of released LXA 4 from group L3 was evaluated through a phagocytotic activity test (Fig. 7). The quantitative analysis showed that LXA 4 released from PIC-PLGA system significantly enhanced the phagocytotic activity compared to the NC and pristine PIC-PLGA vehicle (Fig. 7A). The efficacy of the released LXA 4 was equivalent to the fresh prepared LXA 4 . Microscope inspection showed that more fluorescent beads were internalized in the released LXA 4 group compared to PIC-PLGA vehicle (Fig. 7B).

Descriptive histology of host tissue response
All incisions were fully healed after 7 days. No signs of adverse tissue response or increased stress behavior of the rats were observed during the observation period. Light microscopical analysis of the empty cages (blank control group) revealed that the cages were surrounded by a thin connective tissue capsule (Fig. 8A). The capsule was connected to the struts of the cage and had the appearance of a suspension bridge (Fig. 8B). A seroma-like structure could be observed in the cavity of the cage, which was characterized by the presence of fibrous tissue that showed a very loose structure, as illustrated by the absence of dense collagen fibers (Fig. 8C). Fibroblasts and (micro)capillaries could be recognized. There was no evidence of an inflammatory response and almost no inflammatory cells were present.
The histological sections of the cages that were filled with the various types of PIC-PLGA gel, showed a striking similarity. The cages were all surrounded by a thin collagen capsule with suspension bridge like connections between the struts of the cage (Fig. 8D). The PIC-gels could always be easily identified inside the center of the cage. The gels had shrank to about 75% of its original size. The resulting seroma-like space was filled small cavities and fibrous tissue (Fig. 8E). Around all remaining PIC-PLGA gels, a layer of inflammatory cells could be observed (Fig. 8F). The surface of the gels seemed to erode and occasionally even small portions of gel were loosened from the gel surface (Fig. 8G). Small pores could be recognized inside the body of the PIC-PLGA gel, which might be formed by the residual aggregations of PLGA particles; infiltration of inflammatory cells was also observed (Fig. 8H). No evident difference in inflammatory response existed between the various PIC-PLGA materials.

Expression of inflammatory cytokines
Regarding the cytokine measurements, IL-6 was not detectable in any sample and therefore excluded. Overall, the blank cages showed the significantly lower level of all cytokines than PIC-PLGA groups (Fig. 9). Both pro-inflammatory (GRO/KC, MCP-1, TNF-α, IL-1α, IL-1β, and IL-10) and anti-inflammatory (IL-10) cytokines were up-regulated by PIC-PLGA groups. However, no significant difference could be found after Bonferroni correction for multiple comparisons was further performed to compare any two groups.

Discussion
Topical administration of anti-infectious and anti-inflammatory medication is regarded essential in the treatment of refractory periodontitis. Although many types of delivery systems have been described before, i.e. micro-or nano-spheres [25], polymer chips [26], fibers [27], and gels [28], there is still lack of efficient ways to apply and retain drugs topically, especially in combination with a tunable period of drug release. In this study, we investigated a thermo-reversible PIC hydrogel mixed with drug-loaded PLGA microspheres to serve as a personalized delivery system. The results showed that this system possessed the desired properties for our envisioned clinical application, such as good injectability, a tunable release of bio-active drugs, and not inducing severe inflammation in vivo.
The first prerequisite for successful gel therapy is the ease of administration. Several of the commercially available gels, such as Periocline® (Sunstar Inc) and ATRIDOX® (Tolmar GmbH), are very viscous and not freely flowable, which makes penetration into deep pockets difficult. The currently used thermo-reversible PIC hydrogel has the advantage of water-like consistency. When slightly cooled, the gel can be easily injected and perfuse into deep and irregular pockets, where gelation will take place immediately in situ upon reaching body temperature. The results showed that loading up to 10% wt% blank PLGA microspheres did not compromise the injectability of the PIC material, but the injectability was hampered when higher amounts of microspheres were loaded. Because blank microspheres are bigger than the drug-loaded ones, they are more prone to evoke clogging of the cannula resulting in filter pressing effect and hampering of the flow [29]. Thus, for the finer DOX or LXA 4 loaded microspheres, similar or even better injectability can be assumed. If a filter pressing effect would be observed, a possible solution could lie in the addition of appropriate surfactants to stabilize microspheres in the dispersion [29,30]. Similarly, the addition of such stabilizers could even remove the requirement of having to mix the microspheres with PIC solution directly before use, as was done herein.
A second requirement is the adequate in-situ retention of drug depots. The hydrogel was first formed at the bottom of an Eppendorf tube. The volume of 200 μL was chosen according to the volume of commercial Emdogain® periodontal gel (Straumann, USA). Thereafter, it was incubated in the medium with agitation, which mimicked the microflow in the periodontal pocket. It is worth mentioning that the material inside the periodontal pocket will always experience mechanical forces from mastication, constant turnover of epithelium and gingival crevicular fluid [31], which was not attainable in the in vitro model. Several thermosensitive hydrogel products, mostly based on poloxamers, have previously been developed for periodontal applications [32,33]. Poloxamer gel has much higher storage modulus than PIC gel. However, the structural integrity of hydrogels is not directly related to the value of G'. Based on our previous structural stability test, poloxamer gel was found to disintegrate totally in just a few days [7]; in contrast, the rigid polyisocyanide backbone of PIC gives rise to fibrous semiflexible networks, which maintain structural integrity over weeks [7]. Op 't Veld et al. applied PIC gel on mice for dermal wound dressing and traced the gel by radiolabeling. They found that PIC gel stayed in situ for at least one week, which further supported the structural integrity of PIC [9].
After loading 10% microspheres into PIC gel, the rheology results suggest that the blank microspheres with a relatively larger size did not significantly influence the mechanical properties, nor did they affect the gelation temperature. It indicates that the rheological property of the mixture is determined only by the PIC gel and not affected by the addition of PLGA microspheres by physical mixing. In the drug release test, it was observed that the PIC-PLGA gel was gradually eroded. To investigate the effect of PLGA degradation on the stability of the PIC gel, we carried out additional rheology experiments at pH = 3 (hydrogen chloride, the lowest pH tested in the drug release experiment) and in the presence of the ultimate PLGA degradation products, lactic and glycolic acid (LA/GA, 5% wt each) (Fig. S4). For the freshly prepared PIC gels, the low pH has no effect, whereas an increased transition temperature and a decreased storage modulus G' were observed (0.8 vs 0.27 kPa) in the presence of LA/GA. This phenomenon is in line with the well-known Hofmeister effect [34,35]. After incubation at 37°C for 3 days, the gel in acidic solution slowly and heterogeneously disassembled, indicated by the broad transition, which is attributed to the acid-induced disruption of the hydrogen bonds that stabilize the helical polymer. However, the gel in the presence of LA/GA aged much faster at 37°C (G' = 0.004 kPa), due to the combination of the low pH and the aforementioned Hofmeister effect. It has to be noticed that in an in vivo environment, this erosion process will probably proceed slower because the acidic degradation byproducts will be removed constantly by body or gingival crevicular fluid [36]. Nevertheless, a clinically relevant retention period of the PIC-PLGA gel inside periodontal pockets should be further explored.
A third aim was to facilitate a tunable and sustained drug release. Acid-terminated PLGA degrades faster than ester-capped PLGA, which can be ascribed to both its higher hydrophilicity as well as to the more speedy accumulation of acid in the polymers [13]. As a result, acidterminated PLGA facilitates a much faster drug release [13,37]. Our previous study reported that the acid-terminated PLGA microspheres alone facilitated DOX release for 1w, while ester-capped microspheres displayed triphasic release, i.e. an initial burst phase (0-7d), a lag phase (7-20d) and an accelerated release phase(20d~6w) [13]. In the current study, due to the PIC gel as the secondary encapsulation vehicle, the lag phase was transformed into a sustained release. Furthermore, the PIC-PLGA delivery system was found to have the capacity to manipulate the range of the DOX release period from 1 week to 6 weeks, by simply decreasing the amount of P2A and increasing the amount of P2 microspheres. When more ester end-capped PLGA is used, release will be slower and longer, and vice versa. Thus, in a future practical application, the DOX dose and release period can be tailored based on the individual characteristics of patients. Still, all formulations displayed an initial burst release in the first 4 days, even though for composition D1, this burst release was~25% less than D5. However, it can also be argued that such a burst release is necessary for any type of patient, regardless of their characteristics, in order to achieve an initial bactericidal effect. Thereafter, the long-term sustained DOX release can be adjusted. When the DOX level in gingival crevicular fluid remained well above the minimum inhibitory concentration for periodontal pathogens, it will have a constant antimicrobial effect. On the other hand, its level can be tuned to the sub-antimicrobial dose, which has been proven to have an anti-inflammatory effect as the adjunctive treatment for periodontitis in large randomized clinical trials [16,38]. Many studies suggested the sustained release of doxycycline (or tetracycline or minocycline) did not induce resistance among the normal flora or among periodontal pathogens [39][40][41][42] or only a transient increase in resistant bacteria with no permanent change to the microbiota [43,44]. Still, bearing in mind the global rise of antimicrobial resistance, the usage of antibiotics should only be limited for clear indications. The PIC-PLGA system can be tailored later according to the clinical need.
The effect of the PLGA end groups influenced the LXA 4 release in a similar manner as for DOX. However, the effect size seemed much smaller. To achieve an extension of the release of LXA 4 , the mass ratio of P2 microspheres has to be ≥50%. This observation corroborates with other studies. For example, Reis et al. encapsulated LXA 4 in PLGA microspheres by single emulsion, which resulted in a controlled release of the LXA 4 over two days [45]. Likewise, E. Cianci et al. reported that Fig. 8. Histological sections of the cages after 4weeks of subcutaneous implantation (methylene blue and basic fuchsin staining). Fig. A provides a low magnification overview of blank control cages. The cage is surrounded by a fibrous capsule (yellow arrows) with a suspension bridge-like configuration. The black dots are the struts of the cage. The inner space of the cage, shown at higher magnification in fig. B and C, was filled with loose structured fibrous tissue and capillaries (as indicated by red arrows). Fig. D shows a low magnification overview of a cage filled with PIC-PLGA gel. The gel did shrink during implantation and the created space became filled with fibrous tissue. This was confirmed at higher magnification (Fig. E), which also shows that the remaining PIC-PLGA gel was surrounded by a fibrous capsule. Frequently, a thin layer of macrophages/ giant cells (green arrows) were present at the interface between fibrous tissue and PIC-PLGA gel (Fig.  F). Occasionally, the fibrous tissue contained adipose cells (#). Further, it was noticed that the gels often eroded (Fig. G) and small portions of gels (black arrows) were loosened from the gel surface. Higher magnification (Fig. H) confirmed that small pores (*) were formed inside the body of the PIC-PLGA gel, which might be the residual aggregations of PLGA particles. Inflammatory cells were also seen inside the gel (green arrows in Fig. G and H). Scale bar: 100 μm. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) electrospun poly(ethylene oxide) and poly(D,L-lactide) membranes facilitated the release of LXA 4 for 4 days [46]. We prolonged the release period up to 10 days in the current study. The two-layer encapsulation by the PIC-PLGA vehicle may synergistically contribute to the more sustained release. It is important to note that the clinical effective amount of LXA 4 to treat periodontitis still remains to be determined, which should be done next before further studies on loading, release, and translation can be regarded. Thereafter, the loading dose of LXA 4 in the PLGA microspheres can be adjusted based on the therapeutic requirement. Nevertheless, adjusting the loaded mass ratio of acid-and ester-capped PLGA microspheres in the PIC gels can be used to control the release of LXA 4 to a limited extent. Moreover, all gel carriers maintained integrity longer than the actual drug release period in all formulations, meaning that the carrier system is stable enough for longterm drug release. After release, the antibacterial and phagocytosis tests further confirmed the bioactivity of the released DOX and LXA 4 .
Finally, it is important that the vehicle material does not evoke an obvious inflammatory response. For the in vivo application of PLGA materials, the local accumulation of acidic degradation products (i.e., lactic acid and glycolic acid) is the main cause of inflammation [47]. A rat subcutaneous implantation model has long been used to explore the inflammatory response of the dental biomaterials in an early stage of translational research [48]. This model is well-developed and standardized. Also, it does not cause much discomfort for animals. We found that both pro-inflammatory and anti-inflammatory cytokines were upregulated by PIC-PLGA groups compared to empty cages. The empty cage did not evoke an inflammatory response. The multiple comparison test did not find any statistical differences between any two groups, indicating the inflammation induced by the PIC-PLGA groups was not severe. The histological observation corroborated with the cytokine measurements. PIC-PLGA gel remnants surrounded or infiltrated by only a layer of inflammatory cells were observed, which explains the presence of inflammatory cytokines. In comparison, in an earlier comparable cage model study involving PLGA/PCL electrospun nanofiber membranes [23], we observed a pronounced inflammatory response. A likely explanation for the limited inflammatory response of the PIC-PLGA vehicle is that the inclusion of PLGA microspheres in a very highwater content like the PIC hydrogel in combination with the presence of the body fluid flow results in an enhanced dilution and drainage of the acidic degradation products. Finally, the limited shrinkage or erosion of the PIC-PLGA after 28 days of implantation corroborates with the in vitro study after 28 days. But the erosion rate was less compared with the in vitro results again due to faster drainage of the acidic degradation products.
Although promising results have been acquired in this study, the in vitro evaluation is still far different from the in vivo situation, especially with the special periodontium. This study mainly explored and discussed the tunable drug release model, which composed the early stage of the translation. Herein, the limitations of this study should be addressed. During electrospraying, using water bath to collect the eletrosprayed microspheres could prevent particle aggregation but unavoidably resulted in the drug loss. By reducing the collecting period and quick freeze-drying could decrease the loss of unencapsulated drug to some extent. Future studies can explore the effect of using different collecting methods or surfactants to minimize drug loss. Besides, the rat subcutaneous implantation model could only provide information on the inflammatory response of the PIC-PLGA vehicle. Since the periodontium is a peculiar tissue, a reliable in vivo periodontitis model is of great value to the evaluation of the actual feasibility, retention, drug release, and efficacy of this delivery system. The previous study showed that PIC hydrogel labeled with Indium-111 could be traced by singlephoton emission computed tomography [9]. This technique can be used to non-invasively monitor the distribution of PIC gel in the periodontal pockets after topical application. Moreover, the concentration of drugs in the gingival crevicular fluid could be detected to reflect the real-time drug release. The carrier retention and drug release should be, therefore, tailored and further optimized. For the drug efficacy, whether the long-term low-dose release is sufficient to meet the diverse needs of periodontitis patients should be investigated in the future study. Accordingly, the loading amount of the drug in the PLGA microspheres could be adjusted.  9. The concentrations of detected cytokines in the exudates after 4 weeks in all different groups. Empty cage was used as blank control. Data were presented as 0 when below the detection limit. Data are shown as the median with 25 to 75 percentiles (Boxes) and 5 and 95 percentiles (Whiskers). No significant difference was detected when compared every two groups separately. However, the overall effect of P2, P2A and P2/P2A groups showed significant higher level of all cytokines compared to blank cage group. *, P < .05; ***, P < .001.

Conclusion
In the present study, we have designed and formulated a tunable and injectable local delivery system by loading the PLGA-drug microspheres into the PIC hydrogel. This system exhibited appropriate injectability, long-term structural stability, and no obvious in vivo inflammatory response. Furthermore, the release of drugs could be controlled by adjusting the loaded mass ratio of acid-and ester-end capped PLGA microspheres. As evidenced in these results, we envision that upon further optimization, this novel and simple thermo-responsive PLGA-PIC carrier system may have the potential to be translated as an effective therapy for penalized medicine in periodontal clinics.