Recently, drug-coated solid as well as hollow MNs made from a variety of materials have been investigated to deliver model drugs of various MWs, and nano- or microparticles into the eye either via intracorneal or intrascleral routes [19–21, 22]. Drug-coated stainless steel solid MN was able to deliver only 69 % of the applied dose. The rest of the dose either remained adherent to the MN which was likely due to the incomplete MN insertion into the tissue or may have deposited on the scleral surface affecting dose accuracy and reproducibility . Solid silicon or hollow glass-based MNs are intrinsically brittle and can be broken off accidently during application (thereby remaining in the tissue), and fabrication of these needles is rather complicated and expensive. Furthermore, hollow MNs cannot deliver solutions without MN retraction or use of tissue dissolving enzymes to provide free space for accommodating drug solution that requires careful infusion at a predetermined rate. Drug solution infusion through MNs also requires operation at high pressures (250–300 kPa), which in turn is related to the viscosity of the solution and the geometric properties of the MNs [20, 21]. Uncontrolled retraction from the sclera could lead to removal of the MN and leakage of the drug onto the scleral/corneal surface affecting the amount of drug delivered. Thus, special insertion devices and infusion systems are required to enable MN injection in a controlled manner. Consequently, there is a distinct need for the development of other types of MNs to address these shortcomings. Therefore, this study used PVP-based dissolving polymeric MNs.
A wide range of polymers has been used to fabricate dissolving MN arrays including polyvinylpyrrolidone (PVP) . PVP is a homopolymer formed by the monomer N-vinyl-2-pyrrolidone. It is an FDA-approved biocompatible polymer and has been widely used as a blood plasma expander and in pharmaceutical industry for several purposes including as a tablet binder  and in fabricating dissolving MNs for transdermal drug delivery [25–27]. Its wide use can be attributed to a number of properties, including excellent water solubility, biological compatibility, low toxicity, film forming and adhesion, complexing ability, inert behavior towards salts and acids and resistance to thermal degradation in solution . One of the enticing properties of using PVP as the structural material for polymeric MNs is the ability for rapid dissolution within the biological tissues due to the high water solubility, which is deemed essential for ocular drug delivery as it is very challenging to apply the MNs patch on the eye for long periods of time. FS and FITC-conjugated dextrans have been widely used as model molecules in investigations of permeation across ocular tissues, and it was deemed to be stable and reliable tracers [33–35]. Therefore, we have used these molecules in our study as model compounds.
In literature, some studies have reported using PVP to prepare dissolving MN arrays e.g. PVP K29/32 hydrogel at 40 % w/v  or 50 % w/v of either PVP K15 or PVP K29/32 hydrogel . In this study, PVP MN arrays prepared from various MWs were found to be sufficiently rigid from all tested PVP hydrogel formulations. The microscopic inspection of the fabricated MN arrays (Fig. 1) revealed that conical-shaped MN arrays (3 × 3) with sharp tips were formed that were around 780 ± 60 μm in height, 300 ± 40 μm in base width and 50 μm of interspacing between the MNs. However, higher polymer concentrations were required to fabricate MNs from low MW PVP hydrogels. The required polymer concentration used in the preparation of hydrogels from PVP K15, PVP K30 and PVP K29–32 were 60, 40 and 30 % w/v, respectively. Essentially, we were also able to encapsulate the three model drugs, with increasing MWs and at different concentrations within the MN arrays, while maintaining their mechanical properties (discussed later). However, it was noticed that using the same PVP hydrogels to form the MN baseplates produced very brittle and fragile MN baseplates. Therefore, a high MW PVP (i.e. PVP K90) hydrogel prepared at 15 % w/v was used to from a more a resilient and rigid baseplate for all the MNs tested in this study.
To ensure dosage accuracy and reproducibility, the drug content of MNs was determined (Table 2). Drug recovery from MNs was found to range from 92 ± 5 % to 97 ± 4 %. As shown in Table 2, the content of drug in the MN arrays (i.e. total amount present in nine needles) ranged from 0.96 ± 0.13 to 9.91 ± 0.28 μg for different MN formulations. Importantly, it is of great interest that we were able to incorporate approximately 1.3 and 10 μg of a high MW molecule (i.e. FD 70) in the PVP MNs when loaded at 2 and 10 mg/ml, respectively. A similar trend was noticed for other model drugs. Thus, dissolving MNs could be used to deliver clinically relevant doses of highly potent high MW anti-VEGF drugs such as Avastin® (bevacizumab, ≈149 kDa). For example, a study by Kim et al.  concluded that a dose of 4.4 μg of bevacizumab is needed to provide the same effect as that of 2.5 or 52.5 mg when delivered via the subconjunctival route or eye drops to treat CNV. Moreover, increasing the number of MN arrays can further increase drug content in the PVP MNs. However, it is important to remember that high density of MNs, unlike for transdermal application, can impair MN penetration into ocular tissues due to its curved surface or a bed of nail effect can occur.
Moisture content plays an important factor on mechanical properties of the MNs including rigidity, flexibility and dissolution kinetics within the ocular tissues that in turn effects drug stability and efficacy. High content of moisture will impair MNs ability to penetrate the tissues. In general, the moisture content of the MN arrays ranged from 7.21 ± 0.65 to 10.53 ± 1.03 %, which is dependent upon the PVP MW and type and concentration of the encapsulated drug, as shown in Table 2. Any level of moisture can affect the stability of drug molecules; however, this is particularly important for biologics that are high sensitive to moisture. Although these MN arrays were fabricated under normal lab conditions, fabrication of these MNs in a controlled environment (temperature and humidity) followed by low-temperature drying can further decrease the moisture content of MNs thereby enhancing the storage stability of biomolecules. Nevertheless, at these moisture levels, we found that the MNs maintained their mechanical strength and effectively penetrated the ocular tissues.
MN arrays were also tested for intraocular insertion forces. Prior to this, the isolated porcine corneal and scleral tissues were measured for thickness after hydration for 2 h. The average thickness was found to be 950 ± 70 and 700 ± 300 μm, for the cornea and sclera, respectively. Therefore, the MN height was chosen to be 800 μm to allow intrascleral or intracorneal drug delivery using MN arrays. Applying the MN arrays onto the tissues requires an appropriate force to enable MN penetration which should be maintained until MN array is completely dissolved to deliver its payload into the tissue. Literature review revealed no data has been reported concerning the required force for intraocular MN arrays insertion. Therefore, this study endeavoured to measure the required force (N) to insert MN arrays into either the corneal or scleral tissues in vitro. The applied insertion force/array vs MN insertion depth is reported in Fig. 2. In general, the fabricated MN arrays were sufficiently rigid to be inserted into the ocular tissues. In both ocular tissues, the required force to insert the 3 × 3 MN array was proportional to the insertion depth. This is can be attributed to the conical shape of MNs, where the MN diameter increase from tip to base thereby requiring higher forces with an increase in the depth of penetration. Interestingly, the required force to insert the MN array into the corneal tissue was higher than that for the scleral tissue. For example, the average insertion force required to fully insert the MN array into the corneal tissue was 3.72 N/9 MNs, which was nearly threefold higher than that required to be fully inserted into the scleral tissue (1.34 N/ 9MNs). Thus, the required insertion force per single MN is 0.41 and 0.15 N for the cornea and sclera, respectively.
In the literature, forces to penetrate various areas of the eye wall using hypodermic needle were reported [38, 39]. For example, Matthews et al. reported scleral penetration forces of around 1.0 N, when a single 18 G hypodermic needle with an outer diameter (OD) of 1.27 mm was used. However, the force required to penetrate the central cornea was significantly lower than all other areas i.e. around 0.5 N . Although this contradicts our data, this study was conducted on the human eyeballs, unlike the porcine tissues reported in our study. Tissue type (porcine vs human) and intraocular pressure play a great role in defining the insertion forces. It is well documented that the cornea also has highly ordered parallel fibrils that make the tissue highly resistant to deformation at raised intraocular pressure. Upon the application of a tensile stress, a restoring force is generated by the stretched fibers that balances the applied force and resists deformation, providing the cornea with substantial tensile strength . Thus, higher insertion force is required in the cornea compared to the sclera. But this can be explained on the basis that Matthews et al. have used hypodermic needles which differed from our study in relation to materials, dimensions and geometrics. Pulido et al. found peak needle penetration forces for the anterior sclera of 0.29 and 0.61 N with 30 G (OD 310 μm) and 27 G (OD 410 μm) needles, respectively . The 30 G hypodermic needle is very closely related to our fabricated dissolving MNs in its OD, which is around 300 μm. However, the required insertion force was twofold higher than in our case. This can be attributed to the fact that our MNs are of conical shape while the hypodermic needles are of cylindrical shape. Thus, our MNs allow minimally invasive means of drug delivery.
Further studies were performed to evaluate the ability of MN arrays to penetrate through the ocular tissues using OCT scanning. Images of MN arrays following their in vitro insertion into porcine corneal and scleral tissues are shown in Fig. 3. A previous report has shown that only 200 μm of the 650 μm height of a PVP-based 100-MN array was inserted into cadaver porcine skin . Interestingly, the results here clearly demonstrate that all fabricated MN arrays from the three different PVP polymers i.e. PVP K15, PVP K30 and PVP K29–32 were strong enough to penetrate the ocular tissues in vitro without bending or breaking. The insertion depth was around 75 ± 10.8 % (n = 3) of the total MN height (data not presented). This can be attributed to that of our MN arrays as they are composed of fewer MNs (only nine MNs) with different geometrics, and the use of ocular tissues instead of skin, which has differed mechanical properties. The partial insertion of MNs can be ascribed to the use of the highly elastic Parafilm® layer (to avoid dissolution of MNs during imaging) hindering the MN arrays from being fully inserted into the ocular tissue. However, this should not be of great concern due to the fact that in real situations, MN arrays will be applied directly into the ocular tissues without any barrier. Thus, high percentage (>75 %) of MN insertion will be expected.
The ability of dissolving MN arrays to withstand the forces required for insertion into the designated biological tissue or any additional accidental forces while inserting is the key factor in their successful development and application. Mechanical strength of the dissolving MN arrays is widely affected by several factors including polymer type and concentration, moisture content, encapsulated drug type and concentration and preparation methods [26, 41]. Therefore, in this study, we evaluated the mechanical strength (fracture force) of those MN arrays prepared from either both plain PVP and PVP/drug hydrogel formulations. A Texture Analyser (in compression mode) was used to apply predefined forces including those required for MN insertion and other higher forces (1.5, 3, 6 and 9 N). The percentage of MN height reduction was calculated and plotted against the applied compression force and is presented in Fig. 4. Images of some samples MN arrays after applying predefined forces are presented in Fig. 5.
As represented in Figs. 4 and 5, PVP polymers at the specified hydrogel concentrations can form sufficiently rigid but brittle MN arrays. PVP K15 produced the most brittle MN arrays that showed around 18 % reduction in the MN height at a force of 3 N/array, but PVP K30 and PVP K29/32 MNs showed around 12.9 and 12.4 % reduction in height when exposed to the same force. Additionally, data shows that there is a proportional relationship between the applied compression force and the percentage of reduction in the MN height, and this is more profound in MN arrays fabricated from PVP K15. PVP K29/32 (P1) formed the most rigid and least brittle MN arrays, which can withstand high forces with minimal height reduction. For example, applying forces between 1.5 and 9 N/array caused an average height reduction from 5 to 32 %. PVP K30 (P2) comes second in MN rigidity. However, no significant (p > 0.05) difference in the mechanical strength (fracture force) was noted between MN arrays when fabricated from PVP K30 or PVP K29/32 hydrogels. This can be ascribed to that both polymers have an intermediate and closely related MW i.e. 40 k and 58 k Da, respectively.
Interestingly, encapsulated model drugs of different MWs at a concentration of 2 mg/ml in hydrogel did not significantly affect the mechanical properties when low compression forces were applied such as 1.5 and 3 N/array. However, the effect was more profound with high compression forces such as 6 or 9 N/array. But statistically no significant difference (p > 0.05) was found among the three model drugs in terms of the effect on the mechanical strength within the same MNs and at same concentration (Figs. 4b and 5). On the other hand, increasing the encapsulated drug from 0 to 10 mg/ml in PVP K29/32-based MN arrays produced a gradual decrease in the mechanical strength of the MNs. For example, applying 3 N/array on MN arrays containing no drug, FD 70 at 2 mg/ml or FD 70 at 10 mg/ml, caused a height reduction of around 12.4 ± 2.5, 12.4 ± 3.5 and 15.7 ± 2.3 %, respectively. This can be attributed to a possible interaction between the FD 70 and the PVP polymer which may affect the polymer configuration upon recrystallization during the drying process. Our PVP K29/32-based MNs were still significantly stronger than other PVP-based MNs reported in literature, where Sullivan et al. found the fracture force of MNs to be 0.13 ± 0.03 N per needle . This might be due to the method of MN fabrication which employed the polymerization technique.
Understanding the behavior of MN after insertion into the ocular tissue is essential in the design and development of an effective MN array and to determine optimal conditions in terms of the MN application method. Therefore, the dissolution kinetics of the PVP-based MN arrays was evaluated following insertion into the porcine ocular tissues, as shown in Fig. 6. Although, the water content of porcine corneal tissue (around 78 %) is higher than sclera tissue (around 71 %) [28, 42, 43], our data showed that the PVP-based MN arrays dissolved faster in the scleral tissues (Figs. 6 and 7b) than in the cornea (Figs. 6 and 7a). This could be due to the lipoidal nature of the corneal epithelium, and it is also depended upon the degree of hydration of the tissues prior to experimental setup. With regard to the effect of PVP’s MW on the dissolution rate, results indicate MNs fabricated from PVP K15 were completely dissolved within 30 s following insertion into the sclera and less than 60 s in the cornea. Interestingly, within the first 10 s of insertion, around 80 and 60 % of the PVP K15-based MNs was dissolved within the sclera and cornea. PVP K30-based MN arrays dissolved in similar fashion but at slower rate in comparison with those MN arrays fabricated from PVP K15. In contrast, PVP K29/32-based MN arrays completely dissolved within 120 and 180 s following insertion into the porcine scleral and corneal tissues, respectively. Therefore, given the similarity of porcine and human ocular tissues, we expect that our MN dissolution in the human eye could also be occur within similar magnitude.
Our data revealed that PVP-based MNs dissolve faster in ocular tissues than in the skin. This was expected in comparison to the data in literature, if we take into account that the ocular tissues vary in their properties and water content compared to the skin . In fact, it was reported that after insertion into cadaver porcine skin, PVP-based dissolving MNs showed significant dissolution within 1 min, and after 5 min, the MNs were 89 % (by mass) dissolved . With regard to the effect of the encapsulated drug type and concentration on the dissolution behaviour of the fabricated MN arrays, results (not presented) indicated that drug MW at low concentrations (2 mg/ml) had no significant effect. However, increasing FD 70 concentration from 2 to 10 mg/ml resulted in faster dissolving MN arrays. Overall, results revealed that PVP K29/32 MNs provide rapid dissolution in ocular tissues, which make them as a possible candidate in the fabrication of MN arrays for ocular application. However, using PVP K15 MNs are not feasible as MN arrays could dissolve very quickly even before complete insertion into the ocular tissues thereby depositing its payload on the ocular tissue surface instead within the tissue.
In vitro intraocular drug distribution studies
In addition to insertion of MNs into the tissues, it is also essential to get insight into drug distribution within the ocular tissue after application. Therefore, in this study, the fluorescently conjugated model drugs were traced within the ocular tissue using scanning confocal microscopy, as shown in Figs. 9 and 10. Results showed that MN arrays fabricated from PVP K29/32 polymer were able to penetrate both the porcine sclera and cornea. Although, the MN heights are around 800 μm, they did not completely pierce either ocular tissue. This can be seen clearly in Figs. 8c and 9c. It suggests that dissolving MNs start dissolving as soon as they are inserted into the ocular tissues and do not pierce it completely (as proven in OCT studies). This is consistent with our findings from dissolution studies in the previous section where we found that more than 20 and 40 % of the MNs dissolved within 10 s after insertion either in the cornea or in the sclera. This could be of great benefit, as there will be no concerns of causing damage to the innermost sensitive ocular tissues such as the choroid/retina beneath the scleral tissue. Furthermore, the confocal images show that macromolecules such as FD 70 cannot permeate easily through the scleral or corneal tissues. This can be clearly concluded from images of the corneal and scleral tissues after 1 h from application of FD 70 aqueous solution (Figs. 8d and 9d). Our findings completely agree with data reported in literature that the cornea and sclera form a strong barrier to molecules larger than 5 k Da. Therefore, large molecules such as FD 70 are more likely to accumulate in the outer surface of the sclera and cornea, when they are applied topically [45–47]. Essentially, as we can see in Fig. 9a, dissolving MNs were able to penetrate the epithelium (outer layer of the cornea with thickness of around 50 μm), which is deemed the main barrier for transcorneal drug permeation and then reached the stroma and dissolved to release the payload that led to formation of drug depot (Fig. 8b, c). Similarly, dissolving MNs were able to penetrate the scleral tissue (Fig. 9a, b) to form a drug depot within the tissue that diffused to the other side of the sclera (Fig. 9c).
In terms of the small MW hydrophilic model drugs (i.e. FS), MN arrays successfully delivered the drug into the ocular tissues (data not shown), which then distributed evenly within the tissue. However, no significant difference was seen in distribution of FS within the tissues from either inserting MNs containing FS or application of FS aqueous solution topically. Our finding are in agreement with data reported in literature that FS, a water-soluble compound with a MW of 376 Da, penetrates into and permeates across both the corneal epithelial and scleral membrane [45–47]. Thus, using dissolving MN arrays to deliver such hydrophilic drugs with small MW do not offer great advantages in comparison to topical drops, apart from increasing their retention time in the ocular tissue. Accordingly, dissolving MN arrays could be useful in facilitating delivery of macromolecules by two mechanisms. For macromolecules, by bypassing the main barrier for their permeation either across the epithelium in cornea or in the sclera and reducing the diffusion time through the ocular tissues and by increasing their retention time in the tissue thereby preventing their removal by blinking and tear secretion upon topical application. And, for small molecules, MNs can increase their retention time in the ocular tissues.
In vitro permeation studies
Taking into consideration the properties of rapidly dissolving MN arrays fabricated, from the three PVP polymers i.e. PVP K15, PVP K30 and PVP K29/32, we have selected PVP K29/32 MNs to investigate the efficiency for ocular drug delivery via intrastromal and intrascleral application. Drug permeation studies were performed using either these MN arrays or predefined volume of the respective drug aqueous solutions. Data concerning drug permeation profiles through the cornea and sclera are presented in Figs. 10 and 11, respectively.
For the small MW molecule (i.e. FS), a high permeation (Figs. 10a and 11a) was seen across the corneal and scleral tissues. The percentage permeation across the sclera was nearly the same from either topically applied aqueous solution or following MN application. For example, 86 and 99 % of FS permeated after 24 h from MNs or solution, respectively. In contrast, the percentage permeation across the cornea was significantly higher for the topically applied aqueous solution (79 %) when compared to MNs (53 %). In general, the drug first permeated and increased at a constant rate until 6 h, after which the drug permeation rate decreased significantly. This could be due to FS binding strongly to the scleral and corneal tissues. The binding was likely to be mediated through ionic interactions since fluorescein has three dissociation constants that correspond to its three pKa values: 2.13, 4.44 and 6.36, respectively. At pH 7.4, the anionic forms of FS may have bound to positively charged extracellular matrix molecules (e.g. collagens) and proteins dissolved in interstitial fluid. The binding may effectively reduce the flux of FS in tissues. Our observations are consistent with other literature data . Furthermore, FS diffusion through the corneal tissue from inserted MNs was slower than the topically applied drug aqueous solution. This suggests that PVP polymer from dissolved MNs may encapsulate the drug reducing its diffusion from stroma to the endothelium thus permeation into the receptor solution thereby showing a sustained release profile.
For macromolecules such as FD 70 and FD 150, topically applied aqueous solutions showed high resistance in permeation across both the scleral and corneal tissues (Figs. 10 and 11). Though FD 70 showed a high permeation profile through both the cornea and sclera when compared to FD 150. For example, the average percentage permeation of aqueous solutions of FD70 and FD150 was 9 and 5 % across the corneal tissues. In contrast, average percentage permeation of aqueous solutions of FD 70 and FD 150 was 45 and 6 % across the scleral tissues. Concerning the effect of MNs in enhancing the transocular permeation of both macromolecules, our results indicate that dissolving MN arrays have significantly enhanced transocular delivery of both model macromolecules in comparison with aqueous solutions. The percentage drug delivered through the cornea for FD 70 from topically applied aqueous solutions was 9 % which increased to 83 % after 48 h. Likewise, permeation of FD 150 increased from 5 to 61 %. This suggests that dissolving MN arrays played a significant role in enhancing delivery into the corneal stroma and formed a depot, but the endothelium is still forming a strong barrier. Thus, the drug MW still plays its role in dictating transcorneal permeation. However, intrastromal delivery can be an excellent target in the treatment of corneal diseases such as CNV. On the other hand, increasing the drug concentration of FD 70 in the dissolving MN arrays by fivefold (2 to 10 mg/ml) did not significantly increase the percentage permeation. In the sclera, dissolving MNs improved delivery of both small and macromolecule drugs. For the macromolecules, i.e. FD 70 and FD 150, dissolving MNs successfully delivered the total applied doses from both molecules.
PVP polymer is considered to be biologically inert. In this study, we investigated the biocompatibility of PVP polymer with ARPE cells. Cells were exposed to a range of polymer concentrations (at 0.5, 1, 2, 3 and 4 mg/ml) to mimic a scenario of retinal cells-polymer exposure upon delivering the polymer in the scleral tissue. The percentage viability was found to be >83 % at all concentrations below 2 mg/ml (Fig. 12), which concludes that PVP K29/32 MNs is non-toxic in an in vitro testing to the retinal cells, therefore, deemed as biocompatible. However, cell viability was significantly inhibited when cells were treated with 3 (p = 0.031) and 4 mg/ml (p = 0.041) of PVP. This could be primarily due to excess amount of material that could have inhibited normal cell growth.