Delivery and diffusion of retinal in dermis and epidermis through the combination of prodrug nanoparticles and detachable dissolvable microneedles

To minimize chemical degradation of retinal, we graft this aldehyde on chitosan chains to make them self-assemble into pro-retinal nanoparticles (PRNs), which we then load into detachable dissolvable microneedles (DDMNs) made of 1:1 (by weight) hyaluronic acid/maltose. The presence of PRNs in the hyaluronic acid-maltose needle matrix also helps improve the microneedles’ mechanical strength. Ex vivo administration of PRN-loaded DDMNs on fresh porcine ear skin shows, as observed by stereomicroscopic and confocal fluorescence microscopic analyses of the cross-sectioned tissue pieces, complete deposition followed by dissolution of the needles and diffusion of the PRNs in epidermis and dermis. Rats administered with a single dose of PRN-loaded DDMNs show significantly increased epidermal thickness as compared to rats administered with control DDMNs (no PRN). Both the PRN-loaded DDMNs and the control DDMNs produce no skin irritation in rats.


Introduction
Retinoids are a group of compounds that include vitamin A or retinoic acid and its derivatives such as retinol, retinal, retinyl palmitate, and retinyl acetate. Retinoids are important modulators of immune response and essential regulators of cell proliferation, differentiation, and morphogenesis during skin development [1][2][3]. At the epidermis layer, retinoids regulate epidermal keratinocyte growth and differentiation [4,5]. In the dermis layer, they not only stimulate the growth and differentiation of fibroblast and endothelial cells but also regulate various functions of these cells, resulting in a reduction in matrix metalloproteinase and an increase in newly synthesized collagenous matrix in dermal tissue [6]. These functions make retinoids effective anti-aging compounds, capable of exerting their anti-aging effects in both the dermis and the epidermis [7]. In addition, retinoids have been used in the treatment of numerous dermatological conditions, including acne [8], psoriasis [9], ichthyosis [10], and actinic keratosis [11].
Nonetheless, retinoid topical applications can cause skin irritation [12,13]. Skin irritation from retinoids occurs at the epidermis through an induction of excess keratinocyte which results in epidermal slough [12][13][14], and an increased production of interleukin-1 in growth-stimulated keratinocytes, which results in proinflammatory effects [6]. These epidermal effects result in skin erythema, exfoliation, dryness, burning, scaling, and even alopecia [15]. Because skin irritation from retinoids is dose-dependent [16], a common strategy to minimize those adverse effects is to use retinoids at low concentrations. Nevertheless, low concentrations of retinoids in topical formulations directly induce low retinoid flux into the dermis, producing ineffective anti-aging effects on the dermal fibroblasts. Research has shown that topical application of vitamin A can provide only around 5% bioavailability to skin tissue, and most of the 5% accumulates in the epidermis [17,18].
Another problem for retinoids is their chemical instability in cosmetic and pharmaceutical formulations, with up to 80% and 100% degradation after 6 months when kept at 25 °C and 40 °C, respectively. Retinoid stability is formulation-dependent, wherein it is not dependent on retinoid concentration but dependent on the type of retinoid derivative [19]. Encapsulation of retinoids into polymeric nano-and microcarriers can help increase their stability [20][21][22][23]. Nevertheless, these carriers cannot effectively deliver retinoids to the dermis.
Dissolvable microneedles (DMNs) are arrays of 100-1000-mm-high needles made of materials that are biocompatible and dissolvable in skin tissue. DMNs can be fabricated with drugs embedded in the needles [24,25]. When pressed against skin, DMNs are expected to penetrate skin tissue and be dissolved by interstitial fluid. Dissolution of needles automatically releases loaded drugs into the tissue [26][27][28]. A recent study has confirmed the long-term safety of repeated use of DMNs in human volunteers [29]. Improved retinoid stability by embedding in DMNs has been demonstrated [30]. Effective delivery of retinoid to dermis by DMNs has also been reported [31]. The use of retinoidloaded DMNs for the treatment of seborrheic keratosis or senile lentigo has been demonstrated [32,33].
Here we demonstrate the use of prodrug nanoparticles and detachable DMNs to stabilize and deliver retinal into skin tissue. The study includes pro-retinal nanoparticle preparation, material optimization for DMN fabrication, physical characterization of the pro-retinal nanoparticle-loaded DMNs, and chemical stability of retinoid. The work also covers experiments on DMN dissolution and retinoid diffusion in ex vivo skin tissues and the in vivo effects of proretinal nanoparticle-loaded DMNs on rat skin.

Preparation of pro-retinal nanoparticles (PRNs)
Pro-retinal nanoparticles (PRNs) were prepared by grafting retinal (Sigma-Aldrich) onto chitosan (CS, Mw of 40,000-50,000 Da, Taming Enterprise, Samut-Sakhon, Thailand). The grafting was carried out using our previously reported protocol [34]. In brief, CS was dissolved in 0.1% acetic acid, and the pH of the obtained solution was adjusted to 5.9 using an aqueous NaOH solution. The final chitosan solution contained 45 mg of CS in 19.0 mL solution. The grafting reaction was initiated by slowly adding retinal (15 mg in 1.0 mL of ethanol) dropwise into the two-neck round-bottom flask containing aqueous CS suspension at 5 °C, under ultrasonic (40 kHz), light-proof, and N 2 atmospheric conditions. After stirring for 2 h, the pH of the solution was adjusted to 6.8 using an aqueous NaOH solution, and a PRN suspension was obtained. The particle size of PRN was examined by scanning electron microscopic analysis (SEM, JSM-6400; JEOL, Tokyo, Japan).
PRN-loaded microneedles were prepared by dissolving HA (40 mg) and maltose (40 mg) in 1 ml of PRN suspension (3 mg CS, 1 mg retinal, and 1 mL of 7% ethanol in water). The obtained mixture was poured into the mold and let dry in the light-proof and moisture-controlled (~5%) chamber. Retinal-loaded DMNs were prepared by dissolving HA (40 mg) and maltose (40 mg) into 1 ml of retinal solution (1 mg retinal in 1 mL of 7% ethanol in water). The obtained mixture was poured into the mold and let dry. The morphology of the obtained microneedles was examined under a stereomicroscope (Olympus DP22, Japan). Distributions of retinal or PRNs in the DMN patches were acquired by taking confocal fluorescence images (confocal fluorescence microscope, Eclipse, Ti series microscope, Nikon, Japan) of each DMN patch at all layers and then stacking the obtained images into three-dimensional images.
The mechanical properties of the obtained DMNs were studied using the universal testing machine (UTM, Shimadzu EZ-S, Shimadzu Corporation, Tokyo, Japan). To carry out the testing, the base pad of the DMN patch was attached to an acrylic plate (needles facing upward). The acrylic plate with the DMN was put into the UTM. The maximum compressive force was set at 100 N. During the measurement, the needle side was compressed, and the displaced distance was recorded along with the force.

Detachable dissolvable microneedles for ex vivo and in vivo experiments
A PRN-loaded detachable dissolvable microneedle (DDMN) patch was prepared by dissolving HA (40 mg) and maltose (40 mg) in 1 ml of the PRN suspension (3 mg CS, 1 mg retinal, in 7% ethanol in water). The obtained mixture (0.05 mL) was poured into a silicone mold, and the filled mold was put in a light-proof and moisture-controlled (5%) chamber until dry. After that, 0.05 ml of the viscous polymer solution (without PRN) was dropped onto the dry microneedle array (in the silicone mold) and the water penetrable lint-free sheet (600-μm-thick non-woven polyester sheet) was attached; then, the assembled piece was put in the light-proof and moisture-controlled (2%) chamber until completely dry. The retinal-loaded DDMN patch was prepared similarly, except that the freshly made retinal solution was used in place of the PRN suspension. One milligram of retinal was dissolved in 1 mL of 7% ethanol in water to make retinal solution. The unloaded DDMN patch was prepared similarly, except that water was used in place of the PRN suspension. For all ex vivo experiments, DDMN patches were 8 × 8-mm square patches containing an array of 10 × 10 needles of the tetragonal pyramidal shape with 300 × 300 μm needle base, 650 μm needle height, and 500 μm tip-to-tip distance. DDMN patches used for in vivo skin irritation experiments were prepared similarly with an adjusted amount of PRN to give 8 μg of retinal per patch, and the patch was 8 × 8 mm 2 square containing an array of 10 × 10 needles of the tetragonal pyramidal shape with 100 × 100 μm base, 150 μm needle height, and 500 μm tipto-tip distance.

Retinal stability
The chemical stability of retinylidene moieties in PRNloaded DMNs (50 μg retinal per patch) and in PRN suspension (1 mg/mL of retinal in the suspension), and retinal in retinal-loaded DMNs (50 μg retinal per patch) and in retinal solution (1 mg/mL retinal in 7% (v/v) aqueous ethanol) were examined.
Each DMN sample was kept in its own light-proof package at 4, 25, 40, and 50 °C. At each time point (0, 7, 14, 30, 60, and 90 days), the DMN sample was taken out and subjected to retinal extraction and quantification. To extract retinal from the retinal-loaded DMN patch or PRN-loaded DMN patch, the DMN patch was put in 2 mL of 0.0025 M aqueous HCl under an N 2 atmosphere and light-proof conditions for 10 min at room temperature. The obtained solution was partitioned with ethyl acetate (3 times with 2 mL of ethyl acetate each time). The collected ethyl acetate extract was quantified for retinal by UV-vis spectrophotometry using a maximum wavelength of 370 nm with the aid of a calibration curve of freshly prepared retinal standards. Retinal solution and PRN suspension were kept in an individual vial under light-proof condition at the same temperatures for the same durations. At each time point, the liquid sample was subjected to retinal extraction and retinal quantification with the same protocol.

Ex vivo drug diffusion in skin tissue
Fresh porcine ears from a local slaughterhouse (Lopburi, Thailand) were cleaned with a citrate buffer (0.1 M, pH 7.4). Hairs were shaved, and the shaved ear was rinsed with water and patted with tissue paper. A DDMN sample (650 μm needle height) was pressed against the porcine ear skin, then 1 drop of water was applied, and the patch was pressed for another 1 min. After that, the backing sheet was removed. The ear was kept in a closed petri dish lined with PBS buffer-soaked paper towel (to keep the tissue moist), with the subcutaneous side touching the paper towel, for 0, 1, and 4 h. At each time point, the full thickness of skin was surgically sectioned, and the tissue section was examined under a stereomicroscope and a confocal fluorescence microscope (l ex/emit 488/525 nm, and Eclipse, Ti series microscope, Nikon, Japan).

In vivo skin irritation
The study protocol was approved by the committee on the use of animals for the scientific purpose of Chulalongkorn University animal care (protocol no. 1873021). Eight male rats (Wistar rat, 10-12 weeks, 300-400 g, Nomura Siam International Co., Ltd., Thailand) were housed at Chulalongkorn University Laboratory Animal Center, in an isolated clean room held at 25 ± 2 °C with a relative humidity of 65-75%. The rats were acclimated for 3 weeks before the experiment. Rats were divided into two groups of four rats each, 24-h and 7-day groups.
To start the treatment, rats were anaesthetized and shaved at the treated area. Each rat was treated with three PRNloaded DDMN patches (equivalent to 8 μg retinal per patch) on the right dorsal skin, and three unloaded DDMN patches on the left dorsal skin. A DDMN patch was pressed against the skin using pressure from fingertips for 1 min; then, one drop of water was dropped over the backing of the DDMN patch, and the patch was pressed for another 1 min. After that, the base sheet of the DDMN patch was taken out. The treated area was observed at 1, 6, 12, and 24 h and every day until day 7 post-administration. Full-thickness skin of the test area was sampled at 24 h and 7 days post the single administration. Skin biopsies were carried out on all six applied sites immediately after the sacrifice of the animal. The biopsied skin tissue was fixed in a 10% formalin buffer and processed using routine histopathological techniques. Each biopsied skin tissue was sectioned to obtain at least 3 skin section pieces. The skin sections were evaluated by the Draize scoring system. Erythema and edema were scored as follows: 0 indicates no erythema or edema; 1 indicates very slight erythema and/or barely perceptible edema; 2 indicates well-defined erythema and/or slight edema; 3 indicates moderate-to-severe erythema or moderate edema; and 4 indicates severe erythema and/or edema.
The thickness of the epidermis was measured from digitally expanded images of the cross-sectioned skin tissue pieces. The samples include (1) control group at 24 h postapplication (unloaded DDMN patch), (2) control group at 7 days post-application (unloaded DDMN patch), (3) treated group at 24 h post-application (PRN-loaded DDMN patch), and (4) treated group at 7 days post-application (PRN-loaded DDMN patch). There were 12 application sites for each group (3 sites/animal and 4 animals/group). Three measurements were carried out for each tissue section. The total number of measurements was 36 measurements/group.

Statistical analysis
Difference between epidermal thickness of rats from the unloaded-DDMN-treated group and the PRN-loaded DDMN-treated group was analyzed by unpaired t-test using GraphPad Prism version 9.1.1 (225) software (GraphPad, USA). The p value of < 0.001 was used to indicate a significant difference.

DMNs and mechanical property
The 8 × 8 mm 2 square DMN patch containing 10 × 10 square pyramidal needles (300 × 300 μm square base, 650 μm needle height, and 500 μm tip-to-tip distance) could be successfully fabricated from either HA alone or from the binary composites of HA with each of the following materials, PVP, PVA, CMC, sericin, or maltose. The selection of the square pyramidal-shaped needle over conical-shaped needle was based on the previously reported superior skin penetration [36]. The square pyramidal-shaped needle possesses higher drug entrapment volume as compared to the trigonal pyramidal-shaped needle [37].
Among HA, PVP, PVA, CMC, sericin, and maltose, HA is the only material found naturally in skin tissue and can be degraded in the body by hyaluronidase [38]. Therefore, for maximum skin compatibility, safety, and cosmetic benefits, HA is the most desirable structural needle material. However, DMNs made of pure non-cross-linked HA bent easily and showed a large displaced distance against compressive force (Fig. 1A-1, B-1, red line, D-1). To improve the mechanical strength, we explored the binary composites of HA with PVP, PVA, CMC, sericin, and maltose. The second component was used at 25% and 50% (w/w) of HA because we wanted at least 50% of HA as the needle material (for cosmetic benefits). Among the five materials mixed with HA, CMC showed no improvement in mechanical strength when added at either 25 or 50% ( Fig. 1A-1, A-2, and F-1). Significant improvement in hardness could be achieved by adding PVA at 25% or 50%, or sericin at 25%, or PVP at 50%, or maltose at 50%. DMNs made of these composites showed improved hardness, i.e., smaller displaced distances against compressive force (steeper initial slope in graphs A-1, A-2 of Fig. 1). Although the 25% sericin formulation showed good hardness, obvious skin irritation was observed. DMNs made of 1:1 maltose/HA and 1:1 PVP/HA could withstand up to 50 N of force (Fig. 1A-2). These two formulations were, therefore, selected for the next experiment.

Pro-retinal nanoparticle-loaded microneedles
Pro-retinal nanoparticles or PRNs were prepared as previously described by grafting retinal onto chitosan polymer via an imine linkage and allowing the obtained retinylidene chitosan to self-assemble into water-dispersible particles [34]. Our preparation gave 200-300-nm yellowish PRNs that easily and stably dispersed in water ( Fig. 2A). The loading of retinal in the particles was 212 mg of retinal/g of PRNs.
The obtained PRNs were loaded into the 1:1 PVP-HA-DMNs (DMNs with 50% PVP and 50% HA as needle material) and the 1:1 maltose-HA-DMNs (DMNs with 50% maltose and 50% HA as needle material). Unexpectedly, when the PRNs were mixed with the PVP-HA polymer mixture, aggregation of PRNs took place immediately. As a result, only PRN-loaded 1:1 maltose-HA-DMN was successfully fabricated (Fig. 2B1-D1). The loading content was 59 mg PRN/g of the obtained DMNs. Each 0.36 cm 2 DMN patch contained 0.24 mg of PRN in the needles, which corresponded to 50 μg of retinal in each patch. When observed under stereomicroscope (Fig. 2B1, C1), the PRN-loaded DMNs appeared yellowish and had an even distribution of PRNs in the DMNs (Fig. 2B1, C1). DMNs loaded with free retinal were also prepared at the same retinal loading (Fig. 2B2-D2). The free retinal was found to be well-distributed in the DMNs (Fig. 2B2, C2). Although the PRN-loaded DMN patch and the retinal-loaded DMN patch contained the same amount of retinal, the former gave stronger fluorescence at λ ex /λ em of 488/525 nm. We speculate that interactions between retinylidene moieties and rigid self-assembled chitosan polymeric networks likely limit the rotational and vibrational freedom of the retinylidene moieties. As a result, non-radiative vibrational relaxation of excited retinylidene moieties is less probable compared to that of free retinal. Less non-radiative vibrational relaxation results in stronger fluorescence of the PRNs as compared to that of the free retinal.
The PRN-loaded DMNs, free retinal-loaded DMNs, and the unloaded DMNs (made of 1:1 HA/maltose) showed similar rigidity of the needles (similar slope, Fig. 1A-3). The three DMNs, however, showed small differences in failure forces: 75, 60, and 55 N for the PRN-loaded DMNs, the free retinal-loaded DMNs, and the unloaded DMNs, respectively. Because both PRN and retinal are immiscible with the water-soluble polymer matrix, they likely act as fillers for the polymeric matrix. The presence of filler in the polymer matrix increases the mechanical strength of the needles.

Stability of retinal
We previously reported that the stability of retinal could be improved by converting it into PRN [7]. Nevertheless, this improvement is not enough for liquid formulations of consumer products for which a long shelf life is needed. We hypothesized that by embedding PRNs in solid polymers of microneedles, the stability of the retinoid moieties could be further improved. Experimental results, indeed, confirmed our hypothesis. Comparing retinoid stability between PRNloaded DMNs and retinal-loaded DMNs (Fig. 3A, B), and PRN suspension versus retinal solution (Fig. 3C, D), it is obvious that the self-assembled PRNs play an important part in stability improvement. When comparing between PRN-loaded DMNs and PRN suspension (Fig. 3A, C), the improvement from the solid matrix encompassing the PRNs can be seen clearly. Our mechanistic explanations for the great improvement in stability of the retinylidene moieties of microneedles' embedded PRNs are (1) limited access of oxidizing species to the retinoid moieties inside the PRNs; (2) limited chemical reaction in solid DMNs compared to that in liquid medium; and (3) abundance of maltose, the compound with reducing property, in the needle matrix.

Ex vivo drug diffusion
We monitored the retinoid diffusion from PRN-loaded DDMNs and retinal-loaded DDMNs in the ex vivo porcine ear skin tissue. It should be noted here that the 650-μm PRN-loaded DDMNs were used in this experiment so that PRNs could be deposited into both the epidermis and dermis of the porcine ear skin. The thicknesses of the stratum corneum, epidermis, and dermis of pig ear skin are around 20, 70, and 1500 μm, respectively [39]. A DDMN patch was administered to the skin, and the skin was cross-sectioned at various times postadministration. The cross-sectioned tissue was analyzed under a stereomicroscope and a confocal fluorescence microscope. The yellow color of retinal and PRN was used to track their whereabouts in the cross-sectioned tissue (Fig. 4A, B). The result showed clear embedment of arrays of microneedles in the epidermis and dermis (Fig. 4A1,  B1). At 0 h post the DDMN administration, the shape and depth (distance from the stratum corneum side of the skin tissue) of yellow tissue roughly matched the microneedle dimension (300 × 300 μm square base with 650 μm needle height). After 1 h, a strong yellow color at the originally embedded sites could still be observed, but some paleyellow color at the nearby location was also observed (Fig. 4A3, B3). This indicated that diffusion of PRNs and retinal from the originally embedded sites had already been taking place at 1 h post-DDMN administration. At 4 h post-administration, a more homogeneous distribution of yellow color in the epidermis and dermis was observed Retinoid diffusion was also monitored using a fluorescence signal at 488/510 nm λ ex /λ em . Conforming to the stereomicroscopic images, skin administered with PRN-loaded DDMNs showed detectable fluorescence signals in both the epidermis and dermis (Fig. 4C-E). However, no very clear signal was detected in the tissue administered with retinal-loaded DDMNs. Free retinal, as compared to PRN, likely possesses higher non-radiative vibrational relaxation, which correlates to less fluorescence emission. The obviously high intensity fluorescence spots in the fluorescence images indicate the presence of PRNs and agglomerated PRNs. This implies that not all retinal moieties were released from PRNs at 4 h post-DDMN administration. Low-intensity diffuse fluorescence signals in the tissue point toward the release of some retinal molecules from the PRNs. These results indicate slow release of the retinal from PRNs in the porcine skin tissue.
It has been known that skin irritation from retinoids is dose-dependent and takes place in the epidermis [12][13][14]. We hypothesized that sustained release of retinal from PRNs and the ability to deliver retinoid into the dermis via DDMNs could alleviate the retinoid's skin irritation side effects. As a result, we next tested for skin irritation of the PRN-loaded DDMNs on rats.

Rat skin irritation
The PRN-loaded DDMNs and the unloaded DDMNs were tested for skin irritation on the dorsal skin of rats. Neither erythema, edema, nor any other signs of irritation were observed during the 7-day monitoring period (Fig. 5). On day 7, post-single administration, the group applied with PRN-loaded DDMNs possessed healthy skin with lighter color at the administration site. Images of cross-sectioned histological tissue (hematoxylin and eosin-stained) also revealed healthy skin tissue in both groups. Both the It should be noted here that PRN content in each patch was equivalent to 8 μg retinal per 0.64-cm 2 patch. Therefore, the administration dose was 12.5 μg retinal per cm 2 . For the use of common topical retinoid formulations, applying 0.1 mL of 0.01% retinoid cream to a 1-cm 2 area is equal to applying 10 μg retinoid per cm 2 . The result of no skin irritation in all rats administered with PRN-loaded DDMNs confirms our hypothesis that sustained release of the retinal from PRNs and the ability to deliver retinoid into the dermis via DDMNs could solve skin irritation side effects of retinoid. Interestingly, epidermis of the rat group applied with PRN-loaded DDMNs showed significantly increased thickness at day 7 as compared to that of the controlled group. The result showed average epidermal thicknesses of 62.10 ± 11.25 mm and 43.38 ± 9.65 mm for the PRN-loaded DDMNs and the unloaded DDMNs, respectively (Fig. 5C2, D2). This indicates that although negative irritation reaction was not observed, the retinal delivered into the skin via PRNloaded DDMNs could effectively stimulate epidermal proliferation.

Conclusion
Here, we have demonstrated that microneedles fabricated from the mixture of HA and maltose at a 1:1 weight ratio possess enough rigidity to penetrate porcine and rat skin. We have significantly improved the chemical stability of retinal by (1) grafting retinal onto chitosan chains to make them self-assemble into pro-retinal nanoparticles (PRNs) and (2) loading the PRNs into the HA-maltose microneedles. The PRN-loaded microneedle experiments on ex vivo porcine skin have given us images of skin tissues with evidence that indicates (1) deposition of PRNs in the epidermis and dermis and (2) sustained release and diffusion of retinal from the deposited PRNs in skin tissue. Lastly, in vivo administration of PRN-loaded microneedles on rat dorsal skin has resulted in observable stimulation of epidermal proliferation with no skin irritation.