Dissolving microneedle rollers for rapid transdermal drug delivery

Dissolving microneedle patch (DMNP) is a minimally invasive and painless self-administration device. However, due to skin deformation, it is difficult to apply it on the large areas of skin or curved skin as the patch size increased for DMNP. Here, we propose a polyvinyl alcohol (PVA)–based dissolving microneedle roller (DMNR) device that can be used for delivering drugs rapidly on the large surface areas or curved skin and does not need to be attached on the skin all the time during drug delivery. The hypoglycemic effect of insulin-loaded DMNRs for transdermal delivery of insulin was studied on the type 1 diabetic rat models. It was found that the insulin-loaded DMNR has an immediate and effective hypoglycemic effect that the blood glucose level reduced below to 50% of original blood glucose at 1 h after DMNRs administrated.


Introduction
Transdermal drug delivery offers a noninvasive and good patient compliance route compared with subcutaneous (SC) injection, and avoids some other issues via oral route, such as significant first-pass effect of the liver and drug degradation within the gastrointestinal tract [1][2][3]. However, some biomacromolecule drugs, such as insulin, DNA or vaccine, cannot be delivered into skin effectively by this conventional route because of the strong barrier function of the outermost stratum corneum of skin [4][5][6]. To overcome the stratum corneum limitation, microneedle (MN), one of minimally invasive and painless self-administration drug delivery devices, was introduced to transdermal drug delivery [7]. MN is a type of micron-scale needle with diameter in the size of microns and height up to 1000 μm, which is designed to increase the permeability of drugs to subcutaneous layer of skin by piercing stratum corneum and creating recoverable microchannel on the skin but without stimulating nerves [8][9][10]. Besides, MNs have also been developed for increasing endovascular drug delivery recently [11].
A variety of MNs have been reported in recent years, and they can be roughly categorized into five groups [12] namely solid MNs [13], coated MNs [14], dissolving MNs [15], swellable MNs [16], and hollow MNs [17]. Among these types, dissolving MNs (DMNs) have gained widely attention to deliver protein drugs [18] or for cosmetic applications [19]. Besides, DMNs are mainly prepared by some biodegradable water-soluble polymer materials so that the MNs can be quickly and completely dissolved in the skin after contacting with the skin interstitial fluid. Currently, several dissolvable materials are used to fabricate DMNs, such as polyvinylpyrrolidone (PVP) [20], sodium hyaluronate (SH) [21], carboxymethyl cellulose (CMC) [22], and polyvinyl alcohol (PVA) [23]. DMNs are generally arranged on a small patch with a thickness to forming a device called dissolving microneedle patch (DMNP) [24,25]. However, if the size of DMNP is too large, it will be difficult for DMNP to deliver drugs rapidly and efficiently on a large area of skin or curved skin [26] due to skin deformation [27]. Besides, if the density of the microneedles arranged on the patch is too high, it will also influence the insertion ability of MN patch [28], which is called "bed-of-needles" effect [29]. The reason for this phenomenon is that as the density of microneedles increases, the space for skin deformation becomes smaller. That is to say, in the case of same pressure applied to the microneedles, the skin is not easy to be pierced as the density of microneedles increases [30].
Alternatively, microneedle roller (MNR) offers a drug delivery approach over a large area of skin or curved skin surface because of their special geometry [31,32]. In this scheme, MNs are assembled on the outer circumferential surface of a cylinder roller and fabricated by metallic or solid materials [33]. Generally, MNRs are usually used to induce collagen production for cosmetic purposes [34,35] or for drug delivery systems by creating microchannels to increase skin permeability [36,37].
Combining the advantages of DMNs and MNRs, we proposed a strategy to fabricate tip drug-loaded dissolving microneedle rollers (DMNRs), which can be used to deliver drugs rapidly and efficiently on large areas of skin or curved skin. Similar idea has also been mentioned by Jung-Hwan Park and coworkers [33], but there is no systematic study on the DMNR at present. In this work, PVA, an FDA-approved water-soluble polymer, was chosen to fabricate DMNRs instead of metallic or solid materials. The preparation method of DMNR is simple and low-cost. Additionally, this study further evaluated the mechanical properties and rapid drug delivery performances of DMNRs. In vivo studies were performed to test the insertion rates and drug delivery properties of DMNRs that contain different numbers of MNs. The insulin-loaded DMNRs have good potential to hypoglycemic effect to diabetes. To confirm the administration effect, insulin-loaded DMNRs were used for diabetic rats and the results suggested that the insulin-loaded DMNRs have potential to hypoglycemic effect to diabetes.

Fabrication of PDMS molds
PDMS molds with uniform cavities and different dimensions were initially fabricated using a laser-based and micromolding techniques as our previous study proposed [38]. PDMS prepolymer solution and curing agent were first mixed in a mold at a ratio of 10:1 (v/v), then carried it out for 30 min under vacuum to remove air bubbles generated by stirring, next cured at 60 °C for 5 h to form PDMS sheet. Subsequently, a laser engraving machine (VLS3.50, 50 W, Universal Laser System, USA) was used to prepare a series of conical cavities on the PDMS sheets to form PDMS molds.

Preparation of PVA solution
PVA powder was firstly dissolved in deionized (DI) water and heated with a magnetic stirring at 90 ℃ for 6 h to obtain 32% (w/w) homogeneous solution. Then, sucrose (acts as a stabilizer) was added into the solution (sucrose to PVA weight is 3:4) and kept stirring the mixture at 90 ℃ for 1 h and then cooled it down to room temperature to form 25.8% PVA mixed solution.

Fabrication and characterization of DMNRs
DMNRs were fabricated based on a casting process. Briefly, 200 μl of prepared drug solution (insulin solution or 2 mg/ ml SRB) was first applied to the PDMS mold under vacuum. Then, the residual drug solution was removed from the mold surface and the vacuum was continued to concentrate the drugs on the tips of the cavities of PDMS mold (for 30 min). Third, 25.8% PVA/sucrose solution was applied on the PDMS mold vacuum for 30 min. Subsequently, the PDMS molds were placed on a clean bench with a temperature of 27.3 ± 2 ℃ and a humidity of 54 ± 3% to dry naturally. Finally, PVA/sucrose microneedles with the drug tips were replicated and attached to the customized roller with doublesided tape. SRB was used as a model drug, and insulin solution was used as a therapeutic drug for diabetes. Detailed characteristics of DMNRs were observed using fluorescence microscope (SZX7, Olympus, Japan) and scanning electron microscope (SEM, Hitachi S-4000, Japan).

Demolding time study of DMNR
To ensure that the DMNs can be demolded from the PDMS mold without causing any MN damage, the demolding time of DMNR was first optimized. Briefly, DMNs were demold from PDMS mold after drying 2, 4, and 8 h (27.3 ± 2 ℃ and a humidity of 54 ± 3%). Each DMNR demolded at different times was then weighed immediately. The relative water content of DMNR demolded at different times was determined by comparing the weight change after immediate demolding and after drying under vacuum for 48 h.

Mechanical property tests of DMNRs
Mechanical properties of DMNRs were determined by a displacement-force test station with a mechanical sensor (Mark-10, Force Gauge Model, USA). Both axial force and shear forces of SRB-loaded DMNs with different heights (400, 600, and 700 μm) that mounted on the rollers were tested. Briefly, on the one hand, to access the axial compression performance, a DMNR with only one row containing five DMNs was attached to a flat stainlesssteel baseplate of the sensor probe while the tips of MNs pointing downwards, and the sensor probe was then lowered in the vertical direction. The data of displacement and force measurements were recorded when the tips of the MNs first touched the smooth stainless-steel platform and continued until the force reached 5 N. On the other hand, to test the shear force of DMNRs, the roller with a single MN was fixed on a vertically placed stainless steel plate, and the tips of MNs faced horizontal direction. The sensor probe was lowered in the vertical direction while a shear force was then given at 250 μm from the tips of the DMNs. Displacement and force measurements were recorded when the sensor probe first touched the MNs and continued until the shear force was zero. Subsequently, the ultimate shear force was obtained when the MN was fractured under shear force. After each test, the sensor probe was moved upwards at a speed of 10 mm/min and DMNs were then visualized using fluorescence microscope.

Insertion studies of DMNRs on porcine cadaver skin
To assess the in vitro insertion abilities of DMNRs with DMN heights of 400, 600, and 700 μm, porcine cadaver skin without subcutaneous tissues was used as a human skin model. Before experiments, skin was separated into small pieces of 5 × 5 cm and stored in a refrigerator at − 20 °C. Prior to insertion test, the skin was first immersed in PBS and thawed for 30 min, and excess moisture on the skin surface was absorbed using commercial tissue papers, and then the hair of the pig skin was carefully removed by a disposable razor. SRB-loaded DMNRs were installed on a homemade applicator and then rolled forward once row by row across the skin. To control the force applied on the DMNRs, we referred to the method of previous study [33] that placed the skin on a balance which DMNRs were applied. The skin surface was imaged by fluorescence microscope after DMNR insertion. To clearly evaluate the insertion depth of DMNRs with different height DMNs, the skin samples were sliced from the middle of the DMNs insertion position points and viewed from the side using fluorescence microscope.

Insertion studies of DMNRs on Parafilm M®
To further observe the holes created of DMNR visually, Parafilm M®, a validated skin model [39], was used as an artificial skin simulants. The Parafilm M® sheet was folded to get an eight-layer film (approximately 127 μm thickness per layer). Then, SRB-loaded DMNR (375 DMNs, 600 μm) was installed on a homemade applicator and rolled forward once across the eight-layer film row by row. After insertion, the sheet was unfolded and observed using a handheld microscope (AD7013MZT, Dino-lite, Taiwan, China). The number of holes created in each layer by DMNRs was then calculated under microscope after washing off the fractured MNs in the sheet.

Rapid drug delivery performance tests of DMNRs
The in vitro drug delivery performance of DMNRs was investigated by controlling the insertion time and insertion ways. Porcine cadaver skin described above was used as a skin model. For easy observation, DMNPs and DMNRs with only one row (5 MNs/row, 600 μm height) were used to insert the skin, respectively. The experiments were divided into two groups: (1) DMNP with 5 MNs/row was vertically inserted into the skin and then removed vertically at 1 s; (2) DMNR was inserted into the skin by rolling it once across the skin and removed within 1 s. Immediately, top views and side views of the skin after insertion were imaged using fluorescence microscope. DMNs were visualized using fluorescence microscope and SEM before and after insertion. The DMNs for each group had the same geometry. Each group repeated five times. Moreover, the model drug does not diffuse immediately in a short time after being transported into the skin by DMNs, so the insertion situation can be evaluated by observing the number of red spots left after the drug was delivered into the skin.

Drug release characteristic of DMNRs in vitro
The in vitro drug delivery of SRB was conducted using a Franz diffusion cell. The receptor chamber was filled with of phosphate-buffered saline (PBS; pH 7.4, 7.1 ml) with continuous stirring at 300 rpm and the temperature maintained at 32 °C. In detail, the SRB-loaded DMNR was first applied on the mouse cadaver epidermis. Subsequently, the DMNRtreated skin was placed on a cell with inside of epidermis contacting with PBS in the receptor chamber. Then, 300 µl of sample was withdrawn from the receiver chamber at the predetermined time and supplemented with an equal volume of fresh PBS solution. The fluorescence intensity of the samples was measured using microplate reader. Therefore, the cumulative amount of released SRB was obtained.

Insertion rate and drug delivery efficiency tests of DMNRs in vivo
To assess and compare the in vivo insertion and drug delivery performances of DMNRs and traditional DMNPs, the insertion rate and drug delivery efficiency were measured. UV-sterilized DMNRs containing different numbers of SRBloaded DMNs (225, 375, 525) were separately applied to the dorsal skin (with a certain curvature) of female SD rats, and the DMNPs with the same geometry and same number of DMNs to DMNRs were fabricated as control group. Prior to the experiment, the dorsal region hair of rats was first removed using electric clippers and depilatory cream under anesthesia, and then cleaned by 75% ethanol. Notably, DMNRs were rolled once across the dorsal skin and removed immediately, while DMNPs were inserted into the skin and removed after 2 min. Subsequently, the skin of insertion sites was immediately observed by handheld microscope, and the amounts of red spots created by SRB-loaded DMNRs or DMNPs on the insertion sites were counted to calculate the insertion rate by the following equation: At the same time, the drug residues that remained on the rollers and patches respectively were collected, as well as residual drug on the insertion sites administrated by DMNRs and DMNPs, to measure the fluorescence intensity using a fluorescence microplate reader (Fluoroskan Ascent 374, Thermo Scientific). The amount of residual drug was calculated. Drug loading of per DMNR or DMNP was also determined using fluorescence microplate reader. Therefore, drug delivery efficacy was calculated according to the following equation:

In vivo visualization of drug delivery of DMNR using IVIS imaging tool
To visualize cutaneous permeation of model drug in live SD rats, the live anesthetized rats were first treated with the same size model drug-loaded DMNRs and DMNPs (600 μm (1) Insertion rate = the number of red spots on the skin after insertion the number of microneedles before insertion (2) Drug delivery efficiency = drug loading − drug residual drug loading height), respectively. Then, imaging of the insertion sites of the rat skin at different times (0, 2, 4, 6, and 8 h) after treatment of DMNRs or DMNPs was done using a non-invasive analytical in vivo imaging system (IVIS, Xenogen 200, Caliper Life Sciences, Hopkinton, MA) to visualize cutaneous permeation of drug in live rats.

Rat model of type I diabetes
Sprague-Dawley (SD) rats (6-8 weeks, 200-230 g) were purchased from Beijing Vital River Laboratory Animal Technology Co., Ltd. Rats were housed in laminar flow cages, provided with ad libitum access to food and water, and maintained in a specific environment with temperature of 22 ± 1 ℃, relative humidity of 50 ± 10%, and a 12 h light-dark cycle. All rats were allowed to acclimatize the laboratory environment for 1 week before experiments. After overnight fasting, type 1 diabetic models were induced in SD rats by single intraperitoneal injection with 75 mg/kg streptozotocin in 200 μl sodium citrate buffer (0.1 M, pH 4.5). Rats with fasting blood glucose levels exceeding 300 mg/dl were confirmed as type 1 diabetic rats. All procedures of animal studies were conducted in accordance with the guidelines of China-Japan Friendship Hospital for care and use of laboratory.

Blood glucose control studies in type I diabetic rats
Prior to the experiment, type 1 diabetic rats were fasted for 6 h but received water ad libitum, and the back hair of each group was shaved using an electric shaver under anesthesia. All animals were casually divided into the following 4 groups (n = 4 for each group): (1) control group: DMNRs without insulin loaded were applied to the dorsal skin of diabetic rats; (2) subcutaneous (SC) group: diabetic rats were injected 0.3 IU insulin solution subcutaneously by disposable insulin syringes on the dorsal skin of diabetic rats; (3) DMNRs group: insulin-loaded DMNRs (0.3 IU per roller) were applied to the dorsal skin of diabetic rats; (4) DMNPs group: insulin-loaded DMNPs (0.3 IU per roller) were applied to the dorsal skin of diabetic rats. The blood glucose levels of rats in each group were measured using One Touch glucometer before and 0.5, 1, 1.5, 2, 2.5, 3, 4, 5, and 6 h after insulin-loaded DMNR administration. Besides, an optical coherence tomography (OCT) was used to evaluate the mechanical property of the insulinloaded tip of MNs.

Fabrication and characterization of DMNRs
The fabrication process of DMNR can be divided into the following three steps (Fig. S1, A to G): (1) drug solution was applied to the PDMS mold and concentrated on the tips of MN cavities under vacuum; (2) PVA/sucrose solution was cast into the MN cavities under vacuum for 30 min to fill the drug-loaded cavities of PDMS mold; (3) subsequently, the PDMS molds were placed on a clean bench with a temperature of 27.3 ± 2 ℃ and a humidity of 54 ± 3% to dry naturally. Finally, PVA/sucrose microneedles with the drug tips were replicated and attached to the customized roller with double-sided tape (3 M™ double-sided medical tape).
In addition, demolding time of DMNRs was optimized and the results are shown in Fig. 1. Obviously, DMNs could not be completely demolded after drying for only 2 h (Fig. 1A), but it could be successfully completely demolded without causing any damage after drying for 4 h (Fig. 1B) or even 8 h (Fig. 1C). The relative water content of DMNs demolded at different times was also measured, and the result has shown that the relative water content of DMNs demolded at 2 h was 17.8 ± 1.2%, and as the drying time increased to 4 h and 8 h, the relative water content of DMNs decreased to 9.2 ± 0.3% and 6.5 ± 0.5%, respectively (Fig. 1D).
According to the approach described above, five kinds of DMNRs were successfully prepared (Table S1). Microscope images of DMNRs with different heights MNs are illustrated in Fig. 2, and the overview morphology of DMNR with MN height of 600 μm was recorded as shown in supporting information (Movie S1). Obviously, the MN geometries remained intact and the tips of MNs were not damaged after demolding via rollers. Besides, by observing the seam morphology of DMNR after being stored for 1 day and 45 days after demolding (Fig. S2), it was found that the seam of DMNR after storing 45 days has almost no change compared to the first day after demolding, which is good for DMNR for its long-term storage.

Mechanical performance
Effective insertion is an important prerequisite for MNs to deliver drug into skin effectively. Several physical factors, such as MN height, density, and geometry, have been reported to affect MN mechanical performance [40]. In addition, shear force is also an important factor affecting the performance of the drug delivery for DMNRs. Hence, the axial force and shear force tests of DMNRs with different heights (400, 600, and 700 μm) were performed to characterize the mechanical strength of DMNRs.
The data of the axial force-displacement curves of DMNRs with different height DMNs clearly indicate that the axial force of all the three kinds of DMNRs increased with the increase of displacement (Fig. 2E). Also, it was suggested that the axial force of the DMNRs gradually deteriorates as the MN height increased. Notably, although an array of MNs was used to test the axial force test, tip fracture is differently induced with the same geometry of MNs. Therefore, to further confirm the axial force of the MNs with different heights, an axial force test using a single MN and Similarly, the fracture point for 700 µm is about 0.12 N as the displacement reached 0.16 mm. Therefore, it can be inferred that the 400 µm MN can suffer from a higher axial force before failure and the force become lower as the height increased to 700 µm. This decrease is mainly due to that the higher the DMN height, the smaller the DMN diameter near the tip portion. Moreover, a shear force was given at the position of 250 μm from the tips of the DMNs. The shear displacements at the fracture occurred for all three groups are different. For 400 μm height MNs, the shear displacement is about 180 μm, but for 600 μm and 700 μm group, the shear displacements are around 143 μm and 107 μm, respectively. As shown in Fig. 2G, the ultimate shear forces for each group of different heights MNs of DMNRs were totally different. For DMNR with MN height of 400 μm, the MN was fractured when ultimate force reached about 0.90 N; however, for the MNs of 600 μm and 700 μm height, they were fractured while the average shear force reached about 0.46 N and 0.20 N, respectively.

Insertion ability of DMNRs in vitro
To evaluate the in vitro insertion abilities and optimize the height of MNs for efficient drug delivery of DMNRs, porcine cadaver skin without subcutaneous tissues was used as an in vitro skin model and SRB-loaded DMNRs with DMN height of 400, 600, and 700 μm were prepared (Fig. 3A1 to A3). The porcine cadaver skin surface after administration by DMNRs was imaged in bright field (Fig. 3B1 to B3) and fluorescence field (Fig. 3C1 to C3). Obviously, for 400 μm height DMNs of DMNR, only a small amount of drug was delivered into the skin after rolling the DMNR once across the skin. For 600 μm group, the red spots left on the skin were clear and uniform (Fig. 3B2 and C2). But as the DMN height increased to 700 μm, most DMNs were fractured on the skin surface before insertion. (Fig. 3B3 and  C3). Then, the skin samples were sliced from the middle of the DMNR insertion points and viewed using fluorescence microscope under bright field (Fig. 3D1 to D3) and fluorescence field (Fig. 3E1 to E3). Figure 3 F1 to F3 are the gross observation of the material left on the rollers of all the three different height groups. Obviously, the DMNRs with a height of 600 µm have the least residual drug after administration (Fig. 3F2). For the 400 µm height group, most of MNs remain intact (Fig. 3F1). For the 700 µm height group, most of the needles are broken without piercing the skin  (Fig. 3F3). These results indicate that the rapid drug delivery ability of DMNR with MN height of 600 μm was much better than MNs with height of 400 μm or 700 μm. Therefore, DMNRs with MN height of 600 μm were selected for testing in subsequent experiments.
To further observe the drug delivery way and holes created of DMNRs with DMN height of 600 μm visually, Parafilm M® was used as a skin model. After insertion, it could be clearly observed that the drug-loaded tips of DMNs were broke in the Parafilm M® (Movie S2). The insertion holes

C3
created of DMNRs could be clearly observed (Fig. S4), and it was found that almost all DMNs could penetrate the second layer of Parafilm M® and even over 80% DMNs could penetrate the third layer. Of note, the insertion depth of DMNR on the Parafilm M® is much deeper than that on the porcine skin surface. This difference mainly contributed to tips of DMNs being eroded when they touched the interstitial fluid when they were inserted into the skin while they were not affected when they were inserted into the Parafilm M®. Besides, differences in mechanical properties between the skin and parafilm also affect the insertion result.

Rapid drug delivery of DMNR
Most traditional DMNPs need to be vertically inserted into skin for several minutes or hours to ensure the DMNs completely dissolved [41,42]; therefore, the DMNPs are required to be attached to the skin surface all the time during MN dissolving, which is very inconvenient for patient. In this work, the rapid drug delivery performance of DMNR (MN height is 600 μm) was also assessed. As shown in Fig. 4, compared with the DMNs before insertion (Fig. 4A), only the tips of DMNs produced local surface erosion (Fig. 4B1 to B3) when the DMNP was vertically inserted into the skin and then removed vertically after 1 s (only suffered from an axial compression force). And the top view (Fig. 4B4) and side view (Fig. 4B5) of the skin after insertion show that only a small amount of model drug was delivered into the skin. Differently, when a roller covered with DMNs of the same geometry to DMNP was rolled across the skin, it can be clearly observed that the drug-loaded tips of the MNs were fractured into the skin because of the quick shear force and skin resistance, and there were only a few drug residues on the MNs after insertion (Fig. 4C1 to C3). Besides, the top view (Fig. 4C4) and side view (Fig. 4C5) of the skin images after insertion by DMNR show that the tips of MNs were then dissolved in the skin and the drug was released into the skin, suggesting that it is unnecessary for DMNR to be attached to the skin surface for several minutes or even longer time like a DMNP and it can deliver the drug into the skin in a rapid manner.

Drug release characteristics of DMNRs in vitro
In this experiment, 15 × 25 arrays (drug loading: ~ 7.69 µg) of SRB-loaded DMNRs were used to determine the drug release characteristic of DMNRs. The result is shown in Fig. S5. It was found that the DMNR released almost 50% (about 3.7 µg) of the SRB in the first 0.5 h after application and nearly 65% (about 4.9 µg) at 2 h, which further indicated that the DMNR has the potential to deliver drugs in deep skin.

Insertion rate and drug delivery efficiency in vivo
In addition to shortening drug delivery time, another important goal of ours for DMNR is to expand the application of DMN to enable delivering drugs on the large areas of skin  from 225 to 525 for DMNRs. However, for DMNPs, the red spots could be clearly observed on the rat skin surface after insertion by DMNPs with 225 DMNs (Fig. 5C1), but as the number of DMNs increased to 375, some DMNs near the edge of the DMNP could not be inserted into the rat skin very well (Fig. 5C2), especially when the number of DMNs increased to 525; the number of DMNs near the edge of the DMNP inserted into the skin was significantly reduced due to the skin deformation (Fig. 5C3), indicating that the drug delivery performance of DMNP was seriously affected by the patch size. Subsequently, the number of red spots created by different arrays of SRB-loaded DMNRs and DMNPs on the insertion sites was separately counted, and the insertion rate was then calculated (Fig. 5D). The results show that although the number of DMNs arranged on the DMNRs was different, the insertion rate on the skin of each group was all over 98% and there was no significant difference among each group. It shows that although the number of DMNs arranged on the DMNRs was different, the insertion rates on the skin for DMNRs were all over 98% and almost no significant change among each group; however, for DMN patch, the insertion rate was decreased from 100 to 93% as the DMNP enlarged from 225 to 525 DMNs.
The difference of insertion rate for the DMNR and the DMNP as mentioned above is mainly due to the following two reasons. First, the drug delivery method of DMNR is rolling the roller across the skin row by row, which is less affected by skin elastic deformation. But for DMNP, all MNs that arranged on the patch need to be vertically inserted into skin at once, which is easy for the small patch but difficult for the large patch because of the skin deformation. Another important reason is that the dorsal skin of rat is arcuate rather than flat so that the roller can be rolled along the curved skin and does not affect the MNs of other rows on the roller; when a small DMNP was inserted into the rat skin, the patch could well fit to the skin (Fig. 5F). In contrast, when a larger patch was applied on the arcuate skin, the MNs arranged on the edge of the patch could not contact well to the skin due to that the base of DMNP cannot bend freely (Fig. 5G, marked by the red line circle); therefore, the insertion rate of DMNP on the rat skin was decreased as the MN patch size enlarged.
Moreover, drug delivery efficiency test results are illustrated in Fig. 5E. The average drug delivery efficiency for small DMNP (225 MNs, drug loading: 6.69 µg) was up to 94.7%, but it was dropped to 83.1% and 79.4% as the number of MNs increased to 375 (drug loading: ~ 7.69 µg) or 525 (drug loading: 9.61 µg). However, the drug delivery efficiency of DMNRs was almost the same and maintained 83% approximately even the number of DMNs arranged on the roller was increased from 225 to 525, suggesting that the drug delivery efficiency of DMNRs is more stable than DMNPs. Notably, drug delivery efficiency for DMNP with 225 MNs was much higher than that for DMNR. This difference is mainly due to the tendency of the drug to diffuse from a high concentration to a low concentration during the fabrication process, so that the drug will diffuse from MN tip to MN bottom. Namely, when the drug diffuses from tip to a position below the fracture position of the DMNs of DMNR, those drugs are still on the MNs after insertion (Fig. 5H) and consequently the drug delivery efficiency is affected.

Transdermal drug delivery
The fluorescence intensity of model drug delivered into the dorsal skin of SD rat using DMNR and DMNP was visualized by IVIS. The fluorescent intensity of both DMNP (left) and DMNR (right) treated skin decreased over time (Fig. 5I), indicating that the model drug gradually diffused into the deep tissues of skin. Remarkably, the skin fluorescence intensity at 0 h after DMNR treatment was observed to be more uniform than that of DMNP treatment because the MNs arranged on the roller were still in good contact with curved skin, while the MNs arranged at the edge of the patch did not.

Hypoglycemic effect of insulin-loaded DMNRs for diabetic rats
To assess the practical application of the DMNRs, the insulin-loaded DMNR was applied to deliver insulin to the STZ-induced SD diabetic rats. To evaluate the mechanical characteristics of the insulin-loaded tip of MNs, OCT was used to observe whether the needle tip of DMNR can penetrate the skin and be successfully implanted under the skin. Compared with untreated skin (Fig. S6A), it can be seen that the drug-loaded tip of the DMNR has sufficient strength and is successfully implanted in the skin after DMNR administration (Fig. S6B).
The blood glucose level changes of diabetic rats can reflect whether the drug was successfully delivered into the skin indirectly. The result is shown in Fig. 6. The blood glucose level of control group has almost no change after transdermal administration of insulin-unloaded DMNRs to the diabetic rats (control). In contrast, for insulin-loaded DMNRs administration group (insulin DMNRs, 0.3 IU), the blood glucose level was dramatically reduced blow 50% at 1 h after insulin-loaded DMNR administrated and reached minimum level (~ 28%) at 1.5 h, then returns to initial level after 6 h. For the DMNP group, the blood glucose level was reduced faster than that of the DMNR group at first 0.5 h, but the controlled minimum blood glucose level was higher than that of the DMNR group. This difference is mainly related to the insertion of the two types of MNs. For the DMNP, almost the entire MN body has been inserted in the skin, but only the needle tip was implanted in the skin for DMNR group. That is to say, the insertion depth of DMNP is deeper than DMNR. Therefore, insulin is better absorbed in the first 0.5 h for the DMNP group. However, as described above, due to the skin deformation, part of needles arranged on the patch cannot be inserted in to the skin, especially on the bumpy sites. Thus, the controlled minimum blood glucose level of DMNP was higher than that of the DMNR. Namely, DMNR has a higher drug delivery efficiency than the DMNP in our test size. It also demonstrates that the insulin-loaded DMNR displayed a good potential for insulin delivery to diabetes.

Discussion
To enable DMNs for drug delivery in a rapid and highly efficient manner on the skin, especially on the large areas of skin or curved skin, a DMNR device was designed, which is not required to be attached to the skin surface during drug delivery and can be used for rapid and highly efficient drug delivery because of its high mobility and ability to cover large surface areas. Compared to the traditional DMNP preparation process (Fig. S1, A-E, H, I), the fabrication process of DMNRs has only changed the demolding method using a roller with double-side tape instead of a small patch to arrange the DMNs, which is easy to operate, no heating required and low-cost.
Demolding time of DMNRs was researched firstly, and it was found that the DMNs could be successfully demolded from PDMS molds after drying 4 h or even 8 h. Then, an optimized DMN height was obtained according to the mechanical performance test and in vivo skin insertion test.
The 600 μm height DMN shows excellent skin insertion ability and drug delivery performance. Parafilm M® after insertion by DMNR (Movie S2) and the DMNs morphology of the DMNR after it was rolled once across the pig cadaver skin (Fig. 4) then confirmed that the main way for DMNR to deliver drugs is mainly due to fracture of DMNs under a shear force and skin resistance, and the drug-loaded tips were first fractured into the skin and then dissolved to release drugs, which is convenient and no risk of infection or erythema caused by the patch being attached to the skin for a long time, especially for the skin sensitive patients.
Also, the double-sided sticky tape attached on the roller is 3 M™ double-sided medical tape, which is biocompatible. Before use, the double-sided tape is placed on the clean bench to avoid contamination. Besides, the double-sided sticky tape attached on the roller is not in contact with skin during DMNR administration. Therefore, there is no worry about the skin irritation/undesired adhesion occurred associated with double-sided sticky tape. But it is necessary to keep the surrounding environment clean during demolding.
The insulin-loaded DMNRs were also used for type 1 diabetic rats. The OCT images suggested that the DMNR can successfully implant the tips of MN into the skin. The hypoglycemic effect results showed that the blood glucose level was dramatically reduced below 50% at 1 h and reached minimum level (~ 28%) at 1.5 h, then returns to initial level after 6 h after insulin-loaded DMNR administrated. The change tendency of blood glucose level of insulin-loaded DMNR group was similar to the SC group but hypoglycemic effect of DMNR is weaker than the SC group. Besides, the blood glucose level was reduced slower in DMNR group than that of the DMNP group at first 0.5 h, but the controlled minimum blood glucose level was lower than that of the DMNP group. This difference is mainly related to the insertion depth of DMNR. All the results indicated that the DMNR, as a safe and rapid drug delivery device, could provide a solution to the problems of drug delivery on the large skin or curved skin. For future, it is expected to be optimized for delivering drug for diabetes, and some large skin diseases such as psoriasis or other purposes such as cosmetic.
Funding This work was financially supported by the National Natural Science Foundation of China (51873015), the Joint Project of BRC-BC (Biomedical Translational Engineering Research Center of BUCT-CJFH) (XK2020-05, RZ2020-01), and the long-term subsidy mechanism from the Ministry of Finance and the Ministry of Education of PRC.
Availability of data and materials All data generated or analyzed during this study are included in this work.

Declarations
Ethics approval and consent to participate The experimental protocol was established, according to the ethical guidelines of the BUCT Declaration, and was approved by the Ethics Committee of Beijing University of Chemical Technology. Written informed consent was obtained from individual or guardian participants.

Competing interests
The authors declare no competing interests.