Microneedles for transdermal diagnostics: Recent advances and new horizons

Abstract

Point-of-care testing (POCT), defined as the test performed at or near a patient, has been evolving into a complement to conventional laboratory diagnosis by continually providing portable, cost-effective, and easy-to-use measurement tools. Among them, microneedle-based POCT devices have gained increasing attention from researchers due to the glorious potential for detecting various analytes in a minimally invasive manner. More recently, a novel synergism between microneedle and wearable technologies is expanding their detection capabilities. Herein, we provide an overview on the progress in microneedle-based transdermal biosensors. It covers all the main aspects of the field, including design philosophy, material selection, and working mechanisms as well as the utility of the devices. We also discuss lessons from the past, challenges of the present, and visions for the future on translation of these state-of-the-art technologies from the bench to the bedside.

1. Introduction

Modern medicine has witnessed a continuous growth of needle-based blood collection for laboratory analyses due to high efficiency and low cost. But underneath the flourish are intractable problems: (1) reuse of un- or inappropriate-sterilized needles is common in developing countries to create a serious risk of transmitting blood-borne pathogens (e.g., HIV, HBV, and HCV) [1]; (2) 3.5%−10% of the world’s population expecially children holds a somewhat exaggerated form of needle phobia that may cause them to avoid seeking routine and emergency medical care [2]; (3) hypodermic needles and syringes are difficult for at-home self-administration by untrained personnel with regard to safety and waste management. Even for an easy-to-use blood glucose meter, frequent finger pricks would result in puncture-related pain and discomfort. Prausnitz et al. made remarkable progress in developing an alternative tool by shortening the needle length to around 150 μm using microfabrication technology [3]. The revolutionary design of needles, named microneedles (MNs) thereafter, guarantees not only effective drug transport across the stratum corneum (SC), but also minimal pain by without hitting nerve endings in the dermis. Since then, the research on the development of MNs has moved forward at a rapid pace [4–9].

The primary function of MNs is to gain access to biofluids beneath the skin in a nearly pain-free manner. Human skin is made up of three main layers: (1) the outermost SC layer having a thickness in the range of 10–200 μm [10]; (2) viable epidermis with interstitial fluid (ISF) beneath the SC; (3) dermis composed of blood, lymph vessels, nerve endings, and connective tissue at 300–1500 μm. A typical MN device is installed with a single needle or a needle array with a needle length of 50–2000 μm, a tip diameter of 1–100 μm, and a base width of 25–500 μm [4,6]. The MNs can be fabricated using various materials (e.g., silicon, glass, metals, and polymers) in different structures (solid, hollow, porous, coated, etc.) and shapes (conical, pyramidal, cylindrical, spiky, snake-fang like, etc.) [6]. So far, the MNs have been mainly applied to cosmetics, therapeutics, and diagnostics, among which MNs-based cosmetic treatments travel the furthest distance in practicality to be commercialized for over 10 years [8]. At present, MN-based therapeutics are in clinical trials and most research are focused on the transdermal delivery of drugs for influenza vaccination [11–15] or for treatment of diabetes mellitus [16–18], cancers [19,20], neuropathic pain [21], hair regrowth [22], and obesity [23,24].

It is not until recent years that the MNs have been used as a building block to engineer diagnostic sensors. In general, biosensors for point-of-care testing (POCT) detect medically relevant signals by analyzing external secretions (urine, saliva, sweat, and tear) or body fluids (ISF and blood) [25]. The use of external secretions are intrinsically constrained due to fewer biomarkers at lower concentrations and inaccuracy in disease screening/diagnosis [26,27]. The MNs are deemed ideal biosensing platforms since they enable extraction of or access to cutaneous body fluids in a minimally invasive way for detection of various analytes such as macromolecules, metabolites, and drugs. This new type of device greatly increases the use of ISF as a source of biomarkers for POCT. Recent studies confirmed that cutaneous ISF has similar profiles of proteins, small molecules, and RNA to blood [28–30]. Additionally, ISF contains some specific biomarkers (e.g., exosomes and resident memory T cells) that are exclusive or sufficient compared to blood [30,31], which is especially true for cutaneous disorders (e.g., melanoma, lupus, and psoriasis). Currently, there are two major operating modes of MN-based diagnosis: (1) extract ISF or blood that carries bioanalytes through MNs for follow-up testing; (2) adapt MNs into minuscule capturers or sensors for testing near or at the site of care. The second mode is more advantageous in terms of simplicity and speed, but of narrower applicability at the moment. No matter in which mode, the MNs are rebuilding our expectations of future medical diagnosis. Therefore, it is worthwhile to comprehensively overview the recent advances in MN-based biosensing research. Herein, the objectives of this review are to: (1) summarize the principles of MN-based detection; (2) compare the methods of MN-mediated blood/ISF extraction; (3) expound the utility of MN-based transdermal sensors for bioanalyte detection; (4) discuss the major challenges and future trends in biomedical applications of MN-based sensors.

2. Diagnosis strategies and principles

We categorize MN-based diagnosis into three modalities based on where diagnosis can be made: (1) “Off device”, a MN-device is only for transdermal sampling of biofluids and the samples are transferred for central lab testing; (2) “On device”, a miniaturized analyzer is integrated with a MN-device, so that sample transfer is unnecessary; (3) “On MN”, every single MN mounted in a MN-device functions for in vivo biomarker collector or analyzer. Both “on-device” and “on-MN” modes can be adapted for use in rapid diagnostic tests at the point-of-care.

2.1. Microneedles for transdermal sampling

Biofluids (i.e., ISF and blood) can be drawn out from the skin by MNs based on negative pressure, capillary force, or material absorption. Fig. 1 demonstrates the schematics of hollow (H), porous (P), and solid (S) MNs for skin biofluids sampling [32–34]. Long HMNs of > 1500 μm in length are suitable for blood sampling (see Table 1). They can draw blood from the skin using the capillary force or negative pressure. However, the capillary action is not competent for the situation where a large volume of blood is required. Therefore, a typical blood sampling device comprises a long HMN and an actuator that provides negative pressure to expedite sampling [35]. For short HMNs and PMNs (typical height < 900 μm), the capillarity of the continuous microchannels within the MN patches is preferable to extract ISF in a self-powered manner. In general, the sampling rate based on capillarity can be expedited by reducing the contact angle (θ, see Fig. 1) [36], narrowing the microchannel diameter (2r_e) [32], and/or inserting a wick into the HMN [37]. For SMNs, the ISF sampling can be accomplished by fluid diffusion and material absorption. The SMNs are generally made of hydrogel that is a polymeric crosslinking network containing innumerable hydrophilic groups (e.g., −NH₂, −OH, and −SO₃H) [38]. ISF can continuously diffuse into the hydrogel SMNs, until the swelling force and the elastic network retraction force reach equilibrium. The absorbed ISF can be unloaded from the MNs by centrifugation and/or solvent extraction, while the crosslinking network of the hydrogel remains intact during manipulations [34,39].

Fig. 1.

Differently shaped and structural MNs for transdermal biofluid sampling driven by negative pressure, capillarity, or swelling.

Table 1.

Summary of biofluid extraction using different types of MNs.

MN type	Materials	Length (μm)	Sample	Sampling volume/rate	Sampling time	Sampling method	Test subject	Ref
HMN	Si	100	ISF	1 μL s−1	-	Capillary action	-	[32]
	Metal	1800	Blood	31.3 μL	4 s	Vacuum	Rabbit	[73]
		1500	Blood	31.5 μL	-	PDMS actuator	Rabbit	[74]
	Glass	1500	ISF	20 μL	1–2 h	Mechanical pressure	Human	[30]
PMN	Al2O3	900	Water	-	15 min	Capillary action	-	[75]
	PGMA	700	Water	-	A few seconds	Capillary action	-	[76]
		350	ISF	-	A few seconds	Capillary action	Human	[41]
	PDMS	600	PBS	1.71 nL min−1	-	Compression force	-	[33]
SMN	Metal + paper	750	ISF	2 μL	1 min	Paper absorption	Wistar rat	[77]
	Metal	1000	Blood	103.5 μL	3 min	Vacuum	Human	[78]
Hydrogel MN	PMEV/MA-PEG	600	ISF	0.84 mg	1 h	Swelling	Rat	[79]
		600	ISF	-	5 min	Swelling	Human	[39]
		600	ISF	-	1 h	Swelling	Porcine	[80]
	MeHA	800	ISF	1.4 μL	1 min	Swelling	Mouse	[34]
	Alginate + PLLA	550	ISF	6.5 μL	2 min	Swelling	-	[56]

2.2. Microneedles for transdermal detection

The “off device” diagnosis often suffers from insufficient sampling and follow-up time-consuming procedures. In contrast, both “on device” and “on MN” modes avoid the need for sample transfer, which significantly shortens the turnaround time. These two modes have four common types of designs (Fig. 2a–d): (1) attachment of a sensor to the base of an HMN patch (on device) [40–42]; (2) installment of a sensor into the lumens of HMNs (on device) [43–48]; (3) modification of SMN surface to make itself as a sensor (on MN) [49–51]; (4) metallization of SMNs as dry electrodes (on MN) [52–54].

Fig. 2.

Four typical designs of MN-based sensors. (a) HMNs topped with a specific sensor, (b) electrochemical electrode incorporated HMNs, (c) surface-functionalized and (d) metalized SMNs. (e-i) Working mechanisms of different MN sensors: colorimetry, immunoassay, nucleic acid recognition, and electrochemistry.

The working principles for the four designs of MN-biosensors include colorimetry, sandwich immunoassay, enzyme-labeled electrochemical immunoassay, nucleic acid (NA) recognition, enzymatic/nonenzymic electrochemistry, and skin dry electrodes.

The sensors generally consist of a biorecognition element and an optical or electrochemical transducer. The enzyme and antibody/antigen are mostly used in the MN-based sensors for analytes recognition and capture. In a colorimetric MN sensor, the enzyme (e.g., glucose oxidase) is typically used for both biorecognition and catalysis to produce hydrogen peroxide (H₂O₂) in the presence of analyte. The analyte concentration is readily estimated from the H₂O₂-induced color change of a chromogenic substrate on a paper strip (Fig. 2e) [55]. With respect to the MN immunosensors, the antibody or antigen is generally linked on the MN surface for in vivo protein capturing. Signal transduction relies on in vitro addition of secondary antibody labeled with either a fluorophore for fluorometric assay, or an enzyme for colorimetric assay, such as enzyme-linked immune sorbent assay (ELISA). When transferring ELISA to an electrode to form an electrochemical immunosensor, the analyte concentration can be quantified by monitoring the redox current between the labeled enzyme and a substrate using the electrode (Fig. 2f). Recently, peptide nucleic acids (PNAs) are immobilized on the hydrogel MNs to specifically bind complementary target DNAs via Watson-Crick base pairing (Fig. 2g) [56]. The concentration of the PNA/DNA duplex is determined using DNA intercalator either on MN or off MN after a light-triggered release process. Of note, the MNs combined with the colorimetry, ELISA, NA recognition, or electrochemical immunoassay are for a single use.

For continuous detection, MNs need to be combined with or modified into electrochemical electrodes which mainly consist of sensing materials and conductive electrodes. Most electrochemical MN sensors are enzyme-based, whereas the other sensors are enzyme-free. The majority of the enzyme-based MN sensors monitor the formation of H₂O₂ during the enzyme-catalyzed reaction of analyte, which causes a variation in the current proportional to the analyte concentration (Fig. 2h left, so called the first generation of enzymatic biosensor) [25,57]. The H₂O₂-based sensors are intrinsically influenced by the ambient concentration of dissolved oxygen in skin and require high operating voltage ranging from 0.4 to 0.7 V [44,58–60]. To avoid these issues, MNs have been combined with the second-generation sensing technique [61–63], which utilizes a redox mediator instead of H₂O₂ to shuttle electrons from the redox center of enzyme to the electrode (Fig. 2h right). For example, a low working voltage of 0.15 V was achieved for lactate detection using methylene blue as the mediator [62]. While the redox mediator-assisted MN sensors obviate the limitation of the first-generation sensing technology, they still suffer from susceptibility of enzyme activity to environmental conditions (e.g., pH and temperature). In this context, non-enzymatic catalytic materials have been introduced to MN electrodes mainly for glucose monitoring [64–66]. The MN sensors utilize the non-enzymatic materials as both catalysts and mediators to interact with the analytes and to shuttle electrons between the catalysts and the electrodes (Fig. 2i). Similar detecting principle applies to the pH and nitric oxide (NO) MN sensors [50,67]. For instance, pH responsive metal oxide can directly transfer electrons from its redox reaction with H₃O⁺ to electrode, which induces a potential change for pH detection [68].

Metalized MNs mounting on skin can be used as bioelectrodes for recording of biopotentials, including electrocardiography (ECG) [69], electroencephalography (EEG) [53], and electromyography (EMG) [70]. The biopotentials are monitored by at least two wet electrodes in contact with the skin; however, the traditional wet electrodes containing electrolytes give rise to a high skin impedance and unstable biopotentials during long-term diagnosis. The impedance of wet electrode-skin scheme can be modeled as a combination of an electrolyte resistance (R_e) and three impedances of the electrode to electrolyte (Z_e), the electrolyte to SC (Z_SC), and tissue underneath [71]. The dry MN electrode can impale the SC to directly access ISF, which circumvents the need for electrolyte and the high impedance issue arising from the SC, thus effectively eliminating R_e, Z_e, and Z_SC [72]. Compared to the wet electrode, the MN electrode therefore has low impedance to show superior sensitivity and long-term stability [52,53,69].

3. Biofluid extraction for off-device diagnosis

3.1. Blood extraction

Blood has long been the most important window through which clinicians can identify what is happening to a patient’s health. The dermal vasculature is located at 300–1500 μm beneath the skin surface, and therefore the HMN patch with a micro-cannula of > 1500 μm in length is frequently applied in phlebotomy [9]. Such long HMNs have been fabricated using silicon (Si) [81], polymers [82,83], and metals [84,85]. Among the structural materials, metals are mostly used due to their merits of high Young’s moduli and fracture toughness.

A typical HMN-device involves a metal micro-cannula and a vacuum-generating actuator. Actuators with different operating technologies have been developed for the HMNs, including piezoelectric [86], electrolyte-controlled [36], gel-based [87], thermopneumatic [88], and elastic actuators [89]. The elastic chamber made of polydimethylsiloxane (PDMS) is a simple and desirable actuator as negative pressure can be readily induced by elastic deformation. Jung et al. built up a PDMS chamber with an inlet and an outlet valve [35]. The two valves allowed the formation of negative pressure, rapid blood extraction, and blood transportation. The team further optimized their design with a PDMS chamber and a polymer-sealed HMN (Fig. 3a) [73]. The HMN tip was capped with polyvinyl pyrrolidone (PVP) to form a closed PDMS chamber. With high air permeability of PDMS, negative pressure was created in the closed chamber after incubation in a vacuum box, followed by depositing an airproofing parylene film to maintain the vacuum. Once the vacuumized device was pressed on skin, the PVP sealer got detached and blood was absorbed into the chamber. Animal testing confirmed a high rate of blood withdrawal (~7.8 μL s⁻¹). Inspired by the Jung et al.’s scheme, other groups reported similar HMN models without use of sealers or valves [74]. More recently, Blicharz et al. manufactured a blood collection device by enclosing a SMN array, a microchannel system, a vacuum chamber, and a reservoir in a compact chassis (Fig. 3b) [78]. The stainless steel SMNs were punched into and withdrawn from skin by a bi-stable disc spring. As a result of the vacuum actuation, ~100 μL of blood flowed into the reservoir via the microfluidic channel in 3 min. For the collection process, all a user has to do is to push the actuation button, so that special training is not mandatory.

Fig. 3.

Schematics of (a) a capped HMN with a pre-vacuumized PDMS actuator [73] and (b) a SMN array-based device for blood sampling [78]. Figures reprinted with permissions from Royal Society of Chemistry and Springer Nature.

3.2. ISF extraction

ISF is formed by blood transcapillary filtration and cleared by lymphatic vessels [90]. It is similar to blood in terms of protein diversity (93.3% in common) [28], small-molecule metabolite composition (79.3% in common) [29], and RNA profile (92.5% in common) [30]. Therefore, ISF is ideally alternative to blood for clinical diagnosis, along with advantages of good applicability in continuous monitoring due to its coagulation-free characteristic [91] and high sensitivity towards the change of local tissues. In addition, ISF extraction causes less pain due to the reduction of MN length from 1000–2000 μm to 100–800 μm (Table 1). Shorter length provides MNs with higher yield strength and shear capacity [92,93], so the short MNs for ISF sampling have greater flexibility in design and material selection.

3.2.1. Hollow microneedles

HMN holds promise not only in blood extraction, but also in ISF sampling in a painless manner. Different from extracting blood with a single needle, MN array is generally employed to withdraw enough ISF available in skin (< 1 μL mm⁻² [94]). MN patches, with array density up to 1 × 10⁶ needles cm⁻², MN height of 100–400 μm, and inner diameter of 4–100 μm, have been developed [32,88,95–97]. Short HMNs for fluid collection have been fabricated using Si [32], polymers [88], or titanium [96], but most of them resort to vacuum to extract ISF and do not demonstrate in vivo sampling. Si, although showing high brittleness, has been the most common single material for construction of HMNs [95,98,99], because it allows for accurate microfabrication by photolithography and endows the HMNs with the self-powered ability based on enhanced capillarity. The pioneering work by Mukerjee et al. presented a 20 × 20 array of “volcano-like” HMNs (10 μm in hole diameter) on a bulk Si wafer for in vivo sampling ISF [95]. It took 15–20 min to transfer ISF from human earlobe to a backside reservoir with the aid of capillary action. Strambini et al. further narrowed the inner diameter of each HMN to 4 μm and increased the density to 1 × 10⁶ needles cm⁻² in order to enhance capillary action, so that ISF extraction reached a flow rate of 1 μL s⁻¹ (Fig. 4a) [32]. Recently, cylinder concentric substrate was used to squeeze ISF into a 5-HMN array via local mechanical pressure [30]. A large amount of ISF (up to 20 μL for 1–2 h) from human was successfully collected by the HMN-connected glass capillaries.

Fig. 4.

(a) Optical and SEM images of an HMN patch made of a silicon wafer. Protruding length: 100 μm; pitch: 16 μm; external diameter: 9 μm; internal diameter: 7 μm; HMN area: > 0.5 × 0.5 cm² [32]. (b) Schematic of PMNs-based microfluidic chip for ISF extraction and direct analysis. The optical and SEM images show the PMN made of PDMS and hyaluronic acid [33]. (c) Schematic of hydrogel MNs made of a rapidly swelling hyaluronic acid crosslinked by methacrylic anhydride for ISF extraction [34]. Figures reprinted with permissions from Elsevier, Springer Nature, and John Wiley and Sons.

The high risk of hole clogging is perceived as a major concern. It has been concluded that Si HMNs with straight side-walls have a higher occlusion probability than those with tapered side-walls [40]. However, the tapered HMNs still have the clogging problem since the pore on the HMN tip tends to cut cells during the insertion. Smith et al. moved the pore to the edge of Si MN shank, forming a snake-fang-like HMN, which alleviated the plugging issue [95,97]. Another effective method is to cover the micro-channels on a Si strip with a nanoporous membrane (5 μm in thickness) [100].

3.2.2. Porous microneedles

PMNs offer a high-density network of continuous capillary channels to transfer ISF from the MN-skin interface to the base. Ceramics and metals were used to manufacture PMNs with hole diameter ranging from a few tens of nanometers to several microns [75,101,102]. The formation of porous ceramic or metal structures generally requires high sintering temperatures (e.g., > 1000 °C). Therefore, porous polymers forming at much lower temperatures are a robust alternative. Porous structures for constructing MN have been generated by polymerizing acrylic monomers in the presence of porogens [76,103]. For example, by using PEG as a porogen, Liu et al. prepared a PMN from poly(glycidyl methacrylate) that was crosslinked by triethylene glycol dimethacrylate and trimethylolpropane trimethacrylate with UV irradiation [76]. The PMN patch could trigger biofluid suction by capillary action until ISF infiltrated the patch base. Like Si HMNs, the polymer PMNs are also mechanically fragile. Several teams attempted to strengthen the PMNs through reducing pore diameter and/or density; inevitably, the methods compromised sampling rate and volume [76,101]. More recently, Takeuchi et al. devised an elastic sponge-like PMN using PDMS and hyaluronic acid (HA) to conquer the fragility. The PDMS/HA MN is solid at dry state and turns into an elastic porous structure after HA dissolving, so that it can absorb fluid by manual compression and transfer the specimen to a reservoir with the help of a capillary pump (Fig. 4b) [33]. In another case, Prausnitz et al. proposed to insert porous filter papers between adjacent high-strength stainless steel SMNs [104]. However, the sandwich-structural MN can merely collect 3.3 nL ISF in 20 min. Further to this study, the same team refined their design by attaching the filter papers onto the backside of the SMN, and thus the collected volume of ISF per min reached > 2 μL [77,105].

3.2.3. Hydrogel microneedles

Hydrogel MN is a relatively new type of MNs that is hard under dry state and swells in skin without degradation. An early hydrogel MN patch was manufactured through crosslinking poly(methylvinylether/maelic acid) (PMVE/MA) with PEG [106]. The MN was first utilized for transdermal drug delivery [107], and further used to extract ISF for off-line detection of glucose, lithium, and drugs [39,80]. It required only 5-min insertion to collect sufficient ISF (0.84 μL) from excised porcine skin for analytes detection [79], but the sampling time extended to 1 h for in vivo experiments [39]. In order to shorten the sampling time, Chang et al. employed a super hydrating HA as a hydrogel backbone [34]. They grafted methacrylate on the HA backbones for crosslinking and then initiated free radical polymerization with UV radiation to form crosslinking network (Fig. 4c). The HA MNs successfully reduced the extraction time to 1 min. The large uptake of ISF (1.4 μL) allowed for monitoring subtle changes of ISF glucose/cholesterol levels, showing a similar trend with the real concentrations in blood. Recently, rapid ISF sampling has also been realized via a hydrogel-coated SMN patch, wherein poly(l-lactide)-made SMNs as a mechanical support pierce the skin and Ca²⁺-alginate coating swells to extract fluids with a sampling capacity of ~6.5 μL in 2 min [56,108].

On the whole, significant advances in MN fabrication have demonstrated good feasibility in biofluid sampling over the past 5 years. Advanced MNs have been able to ultra-rapidly extract blood, such as ~100 μL in 3 min using SMNs and > 30 μL in seconds using a single HMN. As for ISF sampling, using capillary effect or material absorption, the different types of superior MNs have already shorten the extraction time to 1 min with the ascending order by sampling volume: hydrogel MN (1–3 μL) ≤ PMN (> 2 μL) < HMN (> 10 μL) (see Table 1). The sampling volume of > 1 μL can meet the need of off-line commercial analysis [34]. Sampling enough biofluids within 1 min already makes MN devices competitive to the traditional needles for blood withdrawal. Among the three kinds of MNs, hydrogel MNs and polymer PMNs are easily fabricated but need additional recovery procedures to unload ISF from the MNs, whereas the ISF extracted by HMNs can be easily transferred out of the MN lumens to the device backside by capillary action. For this reason, HMN collectors have been further developed into on-device sensors of diverse structures, while such devices based on hydrogel or porous MNs have been rarely reported. The following section will describe various integration strategies of HMN-based sensors.

4. On-device diagnosis

4.1. Single-use devices

On-device detection can be realized through affixing a specific analyzer to an HMN patch. The analyzer is typically either face-to-face mounted on the backside of the HMN patch [32,59], or just placed on the backside with connecting fluidic channels, through which biofluids are transferred to the analyzer [89]. Fig. 5a presents a typical HMN-based fluidic platform containing an immuno-modified electrode array for myoglobin/troponin detection [109]. Two syringe pumps enable fluids flow to perform an immunoassay on the antibody-decorated electrodes, followed by applying a substrate solution (3,3′5,5′-tetramethylbenzidine) to the electrodes and a chronoamperometric scan for electrochemical transduction, giving rise to current responses to both proteins in the range of 100–1000 ppb. While this kind of MN sensor minimizes an electrochemical immunoassay into a microchip platform, ISF still needs to be transferred to the microelectrodes via the complex microfluidic channels. The transfer may be limited by small volume of accessible ISF (< 1 μL mm⁻² of skin [94]) and increases the detection time. To avoid ISF transfer and allow low-volume detection, Ranamukhaarachchi et al. integrated an immuno-modified metal HMN with an optofluidic device for vancomycin (VAN) detection (Fig. 5b) [110,111]. The lumen of the HMN was modified using the streptavidin-biotin reaction to bind a high density of peptide for VAN recognition. In this way, the microlumen, as an immunoassay reaction vessel, required only 0.6 nL ISF for analysis. Meanwhile, the high-density peptide allowed a low limit of detection (LoD) of VAN (84 nM) that was estimated from the substrate absorbance in the optofluidic channel using a photo-detector.

Fig. 5.

(a) HMN chip containing microfluidic channels and electrochemical electrodes [109]. (b) HMN optofluidic sensor for drug monitoring: HMN lumen is immuno-functionalized to capture vancomycin [110]. (c) HMN combined with a PDMS switch and a paper-based sensor for multi-diagnosis [89]. Figures reprinted with permissions from John Wiley and Sons and Royal Society of Chemistry.

The HMN-fluidic biosensing platforms show proof-of-concept devices for a single use; however, the function modules of pump, buffer and agent reservoir, optical/electronic components involved in the assay are not assembled into the platforms. These components complicate the device structure and contribute to the cost of the MN sensors, which limits their application as disposable devices. One elegant design to avoid the above issues is the assembly of colorimetric strips on the backside of a HMN patch [89,95,112]. In this design, the fluid is extracted to the testing strip, where analyte is oxidized by enzyme to induce a color change of substrate (e.g., 3,3′5,5′-tetramethylbenzidine). The concentration can be readily graded by naked eye or quantitatively evaluated by color analyzer. Due to simplicity of the colorimetric assay, a full-function sensing system can be easily assembled into a compact device. Fig. 5c presents such a device system including blood processing and multiple-analyte sensing [89]. It is comprised of a PDMS chamber for activation, a metal HMN for blood access, and a paper sensor for diagnostic readout. The PDMS chamber enables the uptake of 30 μL blood onto a sample pad in 30 s. The collected specimen flows through a polysulfone membrane for red blood cell filtration, after which the serum diffuses to two colorimetric reaction zones for glucose and cholesterol. The whole detection procedure is completed within 180 s. While the paper-based MN sensors represent a feasible strategy for disposable POCT diagnosis, they can only detect small-molecule analytes such as glucose and cholesterol. More studies are needed to exploit the test papers enabling protein assay for the HMN platform, so as to detect more biomarkers.

4.2. Continuous monitoring devices

Along with the evolution of disposable MN sensors, a number of research groups have exploited MNs as continuous monitoring platforms by incorporating external electrodes. They have developed a gamut of device architectures, such as defining serpentine Si nanowires on a SMN tip for pH measurement [113], hitching a nanoelectronic thread electrode to carbon/tungsten MN for long-term electrical recording [114], and mounting Ag/AgCl electrodes on a PMN patch for edema monitoring [41].

A commonly used monitoring system is structured by attaching enzymatic electrodes on the base of HMN array [40,99,116]. Fig. 6a presents a typical device for long-term glucose monitoring [59]. A three-electrode system containing a GOx-coated Pt/C working electrode was fixed on the backside reservoir of a Si HMN patch. By relieving skin response with a citrate-containing buffer preloaded in the reservoir, the prototype worked effectively for a wide concentration range (50–400 mg mL⁻¹) and a long period (up to 72 h), as validated by human testing. With similar architecture, Miller et al. extended HMN platform to K⁺ sensing by potentiometry, wherein a solid-state ion-selective electrode (ISE) as transducer was connected to the HMN via fluidic microchannels (Fig. 6b) [115]. The ISE was made up of a K⁺ selective membrane and a 3D porous carbon electrode fabricated by interferometric lithography. The carbon/K⁺ ISE responded linearly to K⁺ (10⁻⁵−10⁻² M) covering the range of normal physiological levels (3 × 10⁻³−6 × 10⁻³ M), indicating that the device held potential in on-body K⁺ monitoring. If aligned with addressable tailored electrodes, a single patch of HMN array can sense multiple physiological markers. Miller et al. developed a highly multiplexed assay of an HMN array/three tailored electrodes sensing system that could detect glucose, lactate, and pH [42].

Fig. 6.

(a) Si HMN patch integrated with three electrochemical electrodes for continuous glucose monitoring [59]. (b) HMN-based microfluidic chip for potassium ions monitoring and SEM images of a 3D porous carbon-based electrode [115]. (c) Pyramidal HMN patch incorporating three electrochemical electrodes for alcohol monitoring [44]. Figures reprinted with permissions from SAGE, John Wiley and Sons, and Elsevier.

Moving electrodes into hollow hole represents a modern strategy in HMN-sensor design. Not only it realizes highly sensitive in situ detection, but also prevents the electrodes from scratch damage during microneedle insertion. An early study reported palladium/aminophenol-modified carbon fibers encased inside the microlumens of a pyramidal HMN array for detecting H₂O₂ and ascorbic acid [66]. Similarly, Wang and coworkers stacked enzyme-modified SMNs on the top of pyramidal HMNs [58]. The SMNs acting as electrodes were electrodeposited with a poly(o-phenylenediamine) (PPD) layer as an entrapping matrix for glutamate oxidase and GOx, as well as an H₂O₂ permselective layer to eliminate other electroactive interferences (e.g., ascorbic acid, uric acid, and cysteine) in ISF. This smart design gave rise to a good linear response to glucose (0–14 mM) or glutamate (0–140 mM). If filled with carbon paste containing metallic micro/nano-particles and oxidases, the HMNs functioned as transducers [43,117]. The renewable carbon paste electrodes supported the actual re-usage of HMNs. Moreover, the high electrocatalytic activity of metal particles towards H₂O₂ allowed for a low operating voltage, e.g., −0.15 V for lactate detection with Rd particles [43]. Carbon paste was also tailored to construct a self-powered biofuel-cell (BFC) glucose sensor [117]. The sensor consisted of a carbon paste-Pt black cathode for oxygen reduction and a carbon paste bioanode containing tetrathiafulvalene (TTF) mediator and GOx. The TTF enabled shuttling electrons from the redox reaction at the enzyme to the carbon electrode [118], so that the BFC sensor continuously harvested power using glucose as a fuel. The power density of the BFC sensor have a linear response to glucose ranging from 0 to 25 mM.

While filling HMN with electrodes exhibits versatility in working electrodes and flexibility in fabrication, these devices do not incorporate counter electrode and reference electrode into a single HMN patch, so external electrodes are required to form a two/three-electrode system to conduct the electrochemical measurement. Mohan et al. presented an integrated alcohol sensor that assembled a Pt wire-based working electrode, a Pt counter electrode, and an Ag/AgCl reference electrode into a single MN patch (Fig. 6c) [44]. The Pt wire was decorated with alcohol oxidase that was sandwiched between size-exclusion PPD film and charged-exclusion Nafion film. The dual permselective layers eliminated interference of common electroactive molecules (acetaminophen, uric acid, etc.) and provided a good linear response to alcohol from 0 to 80 mM.

5. On-MN diagnosis

5.1. Microneedle immunosensors

Procedures involved in the collection/transfer of an ISF or blood sample can be omitted through transforming a SMN into an effective sensor. With the conjugation of antibody/antigen to Si SMNs, Kendall group developed a series of Si SMN-based immunosensors for capturing protein biomarkers from the skin [49,119–124]. They used a hetero-bifunctional carboxyl acid- and sulfhydryl-reactive PEG crosslinker to create linkages between capture antibodies/antigens and gold-deposited Si SMNs. After being applied to the skin for minutes to an hour, the SMNs were peeled off and incubated with dye or enzyme-labeled detection antibodies. The target biomarkers were then identified and quantified by fluorescence microscope or colorimetric enzyme-substrate reactions. This SMN/immunoassay method enabled the sensitive detection of biomarkers, including IgG antibodies [49,120,124], dengue virus NS1 protein [119], and recombinant P. falciparum (rPfHRP2) [122].

Polymers are biocompatible, inexpensive, and easy-processing materials for preparation of MN immunosensors [125–127]. Although polymer SMNs are easily fabricated using micromolding, the bald MN surface requires amine functionalization to immobilize antibodies/antigens. In one study, Yeow et al. utilized electrophilic aromatic substitution and NaBH₄ reduction reaction to yield an amine-enriched surface of polycarbonate (PC) SMNs, to which a hetero-bifunctional PEG derivative (i.e., NHS-PEG-COOH) was linked for influenza vaccine attachment [125]. In another work, antibodies were covalently linked to hexamethylenediamine-modified polylactic acid (PLA) SMNs via homobifunctional crosslinker glutaraldehyde [126]. The former study demonstrated the sensitivity comparable to that of the gold-coated Si SMN probe for anti-influenza IgG detection, and the latter achieved a low LoD in proximity to that of ELISA for interleukins (IL)-1α testing. Moreover, the latter MN patch was functionalized in the “One Row-One Antibody (OPOA)” manner to detect IL-1α and IL-6 simultaneously. A recent work on photonic crystal (PhC) encoded SMNs took a big step forward on multiplex detection of biomarkers (Fig. 7) [127]. Differently colored PhC balls loaded with a specific type of antibody were entrapped in different PEGDA-PEG MNs. Thereupon, captured biomarkers were readily distinguished by reading the refection colors of PhC barcodes. Their concentrations were determined by measuring the fluorescence intensity. Simultaneous detection of three inflammatory cytokines (TNF-α, IL-1β, and IL-6) was performed on sepsis mice using the SMNs contained three colors of PhC barcodes. The PhC-encoded MNs demonstrate greater ease than the OPOA [126] and the LEGO-like design-knitting three smaller patches with different antibodies together into a whole patch [122], since the sensing signals from the OPOA and LEGO-like devices provide only the concentration information.

Fig. 7.

Schematic of (a) photonic crystal encoded MNs and (b) detection of multiple ISF biomarkers [127]. Figure reprinted with permission from John Wiley and Sons.

MN-based immunosensing is a promising method to identify protein markers. The capturing of proteins typically requires from 20 min to a few hours to achieve a comparable sensitivity to ELISA [119,121,128]. Such long sampling time presents a considerable hurdle to its clinical application. Efforts have been endeavored to enhance sensitivity in a short time by engineering MN surface and increasing penetrated surface area of MN. A first attempt was to minimize non-specific absorption by grafting hetero-bifunctional PEG to gold-coated MNs. The PEG could resist over 90% of passive absorption of protein [49]. On this basis, the efficiency for capturing target proteins critically relies on the immobilization strategy of detecting protein. Protein G was proposed as anchor molecules to align the detecting proteins, which improved the capture efficiency but reduced surface density of the proteins as compared to the EDC/NHS crosslinking. The detecting signals suggested that the higher density conquered the better orientation to give a higher sensitivity [119]. The sensitivity could be further improved by optimizing the EDC/NHS crosslinking for the detecting protein [120]. In addition to the surface modifications, Coffey et al. demonstrated that prolonging MN from 40 μm to 190 μm yielded a 4-fold increase in capture amount of specific IgG, enabling rapid extraction of the biomarker within 10 min [121]. Shortly thereafter, the same team increased the MN density from 20,408 MN cm⁻² to 30,000 MN cm⁻² and expanded the array area from 16 mm² to 32 mm² in order to enlarge the penetrated surface area, which obtained a comparable sensitivity in 30 s [123].

5.2. Microneedle electrochemical sensors

For non-metal MNs, conductive treatments are generally required to turn MN surfaces into conductive/sensing electrodes for electrical recording. Multi-walled carbon nanotube (MWCNT) was early reported to selectively grow on a Si SMN array as a working electrode and a counter electrode for continuous glucose detection [64]. In fact, most electrical connections or electrodes were constructed by metal deposition, such as vacuum thermal/electron beam evaporation and sputter coating [53,129–131]. Two types of MNs have been developed to avoid the cumbersome vacuum deposition. One was to blend metal microparticles (palladium) into polymer (PC) solution to directly construct conductive MNs using a micro-molding process. This strategy provided the composite MNs with both catalyzing and monitoring ability towards the ferrocyanide redox [132]. The other one was adoption of hydrogel MN as the entrapping matrix of enzyme (GOx or lactose oxidase) and mediator (vinylferrocene). Caliò et al. demonstrated that the swollen hydrogel (PEGDA) MNs as conduits could couple the electrochemical current of glucose and lactic acid to the electrical contacts on the MN base [61]. Both strategies greatly simplified the preparation process for the polymer MN-based sensors, which involved only the micromolding process.

Selective and sensitive detection of biomarkers depends on redox reactions between target molecules and their enzymes. Enzymes are generally immobilized on the conductive SMNs via an electro-polymerization process, during which the enzymes are entrapped within conductive (e.g., PPD and PEDOT) [58,60] or non-conductive polymers (e.g., chitosan and polyphenol) [44,129]. These polymers provide a stable and biocompatible environment for enzymes, which allows the analyte of interest accessing to enzymes but restrains escape of the large enzymes. For instance, GOx within PEDOT matrix remained a good linear response to glucose between 2 and 14 mM even after the MN sensor was stored in PBS for 7 days [60]. Recently, for sensitivity enhancement, various nanomaterials including metal NPs/polymer nanofibers [133], Cu nanoflowers [134], Au-MWCNTs [62], etc., have been applied to MNs as supporting structures of enzymes. The nanostructures not only provide large electroactive surface areas, but also enhance the conductivities between electrodes and enzymes. For example, Chen et al. deposited Au/Pt NPs, Pt NPs/polyaniline nanofiber, and dual permselective layers on a stainless steel MN for glucose monitoring (Fig. 8a) [133], achieving a low LoD (0.1 mM) with a broad linear range (up to 20 mM), while Bollella et al. construed a sponge-like surface using MWCNTs and polyMB for lactate detection (Fig. 8b) [62], obtaining a high sensitivity of 1473 μA cm⁻² mM with a low LoD of 2.4 μM. However, a typical enzymatic MN sensor has to experience enzyme degradation, complicated immobilization procedure of enzyme, and oxygen limitation.

Fig. 8.

(a) Schematics of (left) functionalized stainless-steel MN (three main components: Au/Pt NPs, PVDF/Nafion layer, and glucose oxidase) and (right) glucose monitoring [133]. (b) Microscopic structure of (left) pMB-MWCNT composite and (right) SMN surface (four layers: gold, Au-MWCNTs, poly-methylene blue (pMB), and lactate oxidase) for lactate detection [62]. (c) Workflow for surface modification of a polycarprolactone (PCL) MN for real-time NO monitoring [50]. (d) Schematic of ZnO NWdecorated stainless steel MNs for H₂O₂ detection and SEM images of vertical ZnO NWs synthesized on MN surface (top) and protected by PVP (bottom) [135]. Figures reprinted with permissions from Elsevier, John Wiley and Sons, and American Chemical Society.

Non-enzymatic MN sensors have been exploited to overcome the critical drawbacks of the enzymatic sensors [136]. While Pt catalyst is highly active for glucose electrooxidation, smooth Pt electrodes are subject to low sensitivity, poor selectivity, and poisoning by adsorbed intermediates [137,138]. Nanoporous Pt black was electroplated at the tips of stainless steel MNs to enlarge the electroactive surface over 440 times larger than that of bare Pt electrode, favoring a high selectivity of 1.48 μA mM⁻¹ cm² towards glucose ranging from 2 to 36 mM [65]. The Pt catalytic surface/ability was further increased by combination of Pt NPs and MWCNTs, giving rise to a higher sensitivity of 17.73 μA mM⁻¹ cm² over the range 3–20 mM [64]. However, biofouling from the foreign body reaction accumulated on the surface of the Pt electrodes during a 7-day rabbit test, for which less and less reliable sensing was witnessed since day 5 [65]. Overcoating a biocompatible membrane would probably improve the biocompatibility of Pt NPs-based electrode.

In addition to glucose sensing, electrochemical on-MN sensors have recently emerged for monitoring NO, pH, and H₂O₂. NO has been recognized as an important messenger molecule under different physiopathological conditions (e.g., a physiological indicator of colorectal cancer) [139]. Keum et al. mounted a hemin/PEDOT-modified PCL MNs on an endomicroscope to construct an instrument that enabled both real-time monitoring of NO in colonic polyps and detailed imaging of the lesions [50]. The PCL MN was exquisitely coated with hemin/PEDOT using polydopamine as a binder (Fig. 8c). The hemin offered a p-type doping effect for PEDOT via π-π stacking as well as high specific affinity toward NO [140]. The presence of NO reduced hemin (Fe³⁺) to heme (Fe²⁺), which resulted in current reduction of the PEDOT coating. Periodical response to NO with micromolar sensitivity was demonstrated by alternately applying the MNs in normal and melanoma tissues. Recent field-effect transistor sensors demonstrated that conjugation of graphene or reduced graphene oxide (rGO) with porphyrins (e.g., hemin) significantly promoted selectivity and sensitivity for NO detection [140,141]. Following this design, Tang et al. synthesized iron-porphyrin functionalized graphene (FGPC) and deposited the FGPC on a PEDOT/Au NPs-coated MN [142]. The composite layer enhanced electron transfer and favored very high specific surface area for NO catalyzing [143,144], endowing the MN sensor with a LoD down to nM level.

Microneedle pH sensors were also developed by surface grafting of various pH-responsive materials [67,130,145,146]. ZnO was reported as pH sensing layer on a tungsten (W) MN because of its biocompatibility, chemical stability, and amphoteric properties [147,148]. The ZnO-W MN sensor could measure pH of 2–9 with a sensitivity of −46.35 mV pH⁻¹, and completed the pH measurement of a mouse brain or bladder within 60 s [67]. A higher sensitivity of −51.2 mV pH⁻¹ was achieved using a composite of molybdenum disulfide (MoS₂) nanosheets and polyaniline (PAN) on a stainless steel MN [146]. The enhanced sensitivity attributed to high specific surface area of MoS₂ provided for H⁺ sensitive PAN. The pH sensitivity could be further lifted to a super-Nernst response of 141 mV pH⁻¹ by modifying a W MN with boron doped diamond (BDD), but the preparation of BBD layer involved microwave plasma-assisted chemical vapor deposition [145], leading to a complex fabrication process. Noted that the aforementioned pH responsive materials were constructed on a single metal MN without base, having disadvantages of poor reproducibility in device fabrication and varied penetration depth in measurements. Combination of a MN patch array (MPA) with pH sensing materials could avoid the issues and, moreover, provide a unique advantage in monitoring spatial distribution of pH, as demonstrated by Zuliani et al. using an iridium oxide (IrOx)-modified MPA for pH distribution mapping of an ex vivo rat heart [130].

The MN sensors decorated by nanomaterials exhibit superior sensing performances arising from the high specific surface area and electrocatalytic activity; however, the surface nanostructures on the MNs are susceptible to mechanical damage upon skin insertion. In a recent scheme by Xie et al., a dissolvable polymer PVP was sprayed over MN working electrodes as protective layers (Fig. 8d). The PVP layer with a thickness of 2 μm could preserve the integrity of vertical ZnO nanowires [135] or nanohybrids of rGO and Pt NPs [149] on the MN surface during insertion and dissolved in the skin in 5 min. As such, the PVP-protected ZnO/MN sensor was three times more sensitive than that of the unprotected ZnO/MN sensor for H₂O₂ monitoring.

5.3. Microneedle biopotential recordings

Biopotentials associated with physiological processes are electrical signals generated inside the body, including, but not limited to, ECG, EEG, EMG, and electrooculogram (EOG). Such biopotentials provide biomarkers for pathological and physiological state and are often recorded using wet electrodes. However, the conventional wet electrodes for biopotential recording require tedious procedures, such as skin preparation and gel usage in order to decrease electrode-skin interface impedance [53,54]. The slow rate of gel drying is prejudicial to long-term monitoring [131,150]. Moreover, motion artifacts arising from electrode loosening and skin-potential variation (SPV) distort the signals [69,150]. Thereupon, microneedle electrode arrays (MNEAs), a novel dry electrode, were exploited to address these issues. The MNEAs penetrated through the SC to fasten the electrodes, which provided stable signals in motion states and low impedance density (e.g., 7.5 kΩ cm²@10 Hz) [53]. To further minimize motion artifacts, other types of MNEAs were developed on curved [151] or flexible substrates (e.g., parylene [53], SU-8 [52], polyimide [150], and PET [70]) to accommodate body movements. Besides, Pei et al. proposed a MN-based strategy to eliminate the SPV effect. The main parts of MN electrodes were overlaid with an insulative parylene layer to avoid contact with the stratum corneum, while bare conductive MN tips exposed to the stratum germinativum [69]. This method effectively suppressed the SPV interference in a motion state. Furthermore, MN patch with high-density electrode array held a unique advantage in mapping complex muscle activity with a special pattern [54,151].

6. Conclusions, challenges, and perspectives

Over the past decades, great strides have been made in structural materials, device designs, and detection strategies to broaden the application of MN-biosensors. Based on a rich variety of detection methods, currently available MN-biosensors are capable of sensing 25 analytes (Table 2), and most are small molecules. Suitable target analytes include, but are not limited to glucose, proteins, ions, drugs, metabolites, biopotentials, and plant DNA [152]. With the rapid development, a few seconds or minutes are enough for biofluid transfer, analyte diffusion, or biorecognition (Table 1). The long collection period for proteins has been reduced to tens of seconds [123]. Recently, hydrogel-coated MNs showed the capablility of transdermal sampling of NA in 15 min and immune cells in 12 h [56,108]. The NA bound to the hydrogel hybridized with complementary target DNA, the concentration of which was determined by spectrophotometry (Fig. 9a). Aside from NA, the hydrogel can be loaded with antigen nanocapsules that solicit antigen-presenting cells (APCs) to recruit antigen-specific T cells into the hydrogel under the skin (Fig. 9b). With the remarkable past work paving the way, the MN-sensor research is on the way to a brilliant future in spite of many hurdles remaining.

Table 2.

Analytes detected by various MN sensors.

Analyte	Sensor structure	Detection method	Test subject	Detection site	Ref
Glucose	Hydrogel SMNs	Glucose assay kit	Mouse	Off device	[34]
	Metal HMN + Paper sensor	Colorimetry	Rabbit	On device	[89]
	Si HMNs + Sensor	Electrochemistry	Human	On device	[40]
	Metal SMNs + PEDOT	Electrochemistry	-	On MN	[60]
Cholesterol	Hydrogel SMNs	Cholesterol assay kit	Mouse	Off device	[34]
	Metal HMNs + Paper sensor	Colorimetry	Rabbit	On device	[89]
Glutamate	Polymer HMNs + SMNs	Electrochemistry	-	On MN	[58]
Lactate	Polymer HMNs + Carbon paste	Electrochemistry	-	On MN	[42,43]
	Polymer SMNs + MWCNTs	Electrochemistry	-	On MN	[62]
Influenza IgG	Si SMNs + Antigen	ELISA	Mouse/Swine	On MN	[49,121,128]
NS1 Protein	Si MNs + Antibody	ELISA	Mouse	On MN	[119]
Pf HRP2	Si MNs + Antibody	ELISA	Mouse	On MN	[122]
IL-lα, IL-6, TNF-α	Polymer SMNs + Antibodies	ELISA-blotting method	Mouse skin, Ex vivo	On MN	[126]
	Polymer SMNs + PhC barcodes	ELISA + Spectra	Sepsis mouse	On MN	[127]
Myoglobin;  Troponin	Polymer HMNs + Immunoelectrodes	Electrochemical immunoassay	-	On MN	[109]
Ascorbic acid	Polymer HMNs + Carbon fibers	Electrochemistry	-	On MN	[66]
K+	Polymer HMNs + ISE	Electrochemistry	-	On device	[115]
	Metal SMN + ISE	Electrochemistry	Chicken/porcine skin, Ex vivo	On MN	[170]
NO	Polymer SMNs + Hemin + Endomicroscopy	Electrochemistry	Melanoma mouse	On MN	[50]
	Metal SMN + FGPC	Electrochemistry	Rat	On MN	[142]
pH	Metal SMN + BDD	Electrochemistry	Mouse stomach, Ex vivo	On MN	[145]
	Metal SMN + ZnO	Electrochemistry	Mouse	On MN	[67]
	Polymer SMN + IrOx	Electrochemistry	Rat heart, Ex vivo	On MN	[130]
H2O2	Metal SMN + Pt/rGO	Electrochemistry	Mouse	On MN	[[149]
Alcohol	Polymer HMNs + Electrodes	Electrochemistry	Mouse skin, Ex vivo	On device	[44]
Vancomycin	Metal HMNs + Peptide	ELISA-optofluidic detection	-	On device	[110]
β-Lactam Antibiotic	Polymer SMNs + g-lactamase	Electrochemistry	-	On MN	[51]
Organophosphate	Polymer HMNs + Carbon paste	Electrochemistry	Mouse stomach, Ex vivo	On device	[44, 46]
Levodopa			Porcine skin, Ex vivo		[48]
Tyrosinase					[47]
DNA-210	Polymer SMNs + PNA oligomers	PNA hybridization	Human skin, Ex vivo	On MN	[56]
T cell	Polymer SMNs + Antigen Nano capsules	Immune response	Human skin, Ex vivo	Off device	[108]

Fig. 9.

Schematics of (a) PNA-modified hydrogel/SMN for sampling NA [56] and (b) Alginate-coated SMNs for capturing T cells. Antigen-loaded nanocapsules are encased in the alginate layer [108]. Figures reprinted with permissions from American Chemical Society and The American Association for the Advancement of Science.

Antibodies and antigens constitute one of the largest and most important classes of disease biomarkers; however, there is relatively a low equilibrium concentration for most proteins in ISF owing to their low microvascular permeability [153]. Our group found that the illumination on the skin for a few seconds with a 532-nm pulsed laser resulted in the 1000-fold concentration increase of circulating biomarkers (e.g., IgG) in the upper dermis [128,154]. When inserting MNs covalently linked with influenza antigens into the laser-treated skin, influenza antigen-specific IgG was readily detected by the MNs using mouse and pig models receiving influenza vaccines, with the sensitivity, specificity, and accuracy comparable to the immunofluorescence assay of blood samples [128]. The laser-induced capillary extravasation depends on the preferable absorbance of specific light by hemoglobin, which causes the thermal-induced dilation of capillary beneath the skin [155,156]. Likewise, Coffey et al. manually applied the local mechanical stimuli (i.e., 1–10 MPa pressure) on the skin to enhance the leakage of proteins through vessels [157]. Increasing the capillary permeability seems to be a viable strategy that enables the MNs to measure blood biomarkers in the epidermis without damage to blood vessels.

MN-based sensing for POCT is still in an immature phase. Only long-needle-based sensors (≥5 mm in length) for monitoring ISF glucose (e.g., Medtronic Guardian™ Sensor and Abbott’s Libre system) have already hit the store shelves [158]. In fact, more MN devices are being commercialized as a blood/ISF collector, such as TAP (Fig. 3b, Seventh Sense Biosystems Inc.) [78], HemoLink (Tasso Inc.), and an ISF absorber (Renephra Ltd.). Such devices depend on a vacuum to painlessly withdraw blood/ISF from the skin with MNs. Of note, TAP comprised of an array of 30 microneedles has received clearance from the U.S. Food and Drug Administration (FDA) in 2017 and CE Mark in 2018. So far, the FDA has just released a draft guidance to detail when a MN device should be subject to the regulation as a product, but the classification has not been made. It may take years or a decade to commercialize MN-based sensors, because it needs to be repeatedly tested for reliability, biosafety, and accessibility prior to clinical use. Design factors affecting the reliability include the length and surface uniformity of MNs and the applied mechanical stimulus [123,157]. These factors may also vary with patients regarding their age, weight, body location, and operative habit. For instance, finger press led to insufficient penetration for the MNs with ~550 μm in length or penetration failure for those with 300 μm in length [159–161]. Although an auxiliary applicator or a force sensor can ensure successful penetration [162,163], inclusion of an adequate control to standardize each assay may be more helpful for reliable results. However, whether this can address the reliability issue remains undetermined. On the other hand, both sterilization and disposal are in need of particular attention. The risk of infection is high due to the inappropriate handling and disposal of MNs [164,165]. For this reason, manufacturers are required to: (1) attach a disinfectant to MN products; (2) introduce a self-sterilizing coating onto MN surface; or (3) explore an additional enclosure for retracting or blunting MNs after use [166,167]. For a workable sensor, data collection, processing, and readout modules are indispensable. The design of MN-based biosensors that are eligible for commocialization has to take more into account, such as simple fabrication, portable size, affordable price, and adequate reliability.

In our view, research in MN-based sensors will grow rapidly in the upcoming decade. One of the key emerging trends would be the convergence of MN-biosensors, optoelectronics, and wireless communication for a comprehensive ubiquitous healthcare solution. For example, Wang et al. built up the first prototype of standalone, wearable MN-biosensor comprising HMN arrays (analyte measurement), a flexible circuit board (data processing), a Bluetooth module (data transmission), and a battery [47]. The strip-shaped device acts as a bandage to adhere to the skin, which significantly enhances a user’s comfort. Similar products came up soon on the horizon that requires the MN patches to perfectly accommodate skin deforming for a good fit [168,169]. It can be envisioned that MN-based biosensors would play a central role in a comprehensive healthcare service in the future, as depicted in Fig. 10. The wearable devices should be embedded with various smart sensors, therapeutic units, and wireless systems: (1) sensors quantify disease biomarkers and upload patients’ medical records to cloud storage; (2) diagnosis is remotely performed by physicians or even artificial intelligence; (3) medical treatment decision triggers the release of preload drugs directly into the patients’ bodies. Apparently, patients would benefit from the decentralization of healthcare to manage their own health at home, which may also help address the physician shortage. Overall, it is essential for academia, industry, and government to work closely so as to transform the innovations in the field into commercial products for better and more effective healthcare.

Fig. 10.

Conceptual diagram of wearable MN-based theranostic system for comprehensive personal healthcare in the near future.