The skin provides crucial regulatory and defensive functions1,2. Skin disorders may put patients at risk of cutaneous and systemic infection, trauma and malignancies3. Although traditional therapeutics have limited capabilities to solve these problems, the latest advances in cell therapy bring hope for previously incurable and untreatable diseases4,5. For example, genetically modified epidermal stem cells can be transplanted to regenerate the skin in patients with epidermolysis bullosa6,7, and melanocytes can be transplanted to treat vitiligo8,9. Dendritic cell (DC)-based immunotherapy has proven to be a safe treatment for melanoma10,11. Currently, these therapeutic cells are delivered by either bolus injection for diseases of systemic health (for example, cancer)11 or scaffold transplantation for local symptoms (for example, eczema)12. However, these methods have difficulty in achieving controlled and precise delivery of targeted cells without sacrificing the comfort of patients, which presents challenges for their translation to clinical applications.

Microneedle (MN) technology is a promising option for realizing the controlled and precise delivery of targeted drugs into specific skin layers and other tissues (for example, mucosa and cornea) in a minimally invasive manner13,14,15,16. MNs are composed of a wide range of materials, geometries and spatial arrangements and function as carriers of small molecular drugs, peptides and proteins (for example, insulin, vaccines and antibodies), oligonucleotides and nanomedicines17,18. Unfortunately, none of these existing MN technologies can carry and deliver living formulations, such as therapeutic cells, into skin, without the assistance of extra devices. To carry and deliver cells, MNs have to maintain the viability of loaded cells and have mechanical strength to penetrate the skin.

Here, we introduce an MN patch that is specifically designed for packaging and delivering living cells into skin (Fig. 1). This MN patch, termed cryomicroneedles (cryoMNs), is fabricated by stepwise cryogenic micromoulding of the optimized cryogenic medium and pre-suspended cells of interest in the pre-designed MN mould. CryoMNs can easily pierce skin and deliver loaded living cells into skin.

Fig. 1: Schematic illustration of the transdermal delivery of cells using cryoMNs.
figure 1

The cryoMNs, loaded with cells (top), can easily penetrate into skin epidermis (bottom left). The loaded cells are released when the cryoMNs melt (right). The released cells subsequently migrate, proliferate, secrete specific cytokines and interact with other local cells in the skin (expanded panel).


Selection of cryogenic medium for the fabrication of cryoMNs

Ice is the solid state of water or an aqueous medium in a cryogenic environment and has strong mechanical properties19. Given that ice can be easily moulded to any shape, ice-based MNs can be fabricated easily in cryogenic environments. Furthermore, cells can be preserved in a cryogenic environment with minimal loss of viability if they are pre-suspended in a suitable cryoprotectant medium20,21.

Dimethyl sulfoxide (DMSO) is one of the most widely used cryoprotectants22 and can minimize damage to cell membranes under a cryogenic environment by hindering the crystallization of ice21. We thus selected cryogenic medium supplemented with DMSO to fabricate the cryoMNs. We found that cryoMNs fabricated from cryogenic medium containing more than 2.5% DMSO (vol/vol) could not be peeled from the mould with their structures intact (observed as leftover rhodamine tips in the mould in Supplementary Fig. 1). However, a concentration of DMSO of less than 2.5% resulted in an obvious reduction in cell viability from ~90 to ~10% (Supplementary Fig. 2).

To solve this dilemma, we incorporated a non-penetrating cryoprotectant into the medium to compensate for the loss of cell viability under low concentrations of DMSO23. In contrast to DMSO, non-penetrating cryoprotectants only act at the extracellular surface of cells and can reduce the amount of water in contact with cells by increasing the osmolarity of the medium24. In addition, DMSO pre-dehydrates cells at non-freezing temperatures (for example, room temperature) to their minimal volume with minimized osmotically active water, preventing cells from being injured during freezing23. As a proof of concept, we selected sucrose to combine with DMSO as the cryogenic medium. The viabilities of six types of human cell—HeLa cells expressing red fluorescent protein (RFP-HeLa), keratinocytes (HaCaT), normal dermal fibroblasts (NDFs), mesenchymal stem cells (MSCs), epidermal melanocytes and peripheral blood CD8+ CD45RA+ T cells (T cells)—were evaluated after cryopreservation. As shown in Supplementary Fig. 2, regardless of the concentration, the addition of sucrose substantially improved cell viability, which was obvious when the concentration of DMSO was less than 5%. This phenomenon was universal to all types of cell. No difference in cell viability was noted between 100 mM and 200 mM sucrose in 2.5% DMSO. In the following experiments, phosphate-buffered saline (PBS) supplemented with 2.5% DMSO and 100 mM sucrose was used as the optimized cryogenic medium for fabricating cryoMNs. PBS was used as the base medium because it did not introduce other molecules (such as proteins) that might induce undesirable immune responses in practical applications.

Fabrication and characterization of cryoMNs

The cryoMNs were prepared using a micromoulding method (Fig. 2a), which is the easiest method with which to shape an aqueous solution into an MN structure. Cells of interest were pre-suspended in the optimized cryogenic medium, and the obtained mixture was cast into a negative polydimethylsiloxane (PDMS) mould. Therapeutic cells are typically rare and precious and it is desirable that they are not wasted during device fabrication, so, because the cryoMN base does not enter the skin layers, the cells need only be loaded into the tips of the cryoMNs. To ensure that the cells had entered the tips of the needles, the cells, suspended in the cryogenic medium, were left in the PDMS mould for 20 min without centrifugation to allow the cells to sink into the tips of the needles by gravity. During this 20-min process, the sucrose simultaneously dehydrated the cells23. A sequential, gradient cryogenic process (from −20 °C to −196 °C) was then applied, and the cryoMN patch was formed after solidification. The entire fabrication process took ~8 h.

Fig. 2: Fabrication and characterization of cryoMNs.
figure 2

a, Schematic of cryoMN fabrication. b, Digital microscopy image of the cryoMNs. Scale bar, 1,000 μm. c, Magnified view of b. Scale bar, 1,000 μm. d, Photograph of the cryoMNs patch. Scale bar, 5 mm. eg, Images of a single MN loaded with cells stained with NucBlue Live ReadyProbes reagent. Yellow arrows indicate individual cells. Image g is an overlay of the bright-field image in e and a fluorescent image (f). Scale bars, 500 μm. h, Compressive strength of cryoMNs (right) tested by an Instron tensile meter (left). The blue area bounded by the dotted line highlights the cryoMNs. PS-MNs, polystyrene MNs; PLA-MNs, polylactic acid MNs; PCL-MNs, polycaprolactone MNs; MeHA-CL-MNs, crosslinked hyaluronic acid MNs. Data are presented as mean ± s.d. (n = 10 independent cryoMN samples). Two-tailed Student’s t-test was used for comparing cryoMNs with other conventional polymeric microneedles (****P = 1.31972 × 10−9 versus PS-MNs; ****P = 1.15268 × 10−5 versus PLA-MNs; ***P = 0.0008 versus PCL-MNs; ***P = 0.0002 versus MeHA-CL-MNs). In eg, three images were taken and all experiments were repeated independently twice, with similar results.

The master template was a stainless-steel MN array (10 × 10), with each MN having a height of 1,200 μm and a base width of 400 μm (Supplementary Fig. 3). The resulting cryoMNs had a height of ~900 μm and a base width of ~350 μm (Fig. 2b–d). The dimensional difference between the original template and cryoMNs was due to shrinking of the elastic PDMS during templating15. However, the pyramid structure and sharpness of the needles were well replicated. To visualize the loaded cells within cryoMNs, the model cells (HaCaT) were stained with Hoechst 33342 beforehand. After being immersed in cold PBS, the MN tips detached from the base. The stained cells were observed inside the needle (Fig. 2e–g and Supplementary Fig. 4). Precipitation for 20 min allowed ~80% of the cells to sink into the cavities (Supplementary Fig. 5), and ~800 HaCaT cells were present in one MN tip (Methods). The cell capacity of a single MN depends on the dimension of the MN as well as the size of the cells. The capacities of the six types of cell used in this study ranged from ~4 × 103 to ~1.93 × 105 per single MN tip (Supplementary Fig. 6).

We also examined the mechanical strength of the cryoMNs using a tensile testing machine (Supplementary Video 1). Four types of polymeric MN that are commonly used for transdermal drug delivery (polystyrene MNs (PS-MNs), polylactic acid MNs (PLA-MNs), polycaprolactone MNs (PCL-MNs) and crosslinked hyaluronic acid MNs (MeHA-CL-MNs)15) were used for comparison. The compression strength was calculated from the slope of the stress–strain curve (Supplementary Fig. 7a,b). Specifically, the compressive strength of the cryoMNs was 24.7 ± 3.8 MPa, which is comparable to the compressive strength of the PS-MNs (66.2 ± 10.9 MPa), PLA-MNs (38.8 ± 6.4 MPa), PCL-MNs (17.6 ± 4.1 MPa) and MeHA-CL-MNs (17.7 ± 3.0 MPa) (Fig. 2h). The load fracture force of the cryoMNs was 0.17 N (Supplementary Fig. 7a), which is greater than the minimum average force (0.058 N) required for normal skin penetration25.

Correlation between the skin penetration ability of cryoMNs and their residence time at room temperature

One potential challenge faced by the use of cryoMNs in practical applications is their limited lifetime, because ice will melt in the operating environment (typically at room temperature (RT)). The melting of cryoMNs definitely has a negative influence on their mechanical strength, which determines their skin penetration ability. When the cryoMNs were removed from their cryopreservation environment (−196 °C, liquid nitrogen) and left at RT (in this case RT was ~24 °C), frost appeared immediately on the cryoMNs (Supplementary Fig. 8a and Supplementary Video 2). After 140 s, the needle tips began to melt. When the cryoMNs were placed on a fingertip, the needle tips melted in 60 s (Supplementary Fig. 8b and Supplementary Video 3).

The correlation between skin penetration ability and the residence time at RT was examined with porcine skin, given its recognized similarities to the anatomy of human skin26. As shown in Fig. 3a,b, the skin penetration ability decreased when the cryoMNs were exposed to RT for longer durations, and they lost their skin penetration ability if they stayed at RT for more than 50 s. At a residence time of 40 s, the cryoMNs could pierce through the stratum corneum, which is ~21–26 μm thick26. The cryoMNs incubated at RT for less than 30 s were able to reach the dermal layer (up to 516 ± 76 μm). These results indicate that cryoMNs should be administered to patients immediately after being removed from the storage environment and that the exposure times to RT should be less than 40 s (Fig. 3a,b and Supplementary Fig. 9). During the experiment, we found that this time was sufficient for the successful penetration of cryoMNs.

Fig. 3: Correlation between the skin penetration ability of cryoMNs and their residence time at RT.
figure 3

a, Representative haematoxylin and eosin (H&E) staining images of porcine skin inserted by cryoMNs. The cryoMNs were removed from their cryopreservation environment (−196 °C, liquid nitrogen) and placed at RT (~24 °C) for 0, 10, 20, 30, 40 and 50 s before skin penetration. Scale bar, 500 μm. b, Quantitative data for the influence of residence time on skin penetration depth. Data are presented as mean ± s.d. (n = 3 independent cryoMN samples). Two-tailed Student’s t-test was used to compare the penetration depth after a residence time of 0s and residence times of 30 s and 40 s (NS, P = 0.2746 versus 30 s; **P = 0.0019 versus 40 s). c, Fluorescent images of skin cryosection after application of RFP-HeLa-loaded cryoMNs. The cell nucleus was stained with NucBlue Live ReadyProbes reagent in advance. The needle shape is outlined with dashed lines. Scale bar, 200 μm. In a and c, three images were taken in each experiment and all experiments were repeated independently twice, with similar results.

As a proof of concept, we delivered RFP-HeLa cells into porcine skin by using cryoMNs with a 10-s residence time. Both nuclear staining and RFP signals confirmed the presence of cells in the dermis layer (Fig. 3c). Similar to the observation in Fig. 3a, a prolonged residence time decreased the skin penetration ability of cryoMNs and the depth that the cells could reach (Supplementary Figs. 10 and 11). CryoMNs could be stored at both −196 °C and −80 °C, which is similar to common cryopreservation conditions. However, at RT, the cryoMNs stored at −80 °C melted more quickly than those stored at −196 °C (Supplementary Video 4) and had to be applied within 20 s for successful skin penetration (Supplementary Figs. 9 and 12).

Survival and proliferation of cells released from cryoMNs

The six types of cell (RFP-HeLa, HaCaT, NDFs, MSCs, melanocytes and T cells) were loaded into cryoMNs with the above-mentioned fabrication process. Cell viability was tested after melting these fresh cryoMNs in PBS (37 °C). According to live/dead staining (Fig. 4a and Supplementary Fig. 13), viability varied among the different types. Specifically, the viability was 38.7 ± 2.7% for RFP-HeLa cells, 34.8 ± 6.5% for HaCaT cells, 30.0 ± 2.4% for NDFs, 30.8 ± 2.2% for MSCs, 53.1 ± 2.4% for melanocytes and 21.4 ± 2.2% for T cells (Fig. 4b). In addition, these living cells could proliferate after being released from cryoMNs (Fig. 4c and Supplementary Fig. 14). At six days, the cell population became 13.8-fold for RFP-HeLa cells, 8.5-fold for HaCaT cells, 2.0-fold for NDFs, 3.1-fold for MSCs, 1.7-fold for melanocytes and 1.9-fold for T cells. These data were similar to the proliferation of cells without any processing (Supplementary Fig. 15), suggesting that the fabrication process did not influence cell proliferation. In addition, there was minimal change in the viability of RFP-HeLa cells and melanocytes in cryoMNs after being preserved in a cryogenic environment for one month (Supplementary Fig. 16), suggesting that the cryoMNs are suitable for long-term storage. In addition, the viability and proliferation ability of cells in cryoMNs from −80 °C were similar to those of cells in cryoMNs from −196 °C (Fig. 4a–c and Supplementary Fig. 17). Furthermore, a three-dimensional (3D) skin model was established by coating agarose hydrogel (as dermis) with a parafilm layer (as epidermis), as in a previous study15. After being delivered into the skin model with cryoMNs, MSCs were able to proliferate within 12 days (Supplementary Fig. 18).

Fig. 4: Viability and proliferation of cells released from the cryoMNs after being removed from their cryopreservation environment (−196 °C, liquid nitrogen).
figure 4

a, Live (green)/dead (red) staining images of the cells loaded in cryoMNs. Scale bar, 200 μm. Four images were taken for each group. All experiments were repeated independently three times, with similar results. b, Quantitative data of cell viability calculated from live (green)/dead (red) staining by dividing the number of living cells (green) by the total number of cells. Data are presented as mean ± s.d. (n = 4 biologically independent samples). c, Proliferation of the cells after being released from cryoMNs and cultured for one, three and six days. Data were obtained by using the alamarBlue assay, and the fluorescence intensity obtained on the first day was used as reference. Data are presented as mean ± s.d. (n = 3 biologically independent samples). Two-tailed Student’s t-test was used for comparing the relative fluorescence at day 6 with the relative fluorescence at day 1 in each group (****P = 2.37951 × 10−8 for RFP-HeLa; ****P = 3.94959 × 10−6 for HaCaT; ***P = 0.0001 for NDFs; **P = 0.0096 for MSCs; *P = 0.0104 for melanocytes; ***P = 0.0002 for T cells). d, Schematic of the application of blank cryoMNs and RFP-HeLa-cell-loaded cryoMNs on the backs of mice. e,f, Photographs of the mouse skin during (e) and after (f) application of cryoMNs. The injection site is bounded by the black dotted square. g, Fluorescent images of cryosections of untreated mouse skin and skin treated with blank cryoMNs and RFP-HeLa-cell-loaded cryoMNs. Skin can induce very light autofluorescence. Scale bar, 200 μm. Two images were taken for each group and all experiments were repeated independently twice, with similar results. h, In vivo fluorescent imaging of mice treated with blank cryoMNs (black circles on the left of each mouse) and RFP-HeLa-cell-loaded cryoMNs (black circled area on the right side of each mouse) on days 0, 1, 3, 7 and 14. i, Quantification of fluorescence intensity on the skins. Data are normalized to the fluorescence intensity at day 0. Data are presented as mean ± s.d. (n = 3 independent animals). Two-tailed Student’s t-test was used for comparing the fluorescence at day 3 with the fluorescence at day 14 (*P = 0.0335).

Next, RFP-HeLa-loaded cryoMNs were applied on the back of mouse skin through a thumb press (Fig. 4d–f). Fluorescent imaging of the skin cryosections (Fig. 4g and Supplementary Fig. 19) revealed that RFP-HeLa cells were present in the dermis with depths ranging from ~20 to 200 μm (the thickness of the mouse epidermis is ~15 μm)27. Blank cryoMNs and RFP-HeLa-loaded cryoMNs were applied on two sides of the mouse back to monitor the fluorescent signals in mouse skin (Fig. 4h). Following transplantation, the RFP signal was consistently observed on the sites treated with RFP-HeLa-loaded cryoMNs, whereas no signal was present on the sites treated with blank cryoMNs (Fig. 4h,i). After one day of implantation, the fluorescence intensity was reduced ~75% compared with that on day 1, which should be due to the death of some cells at the initial implantation and the loss of the cells that were on the surface of the skin. The implanted cells displayed slight proliferation from day 3 to day 14 and continued to maintain viability over 14 days inside the skin of the mice (Fig. 4h,i).

Vaccination by cryoMNs carrying antigen-loaded DCs

DCs are important antigen-presenting cells, and DC vaccines have been widely applied in immunotherapy for cancer treatment11,28,29. Skin is recognized as an ideal target site for vaccination because skin is a highly immunocompetent tissue containing a large population of resident antigen-presenting cells30. Here, we evaluated the therapeutic effect of DC-loaded cryoMNs for cancer vaccination as a proof of concept. Specifically, bone marrow-derived DCs were pulsed with 50 μg ml−1 ovalbumin (OVA; as a model antigen) to obtain OVA-pulsed DCs (OVA-DCs), which were loaded into cryoMNs (Fig. 5a). The levels of the surface marker CD86 and major histocompatibility complex class II (MHCII) were used to confirm the activation and maturation of DCs (Supplementary Fig. 20a,b). Lipopolysaccharide (LPS)-treated DCs (LPS-DCs) served as a positive control. The viability of the OVA-DCs in cryoMNs was 71.4 ± 1.4% (Supplementary Fig. 20b,c). In addition, the viability of the OVA-DCs in cryoMNs did not change even after one month of storage in liquid nitrogen (Supplementary Figs. 20c and 21).

Fig. 5: Vaccination with OVA-DC-loaded cryoMNs (OVA-DC-cryoMNs).
figure 5

a, Schematic of loading OVA-DCs in cryoMNs. b, The protocol of vaccination with OVA-DC-cryoMNs in mice. Vaccinations with subcutaneous (s.c.) and intravenous (i.v.) injections of OVA-DCs were used for comparison. c, Photograph of the administration sites of OVA-DC-cryoMNs on the back skin of mice (left) and an enlarged photograph of the same mouse (right). The sites are bounded by black dashed lines. df, Representative plots (d) and quantification (e,f) of the percentage of CD11c+CD86+ DCs (e) and CD11c+MHCII+ DCs (f) in draining lymph nodes (dLNs) excised from mice in different treatment groups, analysed by flow cytometry. g, In vitro proliferation of extracted splenocytes restimulated with 50 μg ml−1 antigen (OVA). The fluorescence intensity was obtained using an alamarBlue assay. h, Secretion level of IFN-γ in the culture supernatants after 48 h of culture. i, Determination of cytotoxic T lymphocyte (CTL) activity in vitro. Effector cells (splenocytes) and target cells (B16-OVA cells) were co-cultured at different ratios. ‘E:T’ refers to the ratio of effector cells (E) and target cells (T). j, Average tumour growth of mice bearing established B16-OVA melanoma tumours after different vaccination treatments. k,l, Weight (k) and photographs (l) of excised tumours at day 26 from five different mice (left to right) in each treatment group. In ek, data are presented as mean ± s.d. (n = 5 independent animals). Two-tailed Student’s t-test was used to compare vaccination with OVA-DC-cryoMNs with s.c. injection and i.v. injection. In e, ***P = 0.0009 versus s.c. injection, **P = 0.0084 versus i.v. injection. In f, *P = 0.0160 versus s.c. injection, **P = 0.0091 versus i.v. injection. In g, ****P = 2.54937 × 10−6 versus s.c. injection, ****P = 2.51656 × 10−6 versus i.v. injections. In h, ***P = 0.0003 versus s.c. injection, ****P = 2.19911 × 10−6 versus i.v. injection. In i, at E:T = 100:1, ***P = 0.0009 versus s.c. injection, ***P = 0.0004 versus i.v. injection. In j, at day 26, ****P = 5.50942 × 10−5 versus s.c. injection, ***P = 0.0005 versus i.v. injection. In k, ***P = 0.0006 versus s.c. injection, ***P = 0.0001 versus i.v. injection.

We first optimized the administration frequency of OVA-DC-loaded cryoMNs (OVA-DC-cryoMNs) for cancer vaccination in mice. The mice received intradermal vaccination with four patches of OVA-DC-cryoMNs (1 × 105 OVA-DCs per patch) once per week, twice per week and three times per week over a four-week period (Supplementary Fig. 22a). At day 28, the draining lymph nodes (dLNs) were excised, and the maturation of DCs was examined based on the expression of co-stimulatory molecules, including MHCII and CD86. Mice receiving two vaccinations per week exhibited a higher percentage of CD11c+CD86+ DCs (4.21 ± 0.16%) and CD11c+MHCII+ DCs (4.09 ± 0.10%) in the dLNs than the other groups (Supplementary Fig. 22b–d). Splenocytes extracted from mice treated twice per week and three times per week showed better proliferation and higher secretion of interferon gamma (IFN-γ) after restimulation with OVA (Supplementary Fig. 22e,f). These data indicate that vaccination with OVA-DC-cryoMNs twice per week is adequate to induce an immune response.

Next, the optimal dosage was assessed. Following the optimized frequency of vaccination, the mice were administered two patches (2 × 105 OVA-DCs in all), four patches (4 × 105 OVA-DCs in all) and six patches (6 × 105 OVA-DCs in all) of OVA-DC-cryoMNs over the same four-week period (twice per week) (Supplementary Fig. 23a). Analysis of excised dLNs at day 28 revealed that mice vaccinated with 4 × 105 OVA-DCs resulted in a higher percentage of CD11c+CD86+ DCs (6.75 ± 0.08%) and CD11c+MHCII+ DCs (7.82 ± 0.13%) than the other groups (Supplementary Fig. 23b–d). In addition, vaccination with 4 × 105 OVA-DCs and 6 × 105 OVA-DCs via cryoMNs induced better proliferation of splenocytes and higher secretion levels of IFN-γ compared to the other groups (Supplementary Fig. 23e,f). According to these results, a dosage of 4 × 105 OVA-DCs was selected as the optimal dosage, which involves fewer cells without compensation for the therapeutic effect.

After identifying the optimal administration frequency (twice per week) and dosage (4 × 105 OVA-DCs), we compared the therapeutic efficacy of vaccination with OVA-DC-cryoMNs to conventional vaccination methods, namely, s.c. and i.v. injection of OVA-DCs (Fig. 5b). For vaccination with OVA-DC-cyroMNs, each mouse was treated with four patches (that is, the optimized dosage) during one vaccination (Fig. 5c). On day 28, vaccination with OVA-DC-cryoMNs induced 3.59 ± 0.26% CD11c+CD86+ DCs and 3.60 ± 0.64% CD11c+MHCII+ DCs in the dLNs, and these values greatly increased compared with vaccination by s.c. (2.55 ± 0.37% and 2.64 ± 0.35%, respectively) and i.v. injection (2.46 ± 0.68% and 2.40 ± 0.45%, respectively) (Fig. 5d–f). After two days of culture, splenocytes from mice vaccinated with OVA-DC-cryoMNs displayed faster proliferation (Fig. 5g) and secreted higher levels of IFN-γ (Fig. 5h) compared with those from mice vaccinated with s.c. and i.v. injection. The OVA-specific cytotoxic T lymphocyte (CTL) lysis of splenic T lymphocytes from vaccinated mice to the B16 melanoma cell line transfected with ovalbumin (B16-OVA) depended on the ratio of the effector and target cells (Fig. 5i). OVA-DC-cryoMNs induced significantly greater lysis efficiency than s.c. and i.v. injection of OVA-DCs (Fig. 5i). Collectively, vaccination with OVA-DC-cryoMNs could induce more potent antigen-specific immune responses than vaccination with s.c. and i.v. injection.

Finally, we studied the efficacy of tumour prevention of vaccination with OVA-DC-cryoMNs by inoculating mice with B16-OVA cells four days after the final vaccination. All groups vaccinated with OVA-DCs displayed obviously delayed tumour growth (Fig. 5j). On day 26 post tumour inoculation, the tumours from the group vaccinated with OVA-DC-cryoMNs were much smaller, with a volume of 133.3 ± 18.8 mm3 and weight of 57.8 ± 23.7 mg, compared to those from the groups with s.c. injection (240.7 ± 24.6 mm3 and 141.1 ± 24.3 mg) and i.v. injection (375.37 ± 95.2 mm3 and 137.4 ± 9.4 mg) (Fig. 5k,l and Supplementary Fig. 24). This finding suggests that vaccination with OVA-DC-cryoMNs displayed notably stronger antitumorigenic ability compared with vaccination with two conventional standard methods, namely, s.c. and i.v. injection.

Safety and biocompatibility evaluation

To study the biocompatibility of the cryoMNs, they were inserted into the back skin of mice through a thumb press and removed after 10 min. A conventional polymeric MN patch made of hyaluronic acid (HA) (HA-MNs) was applied onto the skin of the same mouse for comparison. HA-MN patches have been widely used for the transdermal delivery of therapeutic agents due to their recognized biocompatibility and safety18,31. After removal, no obvious skin damage was noted at the administration sites of either cryoMNs or HA-MNs. The microholes were clearly observed immediately after application but became invisible within 10 min and completely disappeared after 30 min (Fig. 6a). This fast resealing of skin is essential to prevent the entry of pathogenic microbes or any toxic substances and further reduce the risk of infection. In addition, skin treated with both cryoMNs and HA-MNs did not exhibit any obvious erythema or oedema (Fig. 6a).

Fig. 6: In vivo biocompatibility and safety studies of cryoMNs.
figure 6

a, Recovery of the skin after application of cryoMNs. MNs made of hyaluronic acid (HA-MNs) were used for comparison. The microholes gradually disappeared within 30 min. b, Representative images of H&E staining of skin 30 min or 24 h after MN application. Scale bar, 200 μm. c, Representative images of skin stained with TUNEL assay (green) and Hoechst (blue) after MN application. Skin samples treated with DNase served as a positive control. Scale bar, 200 μm. d, Level of alanine aminotransferase (ALT) in livers of mice after administration of OVA-DC-loaded cryoMNs using different administration frequencies over a four-week period (untreated, one time per week, two times per week and three times per week). Data are presented as mean ± s.d. (n = 5 independent animals). Two-tailed Student’s t-test was used for comparing vaccination with OVA-DC-cryoMNs three times per week with the untreated group (NS, P = 0.9709). e, Weight of mice administered OVA-DC-loaded cryoMNs at different administration frequencies. Data are presented as mean ± s.d. (n = 5 independent animals). In b and c, three images were taken for each group and all experiments were repeated independently twice, with similar results.

H&E staining revealed no obvious inflammatory cell infiltration or pathophysiological response at 30 min and 24 h post-administration of cryoMNs (Fig. 6b and Supplementary Fig. 25). In addition, we stained the skin tissue with the in situ terminal deoxyribonucleotidyl transferase (TDT)-mediated dUTP-digoxigenin nick end labelling (TUNEL) assay and did not observe noticeable cell apoptosis in the skin treated with either cryoMNs or HA-MNs (Fig. 6c and Supplementary Fig. 26).

Finally, frequent administration of cryoMNs neither induced any obvious hepatic damage (Fig. 6d) nor resulted in any weight loss in the mice (Fig. 6e). All the major organs looked normal after application of cryoMNs (Supplementary Fig. 27).


Traditional cell delivery methods mainly include surgical intervention32 and injection with a hypodermic needle33. These methods are associated with considerable disadvantages, including pain or discomfort, scarring, potential risk of infection, stick injuries from sharp waste and the need for skilled healthcare professionals. By contrast, cell delivery with cryoMNs is minimally invasive, generates no sharp hazard, and can be performed by end users with minimal expertise. In MN-based technologies, the packaging and delivery of living therapeutic formulations is one of the most important applications. For example, Bacille Calmette–Guérin (BCG) bacilli are loaded in dry MN patches as vaccines for tuberculosis prevention34,35. Cardiac stromal cells (CSCs) are integrated with the back patch of MNs, which serve as pathways allowing regenerative factors secreted by the CSCs to be released into the injured myocardium36. However, all these existing MN technologies are unsuitable for the transdermal delivery of mammalian cells. So far, the only reported MNs for cell delivery are hollow MNs, which are employed for intradermal infusion of melanocytes into a specific skin layer37,38. However, hollow MNs have intrinsic limitations, including the plugging risk of narrow lumen and the breaking risk of a needle inside skin. Using cryoMNs can avoid these drawbacks. In addition, therapeutic cells can be directly encapsulated in cryoMNs during manufacture and remain viable after long-term storage (Supplementary Figs. 16, 20d and 21). Therefore, complex and redundant procedures, including cell harvesting and the preparation of cell infusing solution, can be circumvented during each administration. Moreover, cell delivery with cryoMNs does not need the assistance of additional equipment to provide infusion pressure, whereas such equipment is usually indispensable during the application of hollow MNs.

In the current design, DMSO was used as one of the raw materials for fabricating the cryoMNs. DMSO has been widely used in pharmaceutical formulations. It can be metabolized to dimethyl sulfone (DMSO2) and dimethyl sulfide (DMS), and these metabolites can be readily excreted from the body, mainly via urine, without any residual accumulation. We evaluated the toxicity of DMSO in several types of skin cell and found that the cell viability remained greater than 80% after 24-h incubation with 2.5% (vol/vol) DMSO (Supplementary Fig. 28). DMSO is generally regarded as non-toxic below 10% (vol/vol)39. In addition, we calculated that the injected dose of DMSO per patch (100 arrays) was ~1.1 × 10−2 mg, which is much less than its median lethal dose (LD50)40. Of note, this formulation of cryogenic medium is not unique and can be further developed by integrating state-of-the-art biotechnologies, such as new freezing protocols23 and cell encapsulation technology41. Our results showed that the application of cryoMNs did not cause localized or systematic abnormalities in mice (Fig. 6). In addition, two volunteers did not report any pain or harsh discomfort when placing cryoMNs on their finger skin without penetrating the skin (Supplementary Fig. 8b and Supplementary Video 3). Previous studies have demonstrated that cooling the skin can reduce patient discomfort during injection42. Thus, cryoMNs could also function as local anaesthetics during administration. However, further detailed evaluation of safety in human subjects is still required to translate this technology to the bedside.

One potential clinical application of cryoMNs is the intradermal delivery of DC vaccines for cancer immunotherapy43,44. In the hospital setting, the protocol includes repeated cycles of blood collection from patients and the generation of DCs45,46. The whole process is time consuming and cost prohibitive, and there are batch-to-batch variations in DC quality. By contrast, DC vaccinations with cryoMNs only require one-time blood collection and generation of DCs, which are then packaged in cryoMNs as ready-to-use aliquots. This process would dramatically improve the practicability of DC vaccination. The intradermal route of administration of DC vaccines has been extensively considered superior to other administration routes, with stronger tumour-specific immune responses and fewer side effects44,47,48. However, intradermal injection is traditionally performed manually by the Mantoux technique, which is technically challenging49. Given the innate features of microdimensions, harnessing cryoMNs greatly simplifies the operation of intradermal injection. Here, we could easily perform intradermal DC vaccination in mice. Administration of the DC vaccine with cryoMNs elicited stronger antigen-specific immune responses and a more potent tumorigenic ability in mice than s.c. and i.v. injection (Fig. 5). We also found that the administration frequency and dosage influenced the immune responses (Supplementary Figs. 22 and 23), consistent with previous reports50,51. Other parameters, such as DC subsets, DC maturation and activation, tumour-associated antigens and cancer type, also affect vaccination outcomes52, which is worthy of further optimization to maximize therapeutic efficacy.

Although we used DC vaccination as a proof of concept, cryoMNs are suitable for packaging other types of therapeutic cell. For example, melanocyte-loaded cryoMNs can be used for the treatment of vitiligo8,9. Stem cell-loaded cryoMNs can be used to promote skin regeneration53. In addition to mammalian cells, cryoMNs can also be used to deliver probiotic bacteria54 or bacteriophages55 for the treatment of topical infection. Finally, this cryoMN technology is not limited to delivering live therapeutics. Similar to other types of MN, cryoMNs can carry bioactive therapeutic agents, including proteins, peptides and vaccines. Due to the organic-free and low-temperature fabrication procedure, cryoMNs can maximally retain the bioactivity of those therapeutics.

Owing to the combination of MNs with cryopreservation technology, compared with the current standard of care, cryoMN technology not only achieves easy intradermal cell delivery with minimal invasiveness, but also simplifies the whole process of cell therapies by permitting multiple treatments with single donations. There are some translational considerations for this infant technology. First, the manufacture of cryoMNs is closely associated with the production of a clinically relevant number of therapeutic cells5. Reproducible manufacturing protocols of desirable cells for each therapy must be established. In addition, the materials for both the cryoMNs and cells have to meet the standards of the United States Food and Drug Administration (FDA) during the FDA approval process4. Second, because cryoMNs will be inserted into human skin, sterilization of cryoMN production is essential to prevent potential infections. CryoMNs can be manufactured under aseptic or sterile conditions. In addition, the feasibility of existing sterilization methods under low temperature is worthy of evaluation. Third, one cryoMN patch has limited cell capacity, which poses a critical issue if a large number of cells are required to induce potent therapeutic outcomes. It might be solved by the administration of multiple patches, but this might decrease patient compliance. An alternative solution is to increase the maximum cell capacity of a patch by either redesigning the cryoMNs with larger dimensions or enlarging the overall size of the patch with more MN arrays. Fourth, similar to COVID-19 mRNA vaccines, cryoMNs require ultracold shipping and storage conditions. Although traditional cryogenic equipment, such as −80 °C refrigerators and liquid nitrogen containers, are commonplace in most hospitals and research institutes, there is still an urgent need to develop new cryogenic equipment with low cost and high efficiency to facilitate distribution and decrease economic challenges, particularly in countries with limited resources. Furthermore, because cryoMNs will melt after being removed from their storage conditions, integrating cryoMNs with a handle or applicator will facilitate their easy manipulation and rapid penetration.


Fabrication of cell-loaded cryoMNs

The stainless-steel MNs were pre-designed and used as a master mould. To obtain the negative mould of the MNs, we replicated the master mould using 10:1 PDMS by pouring the liquid pre-polymer over the master MN structure, degassing in a vacuum oven for 10 min and curing the polymer at 70 °C for 1 h, and finally demoulding. The obtained PDMS mould was treated with O2 plasma and sterilized by ultraviolet exposure for 20 min before making cryoMNs. A 200-µl volume of optimized cryogenic medium (PBS supplemented with 2.5% (vol/vol) DMSO and 100 mM sucrose) was cast into the PDMS mould. Then PDMS mould was centrifuged at 4,000 r.p.m. for 3 min to allow the solution to fill up the needle cavities. The cells of interest were detached by trypsin and resuspended in cryogenic medium (1 × 106 cells ml−1). Next, 100 µl of cell suspension (1 × 105 cells) was added into the mould. The mould was placed aside for 20 min to allow the cells to fill up the needle cavities. The extra medium was removed from the base of the mould using a pipette and fresh cryogenic medium was added into the mould to fill up the base. The PDMS mould was held at −20 °C for 4 h. Then cryoMNs were gently peeled off the PDMS mould, following by freezing at −80 °C for 2 h and at −196 °C (in liquid nitrogen) for 1 h. The number of cells loaded in the MNs was calculated by subtracting the cell number in the removed medium from the cell number in the original cell suspension.

Morphology of cryoMNs

The cryoMNs were imaged by an optical microscope (Leica DVM6) and digital camera immediately after being removed from the liquid nitrogen. The melting behaviour of the cryoMNs was monitored by the digital camera. To visualize the cells (HaCaT) inside the cryoMNs, cell nuclei were stained with NucBlue Live ReadyProbes reagent (Invitrogen) before being loaded into the cryoMNs. The cryoMNs were placed into tissue culture plates and incubated in cold PBS for 30 min once being taken out of the liquid nitrogen. The samples were then visualized and imaged with a confocal microscope (LSM800, Carl Zeiss).

Evaluation of the mechanical strength of cryoMNs

The mechanical strength of the cryoMNs was examined using an Instron 5543 tensile meter. As shown in Supplementary Video 1, the cryoMNs were placed on a flat metal plate. The liquid nitrogen was used to provide a cooling environment for testing and to pre-cool both the probe and the metal plate. A vertical oriented force was applied perpendicularly to the MN tips, facing upward on a flat, rigid aluminium plate, using a 5-mm-diameter flat-head stainless-steel cylindrical probe (at a constant speed of 0.5 mm min−1). Four types of polymeric MN that are commonly used for transdermal drug delivery (PS-MNs, PLA-MNs, PCL-MNs and MeHA-CL-MNs15) were used for comparison.

Skin penetration ability of cryoMNs

CryoMNs loaded with RFP-HeLa cells were held at RT (24 °C) for a designated time (0 s, 10 s, 20 s, 30 s, 40 s, 50 s and 60 s) after being taken out of liquid nitrogen. The cryoMNs were inserted into porcine skin (fresh porcine cadaver skin was purchased from the local supermarket) by thumb press. The skin tissue was further fixed with 4% paraformaldehyde and stained with H&E for histological analysis. To visualize the cell distribution in the skin, the skin tissue was directly processed for cryo-sectioning and observed under a fluorescence microscope. All images were analysed by ImageJ software (version 1.53a, no plugin used).

Viability and proliferation of cells

Cell-loaded cryoMNs were placed into the 48-well plates without supplementation of culture medium and the live/dead cell viability assay was conducted according to the protocol provided by the manufacturers. All fluorescent images were analysed by ImageJ software (version 1.53a, no plugin used). Quantitative data for cell viability were calculated according to live (green)/dead (red) staining by dividing the number of living cells (green) by the total number of cells. In the proliferation assay, cell-loaded cryoMNs were transferred into 48-well plates with supplementation of culture medium. Meanwhile, the same types of cell were seeded into the plates with a density of 1 × 105 cells per well, for use as a positive control. The cells were cultured and the alamarBlue assay was conducted at days 1, 3 and 6 according to the protocol provided by the manufacturer. Briefly, alamarBlue reagent was added into each well to a final concentration of 10% (vol/vol) and the fluorescence intensity (excitation 540 nm, emission 580 nm) was measured after a 4-h incubation by using a plate reader (SpectraMax M5, Molecular Devices). Quantification of cell proliferation was obtained by dividing the fluorescence intensity at each designated culture time by the fluorescence intensity after one-day culture. In addition, cells were visualized by a phase contrast microscope at each designated time point.

Transdermal delivery of RFP-HeLa cells with cryoMNs on NOD/SCID mice

All mice used in the present study were housed at 20–24 °C and 30–70% humidity. The light/dark cycle of the holding room was 12 h/12 h from 8:00 to 20:00. All mice were fed with food and water ad libitum. Non-obese diabetic/severe combined immunodeficiency (NOD/SCID) mice experiments were performed in accordance with ethical approval by the Nanyang Technological University Institutional Animal Care and Use Committee (NTU-IACUC) under protocol A18001. The NOD/SCID mice (male, 6–8 weeks, 20–25 g) were purchased from InVivos. The mice were dehaired under anaesthesia before use. Blank cryoMNs and RFP-HeLa-loaded cryoMNs were applied on the left and right flanks of mouse backs for 5 min. To confirm cell survival after delivery into the mice skin, the MN-treated mice were imaged under an in vivo imaging system (IVIS, Spectrum, Perkin Elmer) using a filter set (excitation 570 nm, emission 680 nm), 2F/stop and 13.6-cm field of view for RFP imaging at the designated times (days 0, 1, 3, 7 and 14). The exposure time was 15 s. Quantitative analysis of the RFP fluorescence intensity was conducted by measuring the average radiant efficiency (photons s−1 cm−2 sr−1 µW−1) in regions of interest (ROIs) that were placed on the application site of cryoMNs. We subtracted the ROI value acquired from untreated skin on the same mouse. The data are normalized to the fluorescence intensity at day 0.

Vaccination with OVA-DC-cryoMNs

Experiments with C57BL/6 mice (female, 6–8 weeks, 20–25 g) were performed in accordance with ethical approval by the Animal Research Ethics Sub-Committee of City University of Hong Kong (internal ref. A-0493). The mice were randomly distributed into one of the following five treatment groups: untreated, cryoMNs, s.c. injection of OVA-DCs, i.v. injection of OVA-DCs (injection through the tail vein without saline flush), OVA-DC-cryoMNs. Each group of mice received vaccination on days 0, 3, 7, 10, 14, 17, 21 and 24. During each vaccination, four patches of OVA-DC-cryoMNs were applied to deliver 4 × 105 OVA-DCs per mouse. For s.c. and i.v. injection, 2.8 × 105 OVA-DCs was injected each time for each mouse (that is, 70% of cells applied for cryoMN vaccination). At day 28, lymph nodes and major organs (heart, lung, liver, spleen and kidney) were excised from the mice. Lymph nodes and partial spleen were homogenized by grinding with the end of a sterile syringe and processed into a single cell suspension. To study the homing of mature DCs to the lymph nodes, cells extracted from the lymph nodes were stained with FITC-conjugated CD11c antibody staining (1:200) to analyse DC infiltration, and APC-conjugated CD86 antibody (1:300) and APC-conjugated MHCII antibody (1:100) to evaluate DC activation and maturation, respectively. The stained cells were measured by a flow cytometer (BD Biosciences, BD FACSDiva software v 8.0) and were analysed by FlowJo software (TreeStar, version 10.5.3)

A splenocytes proliferation and CTL assay were then performed. Splenocytes (5 × 105 per well) were seeded in a 96-well plate and restimulated with 50 μg ml−1 OVA for two days. Splenocytes proliferation was evaluated using an alamarBlue cell viability assay according to the protocol provided by the manufacturer. The production of IFN-γ in culture supernatants was measured by a mouse IFN-γ enzyme-linked immunosorbent assay kit. The CTL assay was conducted following the manufacturer’s protocol (CyQUANT LDH cytotoxicity assay kit). Briefly, the B16 melanoma cell line transfected with ovalbumin (B16-OVA; target cells) was loaded with the fluorescence enhancing ligand. Splenocytes (effector cells) and B16-OVA (target cells) were then co-cultured in U-bottomed 96-well plates with cell number ratios of 10:1. 50:1 and 100:1. After incubation for 2 h at 37 °C, the lysed target cells were quantified. The major organs were cryosectioned and H&E stained to analyse for any organ damage after the vaccination process.

To evaluate the antitumour efficacy of the OVA-DC-cryoMNs vaccination, melanoma tumour models were developed on vaccinated C57BL/6 mice through s.c. injection of 2 × 105 B16-OVA cells in the right flank at day 28. The tumour volume was monitored every other day after cell inoculation by a digital caliper. The tumour volume (cubic millimetre) was calculated as 1/2 × long diameter × (short diameter)2. The tumour tissues and major organs were collected 26 days after cell inoculation following mice euthanasia, and the weight of the tumours was measured.

In vivo skin compatibility and safety evaluation

The cryoMNs patches and HA-MNs patches (as control) were applied to each flank of the dorsal skin of mice and removed after 10 min. To visualize skin resealing and irritation, the treated sites were imaged with a digital camera at pre-designated time points. For histological analysis, the treated sites were collected and processed for cryo-sectioning, then H&E and TUNEL staining were performed on the skin sections.

Statistical analysis

All experiments used biological replicates that consisted of cells in non-repeated, independent cell culture wells or tissue samples from different animals, unless specified otherwise. Quantitative data are represented as mean ± s.d. Statistical analysis was performed by using two-tailed Student’s t-test or original one-way analysis of variance. P < 0.05 was considered statistically significant (*P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001). GraphPad Prism 8.3 was used for data analysis. Microsoft Excel 2019 was used for calculating the exact P value when P < 0.0001.

Additional experiments

Materials, cell culture, optimization of cryogenic medium, generation and stimulation of bone marrow-derived dendritic cells and optimization of vaccination with OVA-DC-cryoMNs are available in the Supplementary Information

Reporting Summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.