Development of an Intrinsic Skin Sensor for Blood Glucose Level with CRISPR-mediated Genome Editing in Epidermal Stem Cells

Biointegrated sensors can address various challenges in medicine by transmitting a wide variety of biological signals. A tempting possibility that has not been explored before is whether we can take advantage of genome editing technology to transform a small portion of endogenous tissue into an intrinsic and long-lasting sensor of physiological signals. The human skin and the epidermal stem cells have several unique advantages, making them particularly suitable for genetic engineering and applications in vivo. In this report, we took advantage of a novel platform for manipulation and transplantation of epidermal stem cells, and presented the key evidence that genome-edited skin stem cells can be exploited for continuous monitoring of blood glucose level in vivo. Additionally, by advanced design of genome editing, we developed an autologous skin graft that can sense glucose level and deliver therapeutic proteins for diabetes treatment. Our results revealed the clinical potential for skin somatic gene therapy.


Introduction
Biointegrated sensors can address various challenges in medicine by transmitting a wide variety of biological signals, including electrophysiological and biochemical signals continuously 1 .
Recent advancement in device designs and fabrication using novel nano-materials have greatly accelerated the development of new soft and stretchable electronics applicable in vivo, including brain, heart, and skin 1,2 . However, it remains technically challenging in the field to improve the stability and biocompatibility of the device and resolve potential infection caused by device implantation. The recent development of genome editing technology [3][4][5] allows us to efficiently engineer cells and potentially transform a small portion of endogenous tissue into an intrinsic and long-lasting sensor of physiological and biochemical signals. However, this intriguing possibility has not been tested before.
The human skin and the skin epidermal stem cells 6,7 have several unique advantages, making them particularly suitable for genetic engineering and applications in vivo ( Supplementary Fig. 1A).
First, skin is the most accessible organ in our bodies, and its surface localization is ideal for signal transmission or monitoring. Secondly, the skin is one of the largest organs on the body surface, making it easy to isolate skin epidermal stem cells and to monitor the tissue for potential detrimental complications after engineering and, if necessary, to remove it in the case of an adverse consequence. The procedure to isolate and culture primary epidermal stem cells has been well established [8][9][10] . Cultured epidermal stem cells can be induced to stratify and differentiate in vitro, and transplantation of epidermal autografts is minimally invasive, safe, and stable in vivo. Autologous skin grafts have been clinically used for the treatment of burn wounds for decades 11,12 . However, despite of the potential clinical applications, the research of skin epidermal stem cells has been greatly hampered by the lack of an appropriate mouse model. Although it has been shown that mouse skin or human skin can be transplanted onto immunodeficient mice [13][14][15][16][17][18][19] , the lack of an intact immune system in the host animals makes it impossible to determine the potential outcomes that the procedure may elicit in vivo. Immune clearance of engineered cells has been one of the major complications for somatic gene therapy 20 . Additionally, it remains technically challenging to perform skin organoid culture with mouse epidermal stem cells and to generate mouse skin substitute for transplantation. To resolve these issues, we have developed a new organotypic culture model with mouse epidermal stem cells in vitro by culturing the cells on top of an acellularized mouse dermis 21-23 . Exposing the cells to the air/liquid interface can induce stratification of the cultured cells to generate a skin-like organoid in vitro 24 . Transplantation of the skin organoids to wild type (WT) mice with intact immune system has led to long-term and stable incorporation of engineered skin in vivo 22 . In this report, with this unique isogenic skin transplantation platform, our results demonstrated that engineered skin grafts derived from genome-edited epidermal stem cells can serve as an internal biosensor for key biochemical signals, such as blood glucose, allowing long-term and non-invasive monitoring of glucose level in vivo.

Glucose sensing with cultured mouse epidermal stem cells in vitro
Obesity and diabetes are acute and growing public health problem around the world 25 . A biointegrated sensor for noninvasive monitoring of blood glucose level in vivo will remove the need for the patients to draw blood multiple times a day [26][27][28] . Additionally, continuous monitoring of glucose allows the patients to better maintain the blood glucose level by altering the insulin dosage or diet according to the prevailing glucose values and prevent potential hyperglycemia and hypoglycemia. If the sensor can be connected with an insulin delivery device, it may create an "artificial endocrine pancreas" that can automatically maintain the glucose level in patients. Currently, most of the continuous monitoring sensors for glucose are enzyme electrodes or microdialysis probes implanted under skin [26][27][28] . These sensors usually require oxygen for activity, and are insufficiently stable in vivo and poorly accurate under the low glucose condition. The presence of interfering electroactive substances in tissues can also cause impaired responses and signal drift in vivo, which necessitate frequent calibrations of current sensors. A fluorescence-based glucose sensor in the skin will likely be more stable, have improved sensitivity, and resolve the issue of electrochemical interference from the tissue 29 .
To engineer epidermal stem cells for glucose sensing, we tested the intracellular expression of a sugar binding protein, glucose/galactose-binding protein (GGBP) 30 . GGBP transports glucose within the periplasm of E.coli, and the binding of glucose to GGBP can lead to a large conformational change in the protein 30 . This property of GGBP has been exploited to develop protein sensors for glucose using FRET (fluorescence resonance energy transfer) or bioluminescence imaging [31][32][33][34][35] . WT (wild type) GGBP has a very high glucose binding affinity (K d = 0.2 μm). To generate a probe corresponding to the physiologically relevant range of glucose, we engineered a CFP/YFP FRET sensor with A213R/L238S double mutant of GGBP (K d = 10 mM) 36 . Although skin stem cells are very susceptible for manipulation with viral vectors, viral infection could lead to genotoxicity and may raise a significant safety concern for the potential applications in vivo 37,38 . The recent advancement of genome editing technology with CRISPR system presents a novel approach to carry out a sitespecific modification of the target genome non-virally 5,39,40 . To test CRISPR-mediated genome editing in mouse epidermal stem cells, we developed DNA vectors encoding the D10A mutant of Cas9 (CRISPR associated protein 9) 41 , two gRNAs (guide RNA) targeting the mouse Rosa26 locus, and a Rosa26-targeting vector. The targeting vector contains two homology arms for the Rosa26 locus, flanking an expression cassette that encodes a GGBP fusion protein ( Fig. 1A and Supplementary Fig.   1B).
Primary epidermal keratinocytes were isolated from CD1 newborn mice, and were electroporated with the Rosa26 targeting vector together with the plasmids encoding Cas9 and Rosa26-specific gRNAs. Clones were isolated upon selection and the correct integration into the Rosa26 locus was confirmed by both PCR screening and southern blot analysis (Fig. 1B). Engineered epidermal cells exhibited robust expression of GGBP fusion proteins in the cytosol (Fig. 1C). To test the glucose sensing in vitro, we exposed the cells to medium containing an increasing amount of glucose. Quantification of FRET ratio by microscopic imaging showed an excellent correlation of FRET ratio with extracellular glucose concentration, ranging from 0-10 mM (Fig. 1D). Our analysis further showed that the GGBP reporter responded to extracellular glucose or galactose, but not to other sugars, such as sucrose, fructose, or ribose (Fig. 1E). The intracellular GGBP probe responded to the change of glucose concentration rapidly. The FRET ratio changed within 30 seconds after replacement of the medium, and remained stable in the same medium. To assess the intracellular GGBP probe more quantitatively, the reporter-expressing cells were perfused with an increasing concentration of glucose. When a stable FRET ratio was reached for each concentration, glucose was removed by perfusion and exchanged with a glucose-free medium. Perfusion of glucose from 0.1 mM to 10 mM led to reversible and increasing FRET efficiency changes ( Supplementary Fig. 1C). When exposed to glucose at higher concentration, the FRET efficiency changes followed a hyperbolic curve, reflecting saturation of the probe at concentration higher than 25 mM ( Supplementary Fig. 1D).
Together, our results suggest that epidermal stem cells expressing the GGBP reporter can faithfully sense the extracellular glucose concentration. When subjected to skin organoid culture 22 , the targeted cells readily produced stratified epithelial tissue ( Supplementary Fig. 1F).

Glucose sensing with skin transplants in vivo
To investigate the potential applicability in vivo, we prepared the skin organoid culture with the engineered epidermal stem cells 22 , and transplanted the organoids to CD1 host animals ( Fig. 2A). No significant rejection of the skin grafts has been observed after transplantation, suggesting that the targeted epidermal stem cells are well tolerated immunologically in vivo. The grafted skin exhibited normal epidermal stratification, proliferation and cell death (Supplementary Fig. 2A-E). To test the glucose sensing capability in vivo, we carried out the IPGTT (intraperitoneal glucose tolerance test) in grafted animals. Fasted animals were administered with a bolus of glucose intraperitoneally.
Fluorescence (FRET) change in the grafted skin was monitored by intravital imaging (Fig. 2B), and 6 the blood glucose level was measured by a commercial glucose monitoring system (Bayer Contour) with the blood samples taken from the snipped tail. Figure 2C and D showed the correlation between the measured glucose concentration and the FRET ratio changes over time. Interestingly, the FRET ratio exhibited a nearly linear correlation with the glucose concentration in vivo (Fig. 2D, R 2 =0.977).
Traditional glucose sensor cannot accurately measure low glucose level in vivo 26,27 . To test the skin sensor under lower glucose conditions, we induced hypoglycemia by insulin administration to fasted animals. Intravital imaging of the grafted skin again showed excellent correlation of the FRET ratio changes with glucose levels ( Fig. 2E and F). To further evaluate the utility of our sensor in a physiological setting, fasted mice were fed with standard chow diet for 30 minutes, and blood glucose level was monitored afterwards. Our results demonstrated the correlation between the blood glucose levels with FRET efficiency changes (Supplementary Fig. 2F-G). The reporter also showed excellent dynamic range when hyperglycemia was induced in vivo by intraperitoneal injection of higher dose of glucose ( Supplementary Fig. 2H). Together, our results strongly suggest that the engineered skin graft with the GGBP reporter can accurately sense the blood glucose level in vivo.

Glucose sensing and delivery of therapeutic proteins with a single skin transplant
The unique advantage of our epidermal stem cell platform allows complex genetic engineering to be applied with ease. Thus, introduction of an expression cassette that encodes both GGBP reporter and a therapeutic protein into the epidermal stem cells may achieve continuous glucose monitoring and diabetes treatment with one single skin transplantation. The hormone GLP1 (glucagon like peptide 1) is released from the gut upon food intake. GLP1 acts both as a satiety signal to reduce food consumption and as an incretin hormone to stimulate insulin release and inhibit glucagon secretion 43 . These benefits have made GLP1 receptor agonists being used to treat type 2 diabetes.
Our recent study demonstrated that skin-derived GLP1 release can effectively reduce body weight and inhibit diabetes development in a high fat diet model 22 . To develop skin transplant with multiple functionalities, we engineered a new Rosa26 targeting vector containing an expression cassette that encodes a GLP1 and mouse IgG-Fc fragment fusion protein together with the GGBP reporter 7 ( Supplementary Fig. 3A). Fusion with IgG-Fc can enhance the stability and secretion of GLP1 when ectopically expressed in cells 44 . The engineered epidermal cells exhibited a robust GLP1 production and secretion in vitro and in vivo ( Fig. 3A and Supplementary Fig. 3B). Since the ectopic expression of GLP1 is driven by a constitutive promoter, the expression and secretion of GLP1-Fc fusion protein by the skin epidermal cells led to a constitutively elevated level of GLP1 in blood regardless of nutrient or food intake ( Supplementary Fig. 3C). The secreted GLP1 fusion protein is functional as the conditioned medium from GLP1-expressing cells can significantly induce secretion of insulin when added to insulinoma cells cultured in vitro (Fig. 3B).
To examine the potential applicability in diet-induced obesity and diabetes, we transplanted GLP1/GGBP-expressing cells and control cells (GGBP alone) to two cohorts of CD1 adult mice, and employed high fat diet (HFD) to induce obesity in grafted animals. To minimize gender difference, only male animals were used in our study. Compared with the animals on regular chow diet, the HFD greatly accelerated body weight gain in the mice grafted with control cells. By contrast, GLP1 expression led to a significant inhibition in body weight increase ( Fig. 3C and quantification in 3D). To examine glucose homeostasis, we conducted IPGTT. Expression of GLP1 significantly reduced glycemic excursion in vivo as determined by both direct measurement of blood glucose or intravital imaging of the GGBP reporter ( Fig. 3E-G). Noninvasive monitoring of the GGBP reporter exhibited an excellent correlation with the conventional glucose measurement in both diabetic animals and the GLP1-treated animals ( Fig. 3F and G).

Glucose sensing with human epidermal stem cells
To test the feasibility of glucose sensing with human epidermal stem cells, we cultured human skin organoids from primary epidermal keratinocytes isolated from human newborn foreskin. The human epidermal keratinocytes can readily produce organoids in vitro, which can be transplanted to the nude mice. When infected with lentivirus, the grafted human cells exhibited robust expression of the exogenous Luciferase gene (Fig. 4A). The grafted tissue showed normal skin stratification when stained for early or late epidermal differentiation markers (Fig. 4B). To examine CRISPR-mediated genome editing in human epidermal cells, we developed vectors encoding two gRNAs targeting human AAVS1 (adeno-associated virus integration site 1) locus, and an AAVS1-targeting vector ( Supplementary Fig. 4A) that encodes GGBP reporter protein. Human epidermal keratinocytes were electroporated with the targeting vector together with the plasmids encoding Cas9 and the gRNAs.
Clones were isolated and the correct integration was confirmed by southern blot analysis (Fig. 4C).
Expression of the GGBP fusion protein in human cells did not significantly affect cell proliferation ( Supplementary Fig. 4B) or differentiation ( Supplementary Fig. 4C) in vitro. The engineered cells stratified and formed skin organoids in vitro, which were transplanted to the nude host. The grafted skin exhibited normal epidermal stratification and proliferation in vivo (Supplementary Fig. 4D-G). With IPGTT in the grafted nude mice, we carried out intravital imaging of the GGBP reporter (Fig. 4D) and observed similar correlation of FRET ratio changes in the grafted skin with blood glucose levels ( Fig.   4E and F). Our data strongly suggest that the skin-based glucose sensor can be used to monitor blood glucose level in human patients.

Discussion
Noninvasive monitoring of blood glucose level in a continuous manner can provide significant clinical benefits to the diabetic patients [26][27][28][29] . The current methodologies of continuous glucose detection mostly depend on electrochemical approaches (enzymatic or non-enzymatic approaches) and optical approach (fluorescence, FRET, ocular spectroscopy, infrared spectroscopy, photoacoustic spectroscopy, et al.) 26,28,[45][46][47] . In order to sense the glucose level in vivo, most current technologies have to use invasive sensors that are implanted subcutaneously or intravenously. However, implantation of the devices can lead to significant tissue inflammation and biofouling-induced sensor degradation, causing loss of sensitivity. Although less invasive approaches have been developed, such as iontophoresis, sonophoresis, smart tattoo, and microneedles, these technologies suffer from various drawbacks, including calibration inaccuracy, biological incompatibility, current-induced skin erythema or irritation, significant subject-to-subject variability, and long warm-up or duration time required to collect sufficient amount of sample 26,28,29,[45][46][47][48] . Thus, the development of novel, noninvasive, reliable, and long-lasting sensor for continuous glucose monitoring will make a tremendous contribution to the field.
Because epidermal stem cells allow efficient genetic manipulation with the minimal risk of tumorigenesis or other detrimental complications in vivo 13,14 , the skin may serve as an ideal intrinsic sensor of our own biochemical and physio-pathological signals. We have recently resolved the technical hurdles and established the unique mouse-to-mouse skin transplantation model with immunocompetent hosts 22 . Our results provide the key evidence supporting the feasibility of cutaneous monitoring of blood glucose level in vivo with mouse models. The skin-based glucose sensor provides an accurate and stable glucose monitoring with a wide dynamic range. Because of low immunogenicity of autologous skin transplants, our probe will have superior durability for long term application in vivo.
Given the enhanced sensitivity and non-invasive nature, fluorescence-based glucose sensing has been actively explored for continuous glucose monitoring 29,47 . Measurement of the fluorescence signals causes little or no damage to the host. With future improvement of the probe, such as increased sensor expression in the epidermal cells and using FRET pairs with longer excitation wavelength, we can further reduce the exposure time of skin to fluorescence light source for repeated measurements, thus minimizing the potential photo-damages of the skin. The same platform can be also exploited to develop other biosensors and stably deliver therapeutic proteins in vivo, such as GLP1, providing a promising treatment for many human diseases 13,14 .
Taken together, our study demonstrates that grafted skin stem cells expressing a foreign reporter protein can be efficiently and stably grafted to the host mice with intact immune systems, and can be used to accurately monitor metabolic changes in vivo. Skin-based gene therapy has been proposed to treat many diseases, including inherited skin disorders 19,[49][50][51] and other terminal or severely disabling diseases [13][14][15][16][17][18] . Our study unravels the tempting potential of cutaneous gene therapy for various clinical applications in the future.

Contact for Reagent and Resource Sharing
Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Xiaoyang Wu (xiaoyangwu@uchicago.edu). Primary human epidermal basal keratinocytes were purchased from Invitrogen. Cells were cultured on mitomycin C-treated 3T3 fibroblast feeder cells with E-media supplemented with 15% serum and a final concentration of 0.05 mM Ca 2+ 8,9 .

Method Details
Plasmid DNA Constructions Lentiviral vector encoding Luciferase and H2B-RFP has been described before 21, 23  Electroporation was carried out with BioRad gene pulser using Ingenio electroporation solution. Cells were electroporated using exponential decay mode at 250 Volts and 950 µF. Electroporated cells were immediately suspended in culture medium and plated. Cells were selected with puromycin (2.5µg/ml) 2-3 days post electroporation.
Mouse and human cells were electroporated at passage 8-10. Mouse epidermal cells at this stage can grow in feeder-free condition, whereas human cells are continuously cultured with feeder cells. Limited trypsinization or brief EDTA (20 mM in PBS) treatment was used to remove the feeder cells before suspension of human epidermal cells.

Skin organoid culture and transplantation
Decelluralized dermis (circular shape with 1cm diameter) was prepared by EDTA treatment of newborn mouse skin 24 . 1.5 X 10 6 cultured keratinocytes were seeded onto the dermis in cell culture insert. After overnight attachment, the skin culture was exposed to air/liquid interface.
For grafting with skin organoids, CD1 males with the ages of 6-8 weeks were anesthetized. A silicone chamber bottom with the interior diameter of 0.8cm and the exterior diameter of 1.5cm was implanted on its shaved dorsal mid-line skin, which was used to hold the skin graft. A chamber cap was installed to seal the chamber right after a piece of graft was implanted. About one week later, the chamber cap was removed to expose the graft to air. A single dose of 0.2mg α-CD4 (GK1.5) and 0.2mg α-CD8 (2.43.1) antibodies was administered intraperitoneally for skin grafting.

Cell cycle analysis
Propidium Iodide (PI) staining followed by Flow Cytometry Assay were used to determine the effect of cell cycle profiles. Mouse and human epidermal cells were cultured in two 6cm cell culture dish for 24 hours, respectively. Cells were trypsinized and 1×10 5 cells from each dish were collected, followed by one PBS wash. Fixation of cells was carried out using 70% (v/v) ice cold ethanol for 1 hour. Then, the fixed cells were centrifuged at 500g at 4 o C for 10 minutes, followed by PBS wash for two times. The cells were then treated with 75µg RNAse A in 100µl PBS and incubated at 37 o C for 1 hour. After incubation, the cells were collected by centrifuging at 500g at 4 o C for 10 minutes, followed