Temporal pressure enhanced topical drug delivery through micropore formation

See allHide authors and affiliations

Science Advances  29 May 2020:
Vol. 6, no. 22, eaaz6919
DOI: 10.1126/sciadv.aaz6919


Transdermal drug delivery uses chemical, physical, or biochemical enhancers to cross the skin barrier. However, existing platforms require high doses of chemical enhancers or sophisticated equipment, use fragile biomolecules, or are limited to a certain type of drug. Here, we report an innovative methodology based on temporal pressure to enhance the penetration of all kinds of drugs, from small molecules to proteins and nanoparticles (up to 500 nm). The creation of micropores (~3 μm2) on the epidermal layer through a temporal pressure treatment results in the elevated expression of gap junctions, and reduced expression of occludin tight junctions. A 1 min treatment of 0.28-MPa allows nanoparticles (up to 500 nm) and macromolecules (up to 20 kDa) to reach a depth of 430-μm into the dermal layer. Using, as an example, the delivery of insulin through topical application after the pressure treatment yields up to 80% drop in blood glucose in diabetic mice.


Transdermal drug delivery (TDD) offers a convenient and patient-friendly way for the treatment of both local disorders and diseases of other organs. It allows drugs to bypass the first pass metabolism while providing sustained and controlled delivery (1, 2). While traditional cream or gel formulations passively deliver small and lipophilic drugs through the stratum corneum (SC) via diffusion, macromolecules (>500 Da) with very low permeability cannot diffuse through the compact lipid-rich matrix of SC (3). Chemical and biochemical enhancement techniques have been developed over the past decades to disrupt SC for the delivery of large and hydrophilic drugs. However, these enhancers are potent irritants at high concentrations and thus affect therapeutic compliance (4, 5). Physical enhancement methods amiable for routine and effective delivery have also been developed (e.g., electroporation, laser ablation, and iontophoresis) (1, 6, 7). Nonetheless, these physical means are accomplished through utilization of additional, specialized equipment alongside trained users, to modulate skin permeability, thereby presenting challenges in terms of patient convenience and home use (8). Another technology worth mentioning is the microneedle (MN), which has received vast attention recently (9, 10). They can be fabricated with different needle length and width (150- to 1500-μm long, 50- to 250-μm wide), with a few drug loading options (10). However, because of their small sizes, the amount of drugs that can be delivered is usually within microgram range (11, 12).

Amidst the popular technologies mentioned above, rubbing serves as a potential method to deliver drugs transdermally. While the mechanism of this method is still unclear (13, 14), the limited literature review on rubbing suggests that the locally applied pressure can benefit TDD. One example is the jet injector. High-pressure liquid jet of ~14 MPa and a diameter of 340 μm can pierce the skin, with a depth of ~2 mm, for delivery of medication such as corticosteroids (15), local anesthesia (16), bleomycin (17), etc. However, the piercing of the skin is a concern to many researchers and clinicians due to risk of infection, pain, or injury of operator’s finger (18). Another technology using similar concept is the ultrasound-mediated TDD that administers an acoustic oscillating pressure wave. The oscillation increases the size of cavitation bubbles in the fluids of the skin. When imploded, these bubbles create an intense local shockwave, disrupting the SC for improved TDD. The range of frequencies is in the range of 20 kHz to 16 MHz to deliver insulin, mannitol, glucose, morphine, and lidocaine (19). While ultrasound-mediated TDD has proven to be effective, it is limited by the complex machinery and associated skin tissue heating, which can damage the deeper tissue (20). Therefore, there still remains a need for simple, cost-effective, and minimally invasive TDD technology to enable convenient transdermal drug administration by the patients.

This article introduces a pressure-based TDD methodology to address this unmet need (Fig. 1A and fig. S1). Tapping on established studies where magnets are used to mimic pressure ulcer injury in vivo due to its well-defined magnetic field and force (21), this study chose neodymium magnets as the proof-of-concept tool to produce the required pressure force to induce temporal skin barrier changes for drug delivery. Specifically, a local pressure (0.14, 0.28, and 0.4 MPa) is generated by applying two neodymium magnets to pinch skin. After a given time, the magnets are removed, and therapeutics in the moisturizer are topically applied to the pressure-processed area. Here, we report the successful delivery of nanoparticles (NPs; up to 500 nm), dextran molecules (up to 20 kDa), and insulin across the skin of mice. The improved penetration is observed through inter- and transcellular routes. We optimize this process and propose 0.28 MPa and 1 min of application as the ideal procedure to achieve an effective penetration without compromising the skin barrier. Furthermore, we demonstrate the effective delivery of insulin into the systemic circulation of both normal and diabetic mice and achieve the 65 and 80% reduction in blood glucose levels, respectively.

Fig. 1 Temporal pressure enhanced transdermal delivery.

(A) Schematic diagram demonstrating the effect of temporal pressure application leading to the occurrence of microphysiological changes, allowing the delivery of drugs across the skin barrier. (B) Representative images of mice with topically applied fluorescent nanoparticles (NPs) after the pressure and MN treatment. (C) Quantification of the fluorescent signal in (A). (D) Quantification of NPs in the dermis. (E) Fluorescence imaging [blue, 4′,6-diamidino-2-phenylindole (DAPI); red, NPs] of histological skin samples in (A). (F) Left: H&E staining of skin samples. Right: The appearance of mouse skin with or without the pressure treatment and MN. Scale bars, 100 μm. n = 3, all data are means ± SD, *P < 0.05. Photo credit: Daniel Chin Shiuan Lio, School of Chemical and Biomedical Engineering, Nanyang Technological University.


The effectiveness of this procedure was initially studied on mice by taking 200 nm fluorescent NPs as the model drugs and comparing this with solid MNs (needle height of 1000 μm, 10 × 10 array with density of 400 needles/cm2). The 0.14-MPa pressure was applied for 1.5 hours, while MN was applied for 5 min as previously reported (22, 23). NPs were applied right after the removal of magnet and MN and were left on the skin for 12 hours. The treated skin was subjected to further analysis. As shown in Fig. 1B, fluorescent NPs were found to be localized only at areas where the pressure and MN treatment were performed (red dashed circle), without spreading to surrounding tissues. The amount of NPs retained in the skin was minimal for the sample without pressure treatment (~14% of the signal from that with pressure and MN treatment, Fig. 1C). No difference in the amount of NPs in the skin was observed between pressure and MN treatment. Histology showed that NPs mainly stayed on the SC layer without pressure treatment but reached the dermis layer upon following pressure and MN treatment (Fig. 1E). Further quantification revealed that pressure treatment increased the NP concentration in the dermis layer at least ninefold (Fig. 1D).

When there was no pressure, NPs entered the skin mainly through appendages (Fig. 1E). The pressure treatment allowed NPs to get into the skin layers through both transappendageal and nontransappendageal routes (trans- and intercellular routes). NPs tended to agglomerate when they entered through follicular and eccrine routes, but pressure treatment permitted NPs to enter through trans- and intercellular routes and distribute more evenly within the skin. On the other hand, application of MN allowed the delivery of NPs into the dermal tissue (Fig. 1E). Hematoxylin and eosin (H&E) staining revealed a thickening of the epidermis after 1.5 hours of 0.14-MPa pressure and overnight application of NPs (Fig. 1F), while no notable immune response was seen for without pressure and MNs.

Pressure is the key element in this methodology. We treated mice with 0.14, 0.28, and 0.4 MPa of pressure, respectively (fig. S2A). Regardless of pressure intensities, NPs were localized only at areas where pressure was applied (fig. S2C). However, greater pressure led to stronger NP signal in the skin. In comparison to the control, 0.14 MPa brought a ~5-fold increase of the NP signal, 0.28 and 0.4 MPa provided seven- and ninefold improvement, respectively (Fig. 2C). A further histological analysis of the skin samples confirmed the improved penetration of NPs in the skin layers (fig. S2C). Similar to previous observations (Fig. 1E), NPs entered the skin through both transappendageal and nontransappendageal routes. The increase of NP signal in dermal layer was less than we observed in in vivo imaging. For example, ~9-fold increase of NP signal was observed for 0.4-MPa treatment, as shown by in vivo imaging system (IVIS) imaging (Fig. 2C), while there was an only 2.25-fold increase in dermal layer as shown by histological analysis (Fig. 2E). The difference comes from the different signal sources. IVIS imaging is capable of collecting signals from both epidermal and dermal layers (24), and thus, fluorescent signal in Fig. 2C comes from both epidermal and dermal layers. In another words, there was also an increase in NP retention in epidermis layer after the pressure treatment.

Fig. 2 Optimization of temporal pressure application.

(A) Overview of the procedure performed to optimize pressure and application time. (B) Representative IVIS images of the mice treated with 0.28 MPa for different application times. (C) Quantification of fluorescent signal in IVIS imaging of mice treated with different pressures (0.14, 0.28, and 0.4 MPa). (D) Quantification of the fluorescent signal in IVIS imaging of mice treated with the same pressure (0.28 MPa) but with different duration (1 min, 5 min, 0.5 hours, and 1.5 hours). (E) Quantification of the fluorescent signal in the dermis of mouse skin treated with different pressures (0.14, 0.28, and 0.4 MPa). ns, not significant. (F) Quantification of fluorescent signal in the dermis of mouse skin treated with the same pressure (0.28 MPa) but with different durations (1 min, 5 min, 0.5 hours, and 1.5 hours). (G) Fluorescence imaging (blue, DAPI; red, NPs) and H&E staining of mouse skin treated with the same pressure (0.28 MPa) but with different durations (1 min, 5 min, 0.5 hours, and 1.5 hours). Scale bars, 100 μm. n = 3, *P < 0.05 and **P < 0.01. Photo credit: Daniel Chin Shiuan Lio, School of Chemical and Biomedical Engineering, Nanyang Technological University.

While 0.4-MPa treatment provided the highest NP retention in the skin, we observed that the SC to stratum basale layers of the mouse skin was damaged under this pressure (fig. S2C). Treatment with 0.14 and 0.28 MPa increased the epidermal thickness to approximately twofold (fig. S3B). The sudden decrease in the epidermal thickness of mice treated with 0.4 MPa confirmed our observation of the loss of SC to stratum basale layers. Under all pressures, the tissue in dermis layer became disorganized with an increase in hematoxylin-stained nuclei (~1.5-, ~1.9-, and ~2.7-fold, respectively) (fig. S3A), which suggests inflammation at the treated areas. As 0.14-MPa treatment did not provide a significant increase (P > 0.05) in NP concentration in dermal layer when compared to the control (Fig. 2C), 0.28 MPa was selected for the subsequent experiments to maximize the penetration of the drugs while minimizing damage to the skin.

Besides the pressure, another factor is the duration of pressure application (fig. S2B). As shown in Fig. 2D, longer treatment brought more NPs into the skin. Compared with the control, 0.5 and 1.5 hours of pressure treatment improved the NP fluorescence signal in the skin six- to sevenfold, respectively. There was ~4-fold increase even with 1 and 5 min of treatment. Histological analysis confirmed a 70 to 90% increase of NPs in the dermis as well (Fig. 2F). Earlier, we observed thickening of epidermis (figs. S2C and S3B) and increased nuclei count with 0.28 MPa after 1.5 hours of treatment (fig. S3A). When the treatment was reduced to 1 and 5 min and 0.5 hours, both the epidermal thickening (Fig. 2G) and inflammation (fig. S4, A and B) were comparable to control.

Next, we examined the suitability of this methodology for various types of drugs including polymer and NPs. The polymer models were dextran with sizes of 3, 5, 10, and 20 kDa (Fig. 3A). They were topically applied on pressure-treated skin at the same mass concentration. We noticed that fluorescence intensities were different for these dextran molecules under the same mass concentration (fig. S5). The fluorescence intensity from 3-kDa dextran is four times that from 10- and 20-kDa dextran. Thus, results in the following experiments were normalized to account for this difference in fluorescence intensity.

Fig. 3 Topical delivery of dextran molecules after pressure treatment.

(A) Schematic of experiments. (B) IVIS image of mice after topical delivery of dextran molecules after the pressure treatment (C → no pressure, 1 → 1 min pressure treatment, 5 → 5 min pressure treatment). (C) Normalized quantification of dextran in treated skin in (B). (D) Fluorescence imaging (blue, DAPI; red, NPs) of histological skin samples in (B). Scale bars, 100 μm. n = 3, all data are means ± SD, *P < 0.05 and **P < 0.01. Photo credit: Daniel Chin Shiuan Lio, School of Chemical and Biomedical Engineering, Nanyang Technological University.

In vivo imaging revealed that pressure treatment increased the penetration of all sizes of dextran molecules into the skin (Fig. 3B). There was ~2-fold increase for both 3 and 5 kDa when treated with pressure. Furthermore, ~3.5-fold increase for 10 kDa and ~1.5-fold increase for 20 kDa were observed, compared to without treatment samples (Fig. 3C). This increasing trend was consistent regardless of the treatment time (1 or 5 min). By normalizing the fluorescence signals in the dermis (Fig. 3D) against the fluorescence intensity of each dextran (fig. S5), we found that there was higher amount of smaller dextran (i.e., 3 to 10 kDa) in the skin than dextran with higher molecular weight (20 kDa) (fig. S6). Histological analysis confirmed that all dextran molecules reached the dermis with the pressure treatment (Fig. 3D), but these molecules only accumulated on the SC layer with no pressure treatment.

Next, we examined the transdermal delivery of NPs ranging from 20 to 500 nm. Different from dextran molecules, different sizes of NPs under the same mass concentration showed similar fluorescence intensity (fig. S7). Without the pressure treatment, the smaller the NPs, the stronger fluorescence the mouse skin had (fig. S8). With the pressure treatment, a drastic increase in the amount of NPs was observed across all particle sizes. Histological analysis showed that pressure treatment notably improved the NP’s concentration in the epidermis and inside the appendages (fig. S9A). NPs were able to enter the dermis and distribute uniformly. Quantification of fluorescence particle count showed that mouse skin treated with 20-nm NPs had the highest fluorescence intensity (fig. S9B), similar to that observed in IVIS (fig. S8). Mice treated with 100- and 200-nm NPs after temporal pressure demonstrated an average of two times more NPs in the dermis. When mice were treated with 500-nm NPs with pressure, particles were found in the SC/epidermis only with negligible NPs in the dermis (fig. S9A). The cutoff size for NPs is 500 nm.

The rabbit is a widely used animal model for skin research due to their relatively large size and docile nature (2527). Using the above established protocol, we treated the rabbit skin with pressure and topically applied NPs in the moisturizer. IVIS imaging showed that the pressure treatment improved the NP fluorescence threefold (Fig. 4, A and C). Histology confirmed the increase of NP concentration in dermis (fig. S10), and no obvious skin damage was observed in H&E-stained sections (Fig. 4B). Aside from the dorsal skin, temporal pressure could also be applied on the rabbit’s ears, delivering drugs on popular in vivo rabbit ear wound model, e.g., keloid scars (28).

Fig. 4 Transdermal delivery of NPs in rabbit skin with the temporal pressure treatment.

(A) IVIS fluorescence imaging of excised rabbit skin after pressure treatment and application of NPs. Circled areas represent the area of interest. (B) Fluorescence imaging (blue, DAPI; red, NPs) and H&E staining of rabbit skin after the treatment. (C) Quantification of NP fluorescence signal in (A). Scale bars, 100 μm. n = 3, all data are means ± SD, *P < 0.05. Photo credits: Daniel Chin Shiuan Lio, School of Chemical and Biomedical Engineering, Nanyang Technological University.

Porcine skin displays similar skin thickness to that of humans (26, 29), and it is a good skin model to test the efficacy of temporal pressure. In addition, MNs were included as the positive control (fig. S11A). Demonstrated in fig. S11B, temporal pressure improved the drug delivery of NPs ~6-fold, while MN improved the delivery of NPs ~7-fold, with no significant difference between the two treatment groups. Histology confirmed the penetration of these NPs into the dermal tissue for treatment groups, while NPs were mostly retained in the SC for without pressure group (fig. S11D). This improvement was quantified to be ~11- and ~14-fold more than without pressure group (fig. S11E). Last, no obvious damage was observed for with pressure group in H&E-stained sections (fig. S11D).

The above experiments confirm that temporary pressure of 0.28 MPa can help polymer and NPs to enter the dermis layer of the skin through nonappendage routes. Immediately after 5 min of temporal pressure (0.28 MPa), high-resolution images revealed microphysiological changes in the skin tissue, with micropores of area ~2.61 μm2 (derived from 21 samples) (fig. S12A). Twelve hours later, although the mice skin did not suffer from any physical damage in this process visually, high-resolution imaging revealed some microphysiological changes (Fig. 5A). In mice skin, numerous micropores of area ~3.21 μm2 (derived from 55 samples) were observed between keratinocytes in the epidermis (black arrows), while no micropores were observed in the untreated mice. These micropores were observed to heal 24 hours later (fig. S17E). A similar phenomenon was observed in rabbit skin, 12 hours after application of 5 min pressure, where numerous micropores of area ~8.25 μm2 (derived from 92 samples) were observed in the epidermis (fig. S14).

Fig. 5 Micropore formation and altered junction protein expression in pressure-treated mouse skin.

(A) H&E staining of the mouse skin after the pressure treatment. The staining was performed 12 hours after the pressure treatment. Confocal images of pressure treated skin stained with DAPI (blue) and Cx43 (red) antibodies (B) and occludin (C). ImageJ quantification of Cx43 (D) and occludin (E) expression in the epidermis layer. Scale bars, 50 μm. n = 3, all data are means ± SD, *P < 0.05.

We further studied the proteins preserving epidermal integrity, connexin 43 (Cx43, gap junction protein) and occludin (tight junction protein). No significant changes were observed in the skin tissue 5 min after temporal pressure (fig. S12, B to E). On the other hand, a slight increase can be observed 12 hours after pressure application (Fig. 5, B and D). Treatment with 0.4 MPa resulted in the highest expression of Cx43 protein (Fig. 5D and fig. S13B). On the other hand, occludin expression was observed to be down-regulated by pressure (Fig. 5, C and E, and fig. S13C).

With the optimized condition (1 and 5 min, 0.28 MPa), we tested whether this strategy could be used for the transdermal delivery of real therapeutics like insulin (5.8 kDa). The protein (1 U or 0.035 mg) was topically applied to mice that were fasted overnight (fig. S15) and treated with 0.28-MPa pressure for 1 or 5 min. We found that 30 min later, there was a slight drop in blood glucose (10%) in the 1-min treatment group, which progressively dropped to ~50% after 3 hours (fig. S16). However, blood glucose was observed to drop to ~35 to 40% 30 min after 5 min pressure treatment and remained at ~20% after 3 hours. We suspect that this could be due to the slow diffusion of insulin molecules from skin to the circulation with only 1 min of pressure treatment. This set of experiments allowed us to choose 0.28-MPa pressure and 5-min treatment in the rest insulin delivery experiments.

The blood glucose level of treated mice was continuously monitored for 5 hours, to study the effect of long-lasting insulin in the systemic circulation (30). As shown in Fig. 6A, 5-min temporal pressure treatment plus topical insulin application decreased the blood glucose in mice to ~35% of its initial physiological state in 30 min, and this was maintained for almost 5 hours. There was no hypodermic shock on the mice, which were physiologically active and normal the next day. On the other hand, when mice were treated with topical insulin only, there was no change of blood glucose level, which was similar to that of control mice, confirming that the protein (i.e., insulin) cannot enter the skin and subsequently into the bloodstream. As a positive control, the same amount of insulin (1 U) was subcutaneously injected and resulted in a drastic decrease of blood glucose to ~15% of initial state in 30 min. The blood glucose was undetectable (0%) at 2 hours onward, indicating a hypoglycemic state. In view of this, the corresponding hypoglycemia index (defined as the decline in blood glucose from the initial value to the nadir, divided by the time taken to reach nadir) was calculated to measure quantitatively the extent to which insulin elicited hypoglycemia. Topical insulin delivery with temporal pressure showed a remarkably reduced hypoglycemic index compared with insulin injection (fig. S17C), suggesting that insulin treatment via temporal pressure has a much lower chance of experiencing hypoglycemic shock as compared to insulin injection (31).

Fig. 6 Blood glucose control in normal mice with topically delivered insulin after the pressure treatment.

(A) The change of blood glucose in the mouse circulation with topically delivered insulin after the pressure treatment, no pressure, and insulin injection over a period of 5 hours. (B) Quantification of insulin in the blood circulation after the pressure treatment and topical delivery of insulin over a period of 5 hours. n = 3, all data are means ± SD, *P < 0.05 and **P < 0.01.

We further performed an enzyme-linked immunosorbent assay (ELISA) immunoassay to quantify insulin, which entered the circulation (Fig. 6B). There was ~15 mU/liter of insulin in the blood at the first and second hours, which was sufficient to cause a drop in blood glucose (32). The concentration of plasma insulin increased to ~160 mU/liter at the fifth hour. The total insulin in circulation accounted to 13.4% of the insulin topically applied on the skin (fig. S17D). However, negligible insulin was detected in the blood without pressure treatment. To confirm that the drop of blood glucose across 5 hours in Fig. 6B was due to the topically applied insulin, we tail vein–injected the same amount of insulin (e.g., 160 mU/liter). As shown in fig. S17A, the blood glucose dropped to a similar level as in the mice treated with pressure and topically applied insulin in the first 30 min. However, blood glucose rose after 1 hour, which should be due to the enzymatic digestion of insulin. This observation is in par with the quantification of plasma insulin using ELISA (fig. S17B). Insulin level increased in the first hour but gradually decreased thereafter to negligible levels in 4 hours. The integrity of skin barrier was studied through H&E staining 24 hours after pressure application, and no inflammation was observed (fig. S17E).

The clinical potential of this methodology was further explored with the diabetic mouse models. Normal C57BL/6J mice were made diabetic by daily streptozotocin (STZ) injection of 50 mg/kg over a 5-day period (33). Increased urination was observed from the diabetic mice in approximately 4 weeks after STZ injection (fig. S18A). In addition, diabetic mice rapidly lost weight (fig. S18B), and their blood glucose was measured at ~25 to 30 mM while normal mice exhibited blood glucose of ~7.5 mM (fig. S18C).

Initially, we used the same condition (1 U of insulin with 5-min temporal pressure) as in Fig. 6 for the diabetic mice. Unfortunately, there was no significant drop in blood glucose over a 2-hour period (fig. S19A). Plasma insulin was measured at ~20 mU/liter at the first hour, similar to the control (fig. S18D). When the insulin amount was increased 10-fold, there was a significant drop in blood glucose from ~25 to ~7 mM for more than 2 hours (fig. S19A).

We benchmarked this methodology with three other insulin delivery methods including subcutaneous injection (34), dissolvable MNs (31), and topical application after MN treatment (Fig. 7). Same amount of insulin was either topically applied or encapsulated in MNs. Two dissolvable MNs was used during each treatment, as only half (~5 U) of insulin was contained in the MN tips. During the 5-day treatment, the positive control (i.e., subcutaneous injection of insulin) consistently controlled the blood glucose level, which dropped from 25.9 to 4.3 and 1.0 mM within the first and sixth hours, respectively (Fig. 7B). The insulin levels in blood were 16.1, 3169, and 2002 mU/liter accordingly (fig. S19B). With the pressure treatment and topical application of insulin, the blood glucose dropped from 28.6 to 9.2 mM at the first hour and then to 1.7 mM in the sixth hour. Plasma insulin increased from 13 to 350 and 1076 mU/liter at the first and sixth hours, respectively (Fig. 7F). If there was no pressure treatment, topically applied insulin will not cause any change of blood glucose (Fig. 7, A and C), and there was negligible insulin in the plasma as well (fig. S19B). When diabetic mice were treated with solid MN [made of poly(methyl methacrylate) (PMMA)] and insulin topically applied (31), blood glucose quickly dropped from 23.2 to 9.6 within the first hour (Fig. 7D) and then to 3.4 mM at the sixth hour. The insulin levels were 46 and 115 mU/liter of plasma insulin accordingly (fig. S19B). If insulin was delivered with dissolvable MNs, a gradual drop in blood glucose was observed from 27.4 to 22.3 mM at the first hour and then to 10.8 mM at sixth hour (Fig. 7E), with 18.6 mU/liter of plasma insulin detected and later increased to 40.5 mU/liter at the sixth hour (fig. S19B).

Fig. 7 Assessing temporal pressure as a method to topically administer insulin, daily, for 5 days.

Blood glucose control in diabetic mice for a period of 5 days with (A) control, (B) subcutaneous injection of insulin, (C) topical application of insulin, (D) topical application of insulin after MN treatment, (E) transdermal delivery of insulin with dissolvable MNs, and (F) topical application of insulin after pressure treatment. n = 3, all data are means ± SD.

We also notice that the skin barrier was compromised in this period (fig. S18E). The continuous pressure treatment thickened the epidermis and the cell-cell bond between stratum basal, and peripheral dermal layer was compromised.


This project developed a new methodology that used temporal pressure to enhance the penetration of all kinds of drugs ranging from small molecules to protein and NPs (up to 500 nm). Although magnets were used here, any pressure-generating device like mechanical actuator would be suitable. It started from the delivery of 200-nm NPs with 0.28-MPa pressure, which was compared to MN as positive control (Fig. 1). NP concentration in dermis layer was improved ~9-fold (Fig. 1D), where penetration occurred through nontransappendageal routes with even distribution within the skin (Fig. 1E). Similar penetration efficacy was also observed in mice treated with MN. However, H&E staining did not reveal any needle puncture in the epidermis (Fig. 1F), this could be due to the recovery and closure of puncture after 12 hours of incubation with Aquaphor as reported by Kalluri et al. (35). Further screening of force of pressure and application time allowed us to optimize the pressure and treatment time. Higher pressure induced better skin penetration (four-, seven-, and ninefold increment for 0.14, 0.28, and 0.4 MPa, respectively; Fig. 2A). A longer pressure treatment also provided a better penetration. Considering the balance between safety and efficacy, we identified 1 and 5 min of 0.28-MPa treatments for the delivery of dextran (3, 5, 10, and 20 kDa) and NPs (20, 100, 200, and 500 nm) in the mouse models. The results were encouraging, in which both 1- and 5-min pressure treatments improved the concentrations of all four-sized dextrans (Fig. 3, C and D) and NPs smaller than 500 nm (fig. S9) in the dermis. The translatability of temporal pressure was further explored in in vivo rabbit (Fig. 4) and ex vivo porcine (fig. S11) models where improvements in penetration were observed in both scenarios. This pressure-aided transdermal delivery is due to the formation of micropores (2.61 μm2 immediately after pressure and 3.21 μm2 12 hours after pressure) in the skin epidermis (Fig. 5 and fig. S12). In addition, there was an increase in Cx43 gap junctions and a decrease in occludin between epidermal cells (Fig. 5 and fig. S13). On the other hand, no significant changes were observed in the junction proteins after 5 min of pressure (fig. S12), this is due to the short time frame for the changes in protein expression to occur. Future works for in-depth molecular studies are proposed toward the end of this discussion.

The clinical indication of this methodology was explored with the insulin delivery model (fig. S15). By applying temporal pressure for 1 min, mediocre effect was observed over a 3-hour period (fig. S16). On the other hand, a simple extension of the pressure treatment to 5 min significantly improved the delivery efficacy. The blood glucose was further monitored for 5 hours (Fig. 6). Mice treated for 5 min with 0.28 MPa showed a drop of blood glucose to ~35% of initial state for 5 hours (Fig. 6B), without any hypoglycemic shock (fig. S17C). After 5 hours, 160 mU/liter of insulin was found in circulation (Fig. 6C), accounting for 13.4% of topically applied insulin (fig. S17D). Unfortunately, similar regulation of blood glucose was not observed in diabetic mice if 1 U of insulin was topically applied after 5 min pressure treatment (fig. S19A). To address this issue, the topically applied insulin concentration was raised 10-fold to increase the amount of insulin entering the blood stream. The insulin was found to be 1100 mU/liter at the second hour, approximately 10-fold of the amount of insulin in mice with 1 U of insulin topically applied (fig. S18D). This effectively decreased the blood glucose in diabetic mice (fig. S19A).

We further benchmarked our methodology with the conventional injection and MN-aided topical deliveries (Fig. 7 and fig. S19B) (36, 37). Looking at the glucose regulation, pressure-aided delivery was as effective as the MN pretreatment plus topical application and subcutaneous injection. All three strategies reduced the blood glucose levels to single digit numbers in 1 hour. For dissolvable MN-aided delivery, the glucose level decreased at a slower speed and reached the single-digit level around 6 hours. When we examined the insulin level in the blood, subcutaneous injection and pressure-based delivery significantly increased the insulin level from two digit numbers to three- and four-digit numbers in the first and sixth hours, respectively. MN pretreatment plus topical application brought the insulin level to two-digit numbers in the first hour and above 100 mU/liter in 6 hours. These suggest that the timely regulation of blood glucose was due to the effective delivery of insulin into the circulation in the injection, pressure- and MN-aided delivery. Although dissolvable MNs have been shown to be efficient in controlling the blood glucose (31, 38), their effect was slower, which might be due to the entrapment by the hyaluronic acid (HA) matrix (31) or aggregation and degradation of insulin monomers during the MN fabrication process (3941). Last, the subcutaneous injection and pressure-aided delivery provided the highest insulin concentrations in circulation. This suggests that pressure-aided delivery is more effective than MNs to bring topically applied molecules into the circulation.

While we demonstrated the efficacy of temporal pressure in delivering molecules and NPs across the skin barrier in a noninvasive manner, we realized that jet injectors had been used since 1990 to deliver medications (18). Despite its accomplishment, it was not widely accepted in routine practices due to risk of infection, pain, or injury of the operator’s finger. Furthermore, technical difficulties such as clogging of the injector, splash, and splatter are among other factors that discouraged health care professionals from using them. In terms of the amount of pressure applied to the skin to achieve effective TDD, currently, in the market, Dermojet produces 1420 psi (equivalent to 9.8 MPa), and Madajet produces 1800 psi (equivalent to 12.4 MPa). Several studies had reported it to be painful (42) with high chance of causing harm to the skin (43). However, our study demonstrated that 0.28 MPa was sufficient to deliver molecules and NPs across the skin and into the systemic circulation, with no pain or major damage observed in the skin. Another similar technology is the use of a pressure wave, generated by intense laser radiation, to permeabilize the SC and cell membrane (44). Lee et al. (45) performed a similar study to deliver insulin transdermally using a pressure wave. The amount of pressure produced was 730 bar (equivalent to 73 MPa) for 100 ns to 1 μs to achieve blood glucose drop for 3 hours. While this technology is promising, the cost of the tool and the replacement of its energetic materials that produce pressure wave are high. In contrast, our study uses cheap magnets to produce the pressure for effective transdermal delivery. To enhance the user’s experience, we aim to develop a wearable device, incorporated with microelectronics to produce temporal pressure.

Impairing skin barrier properties might induce bacterial infection (18). This can occur when the transdermal delivery system induces micropores as part of disrupting the skin barrier for enhancing drug delivery. Technologies using such strategies include MNs and thermal ablation. Studies have shown that MNs with a length of 770 μm can create micropores with a surface diameter of ~71 μm and with a depth of ~153 μm (35). Thermal ablation creates pores of up to 50 μm in depth, reaching into the epidermis (46). Since our technique similarly induces the formation of micropores, there is slight chance of infection. However, we would like to emphasize that temporal pressure creates a pore size of ~3.2 μm2. The formation of small pores can minimize the likelihood of bacteria in the skin. This is supported by the absence of immune adaptive response in the histological study. Furthermore, this pore formation is reversible, and the induced damage is far less than that from needle or MN injection. In addition, subsequent topical cream can be formulated to contain antibacterial substances for additional safety.

In terms of pain and potential skin damage, current jet injectors produced pressure of approximately 9.8 to 12.4 MPa, resulting in a painful experience for some users. In this study, the reported pressure is 0.28 MPa (~44 times lower than jet injectors); therefore, the concern of pain in this study is minimized. Regarding potential skin damage, we performed similar experiments (0.28 MPa, 5 min) on ex vivo porcine skin (fig. S11), with the aim of observing potential skin damage. No notable damage was observed in the porcine skin, suggesting the compatibility of temporal pressure. It was also worth mentioning that this study trialed adding pressure after topical drug (fluorescent NPs) application. However, the applied pressure forced the topical drug out of the intended application site, which resulted in lesser topical drug in contact with the temporally pressurized skin and lesser drugs diffusing across the skin.

We are not only excited by the above discovery but also realize that there are a few questions to be answered before its clinical application. First, the optimized pressure and time (0.28 MPa for 1 and 5 min) in this study was only catered to mice. The optimized parameter was also shown to be effective in rabbit (Fig. 4) and porcine skin (fig. S11). In the near future, more resources would be allocated to optimize pressure parameters in rabbit and porcine skin. Second, while 200-nm NPs were used in initial proof study for temporal pressure (Figs. 1, 2, and 4), the molecule sizes study focused on demonstrating the efficacy of temporal pressure on current available drugs, which are majority measured in Dalton (therefore, the use of fluorescent Dextran and insulin as proof of concept). The study of NP sizes (figs. S8 and S9, n = 1) proved to be effective with temporal pressure as well, and we plan to explore this in the near future using a range of NPs loaded with hydrophilic drugs, therapeutic nucleic acids, etc. Third, the observed micropores in Fig. 5A were present 12 hours after temporal pressure application. This interesting observation raised a possibility where future studies could investigate whether drugs could be reapplied without the need for pressure retreatment within this 12-hour time frame where micropores remained in the epidermis.

Next, this study was performed extensively on rodents due to their ready availability, relatively low cost, easy to handle, and provide a suitable area of skin for permeation studies (29, 47). Rodents are also used in many dermal preclinical studies (26, 28, 48). Despite these advantages, rodents differ anatomically and physiologically from humans (25), they have loose skin and a thinner epidermis compared to human skin (29, 47, 49), and they lack apocrine sweat glands and rete ridges/dermal papillae, both of which are found in human skin (50). In addition, there is a gender difference in rodent skin, male skin is 40% stronger due to a much thicker dermis, while female skin exhibits a thicker epidermis and hypodermis (51). In terms of observing skin damage after temporal pressure application, rodents have a panniculus carnosus muscle, which is absent in humans and pigs, and can contribute to the wound healing contraction process in rodents (52). Minimal damage to the skin was observed. Furthermore, the effective transdermal delivery of insulin may be attributed to the rodents’ rudimentary vasculature network, which consists of one primary network above the panniculus carnosus muscle and a second network below the adipose tissue (49). Future studies of temporal pressure can be carried out in Guinea pigs, which exhibits a thicker epidermis than the rodents, similar to humans (53) and similar dermal blood vessel densities to that of humans (54). In addition, the pig is considered an excellent model for TDD due to similarities in anatomy, physiology, and immunology to humans (53), e.g., similar epidermis thickness, tight-skinned structure (29), and similar number, size, and distribution of dermal blood vessels (25, 53).

Furthermore, it is crucial to study how Cx43 gap junction protein and occludin tight junction protein play their roles in permeability. While only 0 min and 12 hours after 5 min of temporal pressure were covered in this study, it will be important to study the timeline for the skin permeation in relation to the altered expression of junction proteins. Cx43 protein has a half-life of ~2 hours (55), while occludin has a half-life of 4 hours (56). Therefore, an in-depth study can be performed to determine the changes in protein expression, which will potentially improve drug application at optimized time points.

Last, human skin thickness varies between individuals and in different parts of the body; thus, it is necessary to explore the effectiveness of various pressure strengths and time of application on different thickness of human skin and different parts of the body (57). The temporal pressure did effectively deliver insulin into diabetic mice, but there was the hyperthickening of the epidermal layer after daily application for 5 days (fig. S18E). This might be minimized by switching the application sites regularly. Last but not least, daily application of temporal pressure via the magnet might be inconvenient and also pose a problem to sensitive skin. One future development is to have a device that can automatically apply pressure and topical formulation (fig. S20).

Working principle of first-generation device

The device as shown in fig. S20A (first prototype) was designed to apply pressure at the clamp head (made of aluminum) when magnet is placed in the slot as indicated. The device consisted of two aluminum plates, with circular slots in the middle for placement of magnets. The protrusion portion of the aluminum plate act as the clamp head (fig. S20B, red arrow). The aluminum plates are held together by two stainless steel rods, allowing the maneuver of plates along the rods (fig. S20B, blue arrows). A screw was used to separate the clamp heads and the magnets (fig. S2C, black arrow), providing a gap to be placed at skin of interest.


This report introduces a pressure-based transdermal delivery methodology. We discovered that the pretreatment of the skin with pressure would generate microsized pores in the epidermis, which allowed the molecules, NPs, and proteins to diffuse into the deep dermis and vascular system. The pressure and duration were optimized to minimize the damage to the skin while achieving an effective delivery. Using dextran molecules and NPs, we demonstrated effective transdermal delivery of up to 20-kDa molecules and 500-nm NPs. Furthermore, using diabetic mice as a preclinical model, we showed effective insulin delivery for 5 days, without the need for needle penetration. Topically applied insulin (~13.4%) entered the vascular circulation. This concept and its methodology can be further extended to topical application of any formulation after a minute-long pressure treatment and might be acceptable for any patient without the concern of safety and cost.


All chemicals were purchased from Sigma-Aldrich unless otherwise stated. Neodymium magnets of different surface field [1795G (ref number: D81-N52), 3309G (ref number: D82-N52), and 4440G (ref number: D83-N52)] were obtained from K&J Magnetics to produce pressure of 0.14, 0.28, 0.4 MPa, respectively (calculation in table S1). FluoSpheres carboxylate-modified microspheres (20, 100, 200, and 500 nm) (catalog nos. F8786, F8801, F8810, and F8812, respectively) with red fluorescent (excitation, 580 nm; emission, 605 nm) were acquired from Thermo Fisher Scientific (Singapore). According to Thermo Fisher Scientific’s certificate of analysis, these microspheres are suspended in a medium of distilled water with 2 nM sodium azide and have a charge density of 0.3557 meq/g. Rhodamine B–conjugated 3-kDa dextran (catalog no. D3308) was acquired from Thermo Fisher Scientific (Singapore), while dextran of sizes 5, 10, and 20 kDa were acquired from NANOCS (USA) (catalog nos. DX5-RB-1, DX10-RB-1, and DX20-RB-1, respectively). Briefly, water-soluble amino dextran was reacted with succinimidyl ester derivative of Rhodamine B. Following dye addition, unreacted amines on the dextran were capped to yield a zwitterionic dextran and purified through size chromatography to remove nonconjugated dye. All animal studies were performed in compliance with the guidelines set by in Institutional Animal Care and Use Committee (IACUC; reference number: A18025), NTU. Normal bright-field and fluorescence imaging was performed with a LX1 inverted microscope (Olympus) equipped with Retiga-2000R charge-coupled device (CCD) camera. The settings (exposure time, 200 ms; gain, 8) were fixed. Confocal imaging was performed using a Leica SP8 upright confocal microscope (Germany).

Transdermal delivery of NP and dextran with pressure treatment in mouse model

Six-week-old C57BL/6J male mice (InVivos Pte. Ltd., Singapore) were randomly grouped and anaesthetized using 2% isoflurane in oxygen. Next, the mice were shaved using Moser ChroMini Pro and followed by Veet Hair Removal Cream Sensitive. This is to ensure that the hair does not trap the cream. The dorsal loose skin was then pinched (fig. S1), and magnets (3309G to produce 0.28 MPa) were placed carefully on each side to prevent injury to the mice. Magnets were placed for 1 min, 5 min, 0.5 hour, and 1.5 hours to produce the intended pressure. The procedures hereafter are similar for both dextran molecules and NPs topical application. After pressure application, 0.05 mg of dextran molecules or 0.2 mg of carboxylate-modified NPs was mixed with Aquaphor 1:1 ratio (w/v). Aquaphor Healing Ointment is a commercial formulation for dry and cracked skin. It contains 41% petrolatum, panthenol, and glycerin to moisturize, nourish, and protect the skin. Magnets were removed carefully by sliding them apart after stipulated time. The formulation containing NPs or dextran was then topically applied on the skin marks caused by the pressure, followed by Tegaderm and Coban application to secure the mixture overnight. The next day, mice were anaesthetized, and Tegaderm and Coban were removed. The excess topical mixture was removed by gently wiping with gauze and phosphate-buffered saline (PBS). Mice were imaged with IVIS. Thereafter, mice were euthanized by CO2 asphyxiation, and the skin was excised and fixed in 10% formalin solution for histological studies.

Transdermal insulin delivery on normal mice

Experimental test group sizes were three mice per treatment, to ensure a balance between sufficient replication of results and a reduction in mouse number. Mice were fasted for 12 hours before treatment. Then, the mice were anaesthetized and shaved as reported in the previous section. After pressure application using magnets (3309G to produce 0.28 MPa), topical insulin (1 U of insulin, Mixtard 30 HM Penfill mixed with Aquaphor, 1:1 ratio, w/v) was applied onto pressure marked skin. Mixtard 30 HM Penfill is a suspension of active human insulin containing 30% soluble insulin and 70% isophane insulin. The topical application was secured using Tegaderm and Coban. Blood glucose was measured using One Touch UltraMini by snipping off a small portion of the tail. Blood glucose levels were measured at time 0, 0.5, 1, 2, 3, 4, and 5 hours.

Transdermal insulin delivery on STZ-induced diabetic mice

Male C57BL/6J mice were weighed at week 7 and marked with an ear puncher. They were fasted overnight (~16 hours), followed by measurement of fasting blood glucose (FBG) using a glucometer (Accu-Chek Guide Wireless Blood Glucose Meter, Roche). Mice with FBG of >11.1 mM were excluded from the injection. One day before STZ administration, mice were fasted overnight. STZ was prepared in a concentration of 0.02 g/ml of buffer containing 0.1 M citrate buffer. STZ was then administered by intraperitoneal injection at 50 mg/kg to the STZ-induced diabetic mice, while only citrate buffer was injected into mice for the control group. This procedure was repeated each day for 5 days, followed by 4 weeks of monitoring before mice were tested again for FBG. Mice with blood glucose greater than 21 mM were considered successfully induced with diabetes.

In the transdermal delivery experiments, diabetic mice were anaesthetized and shaven as reported in the previous section. Next, 10 μl of 10 U of insulin (Mixtard 30 HM Penfill) was mixed with Aquaphor 1:1 (w/v) ratio before application. After removing the magnets (0.28 MPa) for 5 min, the mixture of insulin was then topically applied on the skin marks caused by the pressure, followed by Tegaderm and Coban application to secure the mixture while monitoring the blood glucose at 0, 0.5, 1, and 2 hours.

To mimic a clinical setting, the above-mentioned study was performed for a period of 5 days. The diabetic mice were fasted for 4 hours before each day’s experiment. As a comparison, topical application after pressure treatment was performed alongside subcutaneous injection of 2 U insulin on the dorsal portion, topical application with 10 U insulin, transdermal delivery with dissolvable MNs for 3 min, and control. In another group, drug-free solid PMMA MNs were thumb-pressed on dorsal skin for 3 min before the insulin-Aquaphor formulation was applied as mentioned above. Blood glucose was measured at 0, 1, 2, 4, 6 hours of the first and second days, and 0, 2, and 6 hours for subsequent days. Plasma insulin level was measured by drawing the blood using capillary blood collection tubes, and insulin levels were measured using a human insulin ELISA kit (Crystal Chem).

Fabrication of insulin-containing dissolvable MN

Insulin-containing dissolvable MNs were made via micromolding. Poly(dimethylsiloxane) (PDMS) micromolds were first prepared by pouring PDMS mixture [SYLGARD 184 Silicone Elastomer Kit, Tat Lee Engineering Pte. Ltd.; 10:1 (w/w) ratio of prepolymer to curing agent] into a stainless-steel master (consisting of 10 × 10 arrays of sharp pyramids needles, with height of ≈1000 μm, interspacing of ≈700 μm, and base width of ≈300 μm; Micropoint Technologies, Singapore), followed by degassing with a vacuum oven and curing at 70°C for 3 hours. Subsequently, this micromold was filled with HA solution [0.5 g/ml; sodium HA (Mw, 3 kDa), Freda Biochem Co. Ltd. (China); ≥95.0% purity] containing 10 U insulin through centrifugation method (4000 rpm, 3 min). Following overnight drying at room temperature, a backing layer of bare-HA solution (0.5 g/ml) was added as supportive base layer. The final patch was peeled off from the micromold after overnight drying and stored at 4°C in dehumidified conditions.

Solid MNs

Solid, nondissolving PMMA MNs were obtained from Micropoint Technologies Pte. Ltd., Singapore. Briefly, PMMA (Mw, 75 kDa) was caste in a mold to have needle height of 1000 μm, in 10 × 10 array with a density of 400 needles/cm2. The individual MNs have a square pyramid shape.

Transdermal delivery in the rabbit model

Pathogen-free New Zealand rabbits were purchased from Prestige World Genetics (PWG) Laboratories Singapore and housed under standard laboratory conditions (22° ± 1°C; relative humidity, 40 to 55%) at PWG. Rabbits were fed with normal rabbit chow and watered ad lib. The in vivo rabbit experiments were performed under the regulations of Good Laboratory Practice Compliance Programme and Association for Assessment and Accreditation of Laboratory Animal Care International (IACUC reference number: BN16098). Rabbits were anaesthetized using ketamine (60 mg/kg) and xylazine (5 mg/kg). The dorsal hair was removed by shaving and depilation, followed by pressure application using magnets for 1 or 5 min, followed by 12 hours of topical drug application. Given the bigger dorsal skin surface area as compared to mice, four different sets of time periods were performed for each rabbit. Positions of each set of time periods were randomized on the rabbit dorsal skin. Carboxylate-modified NPs (0.2 mg of 200 nm) were mixed with Aquaphor 1:1 (w/v) ratio before application. The mixture was then topically applied to the skin marks caused by the pressure, followed by Tegaderm and Coban application to secure the mixture for 12 hours.

Transdermal delivery in ex vivo porcine skin model

Porcine skin was purchased from SingHealth Experimental Medicine Centre. The experiment was performed as soon as the porcine skin was harvested. Similar to the procedures in mice and rabbits, the skin was shaved using Moser ChroMini Pro and followed by Veet Hair Removal Cream Sensitive. Next, the skin was cut into dimensions of 1 cm by 1 cm to facilitate the different treatment groups. Following which, pressure and MN were applied for 5 min, followed by topical NP application (0.2 mg of 200-nm carboxylate-modified NPs) and overnight incubation in a 37°C incubator. Thereafter, the excess NPs were removed by wiping, followed by IVIS imaging.

Molecular study on the impact of temporal pressure on the skin barrier

To study the molecular changes in the skin barrier enabling the penetration of molecules and NPs, the mice were shaved and depilated as stated in the previous section. Two sets of experiments were performed. First, the dorsal loose skin was pinched (fig. S1) with 0.28 MPa for 5 min, and the mouse was euthanized by CO2 immediately, followed by harvesting of skin. In the second set of experiments, different strengths of magnets (0.14, 0.28, and 0.4 MPa) were applied for 5 min, followed by application of a mixture of PBS in Aquaphor in a 1:1 (w/v) ratio on the pressurized skin to mimic the use of aqueous drug. Next, Tegaderm and Coban were applied to prevent the mice from removing the topical application. After an overnight (12 hours) application, the mice were euthanized by CO2 asphyxiation, and the skin was excised and fixed in 10% formalin solution for histological and immunostaining studies.

Histological sectioning and H&E staining

For fluorescence microscopical observation of transdermal delivery of dextran molecules and NPs into the skin, the excised skin samples were cryosectioned to minimize quenching the fluorescence in paraffin processing. Briefly, skin tissues were embedded in optimum cutting temperature (Sakura Finetek) first. Then, they were sectioned at 10 μm using Leica CM3050 S Cryostat and stained with H&E under standard protocols. The H&E-stained cryosections can be found in Figs. 1E, 2G, and 4B and figs. S2C, S11D, S14A, S17E, and S18E.

To look for microscopic histological changes, samples were sectioned using paraffin-embedded tissue. Briefly, skin tissues were transferred into 70% ethanol after fixation and processed for paraffin embedding using an automated tissue processor (Leica Tissue Processor HistoCore Pearl). Paraffin-processed tissues were immediately embedded into paraffin blocks and sectioned at 5 μm. Hematoxylin (catalog no. 3801570, Leica) and eosin (catalog no. 3801615, Leica) histological stain was performed and mounted using Organo/Limonene mount (catalog no. O-8015, Sigma-Aldrich). Slides were allowed to dry overnight and imaged using an Axioscan.Z1 slide scanner (Zeiss). The H&E-stained paraffin sections can be found in Fig. 5 and figs. S12A and S13A.


Paraffin-embedded skin sections were first deparaffinized and rehydrated using decreasing concentrations of ethanol (100, 95, and 70%) followed by water and then 1× PBS. Tissues were stained overnight using rabbit anti-Cx43 primary antibody (C6219, Sigma-Aldrich) of 1:1000 dilution after being permeabilized and blocked. This was followed by 1-hour incubation of 1:500 goat anti-rabbit AF488 secondary antibody (A11008, Thermo Fisher Scientific, Singapore). Tissue sections were counter stained with 4′,6-diamidino-2-phenylindole (DAPI). To stain occludin, tissues were deparaffinized and processed with enzymatic antigen retrieval using proteinase K. Similar to Cx43 staining, tissues were stained overnight using rabbit anti-occludin primary antibody (71-1500, Thermo Fisher Scientific, Singapore) of 1:100 dilution after being permeabilized and blocked. This was followed by 1-hour incubation of 1:500 goat anti-rabbit AF555 secondary antibody (A21428, Thermo Fisher Scientific, Singapore) and subsequently counter-stained with DAPI.

Fluorescence and confocal microscopy and image analysis

For all figures (except for Fig. 5 and fig. S11), slides were imaged using an Olympus IX71 inverted microscope. All images were taken using a 10× objective. For quantification of signal, three representative images of each sample were analyzed using ImageJ. Data were plotted and statistically analyzed using one-way analysis of variance (ANOVA) with post hoc Tukey test, unless otherwise stated. For Fig. 5 and fig. S11, slides were imaged using a Leica SP8 upright confocal microscope. All images were taken using a 40× objective. For quantification of signal, three z-stack images of 3 μm were taken in the relevant regions and analyzed using ImageJ. Data were plotted and statistically analyzed using Prism6 software. Kruskal-Wallis test followed by Dunn’s post hoc test was performed, and statistical significance was achieved when *P < 0.05.

Statistical analysis

Statistical comparisons between groups, unless specified, were determined by one-way ANOVA with post hoc Tukey test using a web-based statistics calculator from: For all tests, *P < 0.05 was considered statistically significant. For Fig. 5 and fig. S11, data were plotted and statistically analyzed using Prism6 software. Kruskal-Wallis test followed by Dunn’s post hoc test was performed, and statistical significance was achieved when *P < 0.05.


Supplementary material for this article is available at

This is an open-access article distributed under the terms of the Creative Commons Attribution-NonCommercial license, which permits use, distribution, and reproduction in any medium, so long as the resultant use is not for commercial advantage and provided the original work is properly cited.


Acknowledgments: Funding: C.X. acknowledges the funding support from Singapore Agency for Science, Technology and Research (A*STAR) Science and Engineering Research Council Additive Manufacturing for Biological Materials (AMBM) program (A18A8b0059) and internal grant from City University of Hong Kong (#9610472). D.L.B. acknowledges the funding support from A*STAR under its Industry Alignment Fund–Pre-Positioning Programme (IAF-PP): (1) Wound Care Innovation for the Tropics Programme, Singapore (WCIT): (H18/01/a0/0I9) and (H17/01/a0/0C9); (2) The Skin Research Institute of Singapore, Phase 2: SRIS@Novena (“IAF-PP SRIS2 Grant”, H17/01/a0/004). X.W. acknowledges the funding support from the National Medical Research Council Singapore Large Collaborative Grant DYNAMO (NMRC/OFLCG/001/2017); and National Medical Research Council Singapore Large Collaborative Grant TAAP (NMRC/OFLCG/004/2018). Author contributions: D.C.S.L., X.W., D.L.B., and C.X. conceived and designed the experiments. D.C.S.L., R.N.C., M.S.Y.K., C.W., L.E.M., and S.M.A.K. performed the experiments. D.C.S.L., M.S.Y.K., S.M.A.K., X.W., D.L.B., and C.X. analyzed and interpreted the data. D.C.S.L., H.C., X.W., D.L.B., and C.X. wrote the manuscript. All authors read and approved the final manuscript. Competing interest: The authors declare that they have no competing interest. Data and materials availability: All relevant data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. Additional data related to this paper may be requested from the authors, as a source data file that includes data underlying Figs. 1 (C and D), 2 (C to F), 3C, 4C, 5 (D and E), 6 (A and B), and 7 and figs. S3A, S4B, S5 to S11 (B and E), S12 (D and E), 14B, S16, S17 (A to D), S18 (B to D), and S19 (A and B).

Stay Connected to Science Advances

Navigate This Article