Improved Magnetic Delivery of Cells to the Trabecular Meshwork in Mice

Abstract

Glaucoma is the leading cause of irreversible blindness worldwide and its most prevalent subtype is primary open angle glaucoma (POAG). One pathological change in POAG is loss of cells in the trabecular meshwork (TM), which is thought to contribute to ocular hypertension and has thus motivated development of cell-based therapies to refunctionalize the TM. TM cell therapy has shown promise in intraocular pressure (IOP) control, but existing cell delivery techniques suffer from poor delivery efficiency. We employed a novel magnetic delivery technique to reduce the unwanted side effects of off-target cell delivery. Mesenchymal stem cells (MSCs) were labeled with superparamagnetic iron oxide nanoparticles (SPIONs) and after intracameral injection were magnetically steered towards the TM using a focused magnetic apparatus (“point magnet”). This technique delivered the cells significantly closer to the TM at higher quantities and with more circumferential uniformity compared to either unlabeled cells or those delivered using a “ring magnet” technique. We conclude that our point magnet cell delivery technique can improve the efficiency of TM cell therapy and in doing so, potentially increase the therapeutic benefits and lower the risk of complications such as tumorigenicity and immunogenicity.

With nearly 80 million cases worldwide, glaucoma is the leading cause of irreversible blindness. The most common subtype of the disease is primary open angle glaucoma (POAG) which is associated with an elevated intraocular pressure (IOP). IOP is governed by the rate of aqueous humor (AH) production and the dynamics of its subsequent drainage through the outflow pathways. A key component of the conventional outflow pathway, the main drainage route for AH, is the trabecular meshwork (TM) which undergoes significant changes in POAG, including a loss of cellularity (Alvarado et al., 1984; Coulon et al., 2022; Gong and Swain, 2016). There has thus been great interest in recellularization and refunctionalization of the TM as a potential long-term treatment for ocular hypertension associated with POAG (Xiong et al., 2021; Yun et al., 2018; Zhu et al., 2017, 2016). To date, such work is at the preclinical stage and has been carried out by intracameral injection of stem cells into various glaucoma models.

Mice are useful model organisms for studying ocular hypertension, yet only a few reports exist on TM cell therapy in mice. Du and colleagues suggested that delivering TM stem cells into Tg-MYOC^Y437H mice led to a reduction in IOP, an increase in outflow facility, improved TM structure, increased TM cellularity and neuroprotection, as compared to the saline-injected (sham) controls (Xiong et al., 2021; Yun et al., 2018). Zhu et al. delivered iPSC-TM cells into the eyes of Tg-MYOC^Y437H mice and reported a marked decrease in IOP and an increase in outflow facility for up to 12 weeks after injection as compared to saline-injected (sham) controls. The treatment also increased TM cellularity through promoting the proliferation of endogenous cells (Zhu et al., 2017, 2016). Similarly, Du and colleagues suggested that delivering TM stem cells into Tg-MYOC^Y437H mice led to a reduction in IOP, an increase in outflow facility, improved TM structure, increased TM cellularity and neuroprotection, as compared to the sham control eyes (Xiong et al., 2021; Yun et al., 2018).

Despite the promise of TM cell therapy, there is much room for improvement, specifically in the quality of cell delivery. All the studies mentioned above rely on hydrodynamic forces generated by the flow of AH to passively carry injected cells, resulting in only 8% of the cells being delivered to the relative proximity of the TM (Wang et al., 2022). Off-target delivery limits the therapeutic potential of the treatment, requiring more cells to be injected which in turn increases the risk of tumorigenicity and immunogenicity (Coulon et al., 2022; Zhou et al., 2020). A further concern arises when considering the cells that do reach the TM: due to the segmental nature of AH outflow in the conventional outflow pathway (Chang et al., 2014), passively delivered cells will be spatially limited to only a part of the TM filtration area. Magnetic cell steering has previously been proposed to overcome these shortcomings (Snider et al., 2018). In mice in particular, Wang et al. labeled iPSC-TMs with magnetic nanoparticles and steered them towards the TM under the forces generated by a ring magnet. While they report a 4-fold increase in the proportion of cells delivered in proximity to TM, still about two-third of the cells were delivered off-target (Wang et al., 2022).

In this study we introduce a new delivery technique that uses a magnetic apparatus with a focused magnetic field (“point magnet”). We compare the performance of this technique with the previously used “no magnet” and “ring magnet” methods discussed above, focusing on the circumferential uniformity of cell delivery and delivery specificity to the TM region.

Human adipose-derived mesenchymal stem cells (MSCs) were purchased commercially (Lonza Bioscience, Walkersville, MD) and maintained in α-MEM supplemented by 10% FBS and 1% penicillin and streptomycin and 2 mM L-glutamine at 37° C and 5% CO₂. Cells were passaged by treating with 0.05% trypsin (25–053-CI, Corning Inc., Corning, NY) and seeding at 5000 cells/cm² in cell culture flasks. MSCs at passages 5 or 6 and 80% confluency were magnetically labeled by incubation overnight with 50 μl of 150 nm amine-coated superparamgnetic iron oxide nanoparticles (SPIONs; SA0150, Ocean NanoTech, San Diego, CA; 5 mg/ml stock solution) in a T-25 culture flask. Successful labeling was confirmed by light microscopy in a preliminary study. After incubation, we trypsinized the cells for 5 minutes, followed by addition of cell culture media and centrifugation at 2100 g for 5 min. The cells were then resuspended in PBS and were labeled with 5 μM carboxyfluorescein succinimidyl ester (CFSE) to allow fluorescent tracking in the eye (65-0850-84, eBioscience, San Diego, CA). The cell solution was then transferred to a 1.5 ml tube (Fig. 1A, left) and a 0.25” cubic N52 neodymium magnet was placed on the side of the tube, which resulted in the formation of a cell pellet adjacent to the magnet within seconds (Fig. 1A, center). To ensure all injected cells were magnetized, any cells not in the pellet were then removed by aspirating the supernatant. The cell pellet was resuspended in PBS to a final concentration of 1k cells/ul, since higher concentrations resulted in cell clumping inside the injection needle and inferior adhesion to the TM after delivery.

Figure 1.

Uniformity and specificity of cell delivery to the TM assessed from en face images. A) Preparatory steps, including MSC labeling with SPIONs (left), isolation of sufficiently magnetized cells with a magnet (center; white arrow indicates the cell pellet), and the magnets used for different delivery methods (right). B-D) Representative en face images of the delivered cells (red) using different methods. “C” marks cornea. E) A representative quadrant (zoomed region marked by orange box in panel D) subdivided both radially and angularly (green and blue meshes). Scale bars show 1 mm. Blue subregions show the region of interest (ROI) at the limbus. F) Representative plots of circumferential cell distributions using different delivery methods. The “normalized fluorescent pixel count” (NFPC) is calculated from the ROI in panel E. Note that each plot belongs to an individual eye. G-H) Metrics of circumferential delivery uniformity for each method, as described in text. A higher Delivery Adequacy (panel G) and a lower Coefficient of Variation (CV; panel H) correspond to a more uniform circumferential delivery of cells. I) Delivery specificity was evaluated by computing a circumferentially-averaged NFPC value for each delivery method. Higher values indicate more cells being delivered to the limbal region. In panels G-I, each point represents a single eye. In bar graphs, data shown as mean ± standard deviation. For details of methods and statistics refer to text. *p < 0.05

Injection needles were fabricated as follows. We pulled glass micropipettes with a pipette puller, then scored, broke, and beveled the micropipette tips on a revolving diamond abrasive plate. The resulting needles had an OD of 100 μm or less, and a bevel angle of 30°. To improve sharpness and ease of penetration into corneal tissue, we rotated the needle on both lateral sides of the beveled surface and continued grinding until a sharp pointed tip was achieved. To avoid cell adhesion to the needle lumen, the needles underwent plasma cleaning and trichlorosilane treatment, followed by coating with 0.02% Pluronic F-127 (P2443, Sigma-Aldrich, St. Louis, MO). The needle was then loaded into an injection assembly (MMP-KIT, WPI, Sarasota, FL) which itself was mounted on a micromanipulator and connected to a microsyringe pump (PHD Ultra, Harvard Apparatus, Holliston, MA).

All animal procedures were conducted following guidelines approved by the Georgia Tech institutional animal care and use committee and consistent with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. Eleven eyes of 7 wildtype C57BL/6 mice (Jackson Laboratory, Bar Harbor, ME) aged 3–5 months were used. For each injection, a tropicamide eyedrop was instilled and anesthesia was induced using isoflurane, after which anesthesia was maintained through a nose cone while the animal was strapped on a heated bed at 37°C. The micropipette was sterilized with 70% ethanol and was loaded with 3 μl of cell-containing solution. Topical anesthetic (tetracaine) was applied to the eye, and the eye was proptosed using a pair of non-magnetic forceps while the needle was advanced into the AC using the micromanipulator. 1.5 μl of cell solution was injected into each eye, with the specifics of delivery varying based on the candidate delivery method, as outlined below.

• Non-magnetic (passive) delivery: the cell solution was injected into the eye over 5 seconds, corresponding to a flow rate of 18 μl/min, as reported in previous studies in which non-magnetic polystyrene microbeads were delivered to the mouse TM (Calkins et al., 2018). We chose this flow rate because the microbeads used in the above-referenced study had a diameter similar to MSCs (~15 μm), and because we could not find a reported flow rate in previous studies that attempted passive cell delivery to the TM. After a 15–20 second wait to minimize AH backflow, the needle was withdrawn, and injection was complete.

• Ring magnet method: The approach was similar to the passive delivery approach, except that after removing the needle a commercial N52 ring magnet (ID = 3 mm, OD = 4 mm and 1mm thickness; R0545–10, SM Magnetics, Pelham, AL; Fig. 1A, right) was placed over the limbus for 15 minutes according to previous studies (Snider et al., 2018).

• Point magnet: The needle was kept in the eye while the pump delivered the cell solution in 4 aliquots of 0.375 μl at 2.4 μl /min. This lower injection flow rate gave the experimenter enough time to steer the cells to the TM in small aliquots while avoiding clumping of the cells, thus yielding a more consistent delivery. The experimenter triggered the injection of each aliquot through a foot pedal and gradually dragged the injected cells towards the TM by placing the tip of the point magnet (consisting of a cubic N52 neodymium magnet connected to a thin stainless-steel rod; Fig. 1A, right) on the limbus and slowly moving the tip along the circumference of one quadrant of the cornea. The same procedure was repeated for the remaining quadrants, and the needle was kept in the eye for an additional 15 minutes, for reasons discussed in detail below.

For all methods, the injected eye received topical antibiotic ointment (neomycin, polymyxin, and bacitracin ophthalmic combination) and the animal was allowed to recover on a heated pad.

The injected animals were euthanized by intraperitoneal injection of sodium pentobarbital 48 hrs after injection, and the eyes were carefully enucleated and immersion fixed in 10% formalin (Fisher Healthcare, Waltham, MA) overnight at 4°C, after which the anterior segments were dissected into 4 leaflets for wholemount fluorescent imaging. To visualize cell distribution around the circumference of the eye, fluorescent en face tile scans of the anterior segment wholemounts were taken using the 20X objective on a Leica DMB6 epifluorescent microscope (Leica Microsystems, Wetzlar, Germany). Next, at least two of the leaflets from each wholemount were cryoprotected by sequential immersion (15 min each) in 15% sucrose (Sigma) in PBS, 30% sucrose, and a 1:1 solution of 30% sucrose and optimal cutting temperature (OCT) media. The specimens were then embedded in OCT and were flash frozen in a 100% ethanol bath cooled by dry ice. A CryoStar NX70 cryostat (ThermoFisher Scientific, Waltham, MA) was used to cut 10 μm-thick sagittal sections. Tile scans of the sagittal sections were taken using the same microscope as above.

We considered two metrics of cell delivery: the uniformity of cells around the entire anterior segment circumference (“uniformity of cell delivery”), and the proximity of cells to the TM (“specificity of cell delivery”), as follows.

Uniformity of cell delivery:

En face images (Fig. 1 B–D) were analyzed using a semi-automated MATLAB v2020 (MathWorks, Natick, MA, USA) algorithm, as follows.

• We manually marked the cut edges of the iris for each quadrant (Fig. 1E).

• We subdivided the marked region into subregions, such that the entire eye (assembly of the four marked regions) was made up of 10 radial “rings” and 21 “wedges” (Fig. 1E, green grid shown for one quadrant). This meant that the number and the size of subregions in each quadrant depended on the size of the marked area in that quadrant compared to the others. For example, if the length of the arc that marked the limbus in quadrant A was half of that of quadrant B, quadrant A was assigned half the number of wedges as quadrant B.

• We picked the two most exterior rings (Fig. 1E, blue grid) to be the approximate location of the limbus and thus our region of interest (ROI). A normalized fluorescent pixel count (NFPC) for each of the subregions in the ROI was calculated by counting the number of fluorescent pixels in the subregion divided by the total number of pixels in the subregion.

• Inside the ROI, each wedge was assigned the maximum NFPC of its two ring subregions, resulting in a single NFPC value for each of the 21 wedges.

• This process was repeated for all the injected eyes (3 no magnet, 4 ring magnet, and 4 point magnet) providing an angular distribution of NFPCs for each eye (Fig. 1F). We then derived two metrics of circumferential uniformity from such plots.

  • Delivery adequacy: Even if cell delivery is not perfectly uniform, all wedges in an eye should ideally receive an “adequate” number of cells. To determine a threshold for this “delivery adequacy” we first calculated the mean NFPC over the 21 wedges of each eye. We found that the eyes in which cells were delivered using the point magnet method had, on average, the largest mean NFPC value (0.20), and we assumed any wedge that had an NFPC value less than 20% of this value to have had inadequate delivery, i.e. threshold NFPC = 0.04. “Delivery adequacy percentage” for each eye was then reported as the ratio of adequately delivered wedges over the total number of wedges.
  • Coefficient of variation (CoV): We calculated the CoV for each eye by dividing the standard deviation of NFPC distribution over all the wedges in an eye by the mean NFPC for that eye.

For the above analyses, we tested the assumption of normality for each group using the Wilk-Shapiro test, and then performed one-way ANOVA followed by Tukey’s multiple comparison to compare various delivery methods.

We found that cell distributions in en face images of anterior segments were less uniform for the no magnet and ring magnet groups vs. the point magnet group (Fig. 1B–D). This qualitative observation was confirmed by the NFPC distributions for each eye (Fig. 1F) and delivery adequacy percentage for each method (Fig. 1G). More than 90% of the wedges received “adequate” cells using the point magnet, whereas less than 10% did for the other methods (point magnet vs. no magnet or ring: p < 10⁻⁵). Statistical comparison between the coefficient of variation for the point magnet method (0.8 ± 0.2) vs. both the ring magnet (3.1 ± 1.5, p = 0.02) and no magnet (3.2 ± 0.3, p = 0.02) methods showed that the point magnet method had significantly less spatial variability (Fig. 1H).

By using the point magnet method, we expected to see a more uniform distribution of cells around the circumference of the limbus compared to both the no magnet case (due to segmental outflow) and the ring magnet method (which required the cells to be injected exactly in the center of the ring to spread evenly – a condition that was almost impossible to achieve experimentally). Consistent with our expectations, the point magnet approach yielded a lower CoV and higher ratio of subregions with adequate cell concentration, indicating a significant improvement in the uniformity of delivery. This observation is contrary to the findings of Snider et al., who previously reported a significant improvement in uniformity using a ring magnet compared to passive delivery in a porcine anterior segment perfusion model (Snider et al., 2018). This might be due to size differences between the mouse and porcine eye, or because the porcine anterior segment preparation used by Snider et al. lacked most of the anterior segment structures that can interfere with cell delivery, such as the iris. The fact that en face images for the ring magnet method show a strong signal in the form of a ring closer to the pupil than the limbus is consistent with the existence of a “blocking” effect by the iris (Fig. 1C).

Specificity of cell delivery (proximity to the TM):

Circumferential uniformity is an important metric, but not the only measure of cell delivery quality. We also evaluated delivery specificity (proximity of cells to the TM), using two approaches. In the first, we simply evaluated the magnitudes of previously calculated mean NFPC values for our three delivery methods (Fig. 1I). In the second, we analyzed images of sagittal sections of the anterior segment for each injected eye (Fig. 2A–C), using a custom image processing algorithm in MATLAB v2020. The algorithm quantified the specificity of delivery by calculating an “off-target index (OTI)” that we defined as:

$$\begin{gathered} \text{OTI}=\sum AB \\ A=\frac{\text{distance}\text{of}\text{pixel}\text{from}\text{TM}\text{measured}\text{along}\text{iris}}{\text{total}\text{length}\text{of}\text{iris}\text{contour}}\hspace{1em}B=\frac{\text{fluorescent}\text{intensity}}{\sum \text{fluorescent}\text{intensity}} \end{gathered}$$

Figure 2.

Specificity of cell delivery to the TM evaluated in sagittal sections. A-C) visualization of fluorescent cells (red) delivered to the AC using different methods. Insets show magnified view of cell-containing regions (red). D) Manual segmentation of the iris (green). E) Calculation of normalized distances by the algorithm. White asterisks show the projection of red fluorescent pixels (cells) onto the nearest point on iris contour. S and E mark the start of the iris contour at the TM and its end at the posterior side of the iris where the ciliary processes emerge, respectively. F) Off-target index (OTI) for different delivery methods. Each dot is one sagittal section, grouped by color for each eye. Lower OTI values mean better delivery specificity of injected cells. *p < 0.05. Scale bars: 300 μm.

In order to calculate A, the boundary of the iris was marked by the user as a continuous contour (Fig. 2D) starting from the TM and ending at the posterior side of the iris at the ciliary processes (Fig. 2E). Next, the projection of each fluorescent pixel, representing labeled cells, onto the closest location on the iris contour was found (Fig. 2D). The length of the contour segment bounded in between this projection point and the TM formed the numerator in quantity A.

The quantity B was the normalized fluorescent intensity for each pixel (note that the denominator in the definition of B is the sum of intensity for all pixels), which accounted for the cell concentration through the depth of the cryosections. The value of OTI, by construction, lay in the range from 0 to 1, with lower OTI values meaning more specific delivery to the TM for that section. We calculated the OTI for at least two quadrants per eye and at least one sagittal section per quadrant in all the injected eyes.

For statistical analysis, we confirmed the assumption of normality for each delivery method using the Shapiro–Wilk test. A linear mixed-effect model was then used to compare OTI values, with the fixed effect being the delivery method (ring magnet, focused magnet, or no magnet) and nested random effects being the eyes and sagittal sections (replicates) within each eye.

We found that the delivery method NFPC, a measure of the concentration of cells in the limbal region, was nearly 20-fold higher for point magnet compared to either of the other methods (Fig. 1I). In a more direct evaluation of delivery specificity, sagittal sections of injected eyes showed very different cell distributions between the various injection methods (Fig. 2A–C). Calculation of OTI showed that the point magnet approach gave a ~5-fold improvement in the specificity of cell delivery: 0.07 ± 0.07 (mean ± SD) vs. ring magnet (0.37 ± 0.13) or vs. no magnet (0.36 ± 0.10) (p < 10⁻¹⁰; Fig. 2F). We also observed a smaller OTI standard deviation in point-magnet steered eyes, indicating improved delivery repeatability using the point magnet.

The specificity of cell delivery to the TM has important implications for safety of TM cell therapy, since, depending on the delivered cell type, off-target events can cause immunogenicity and tumorigenicity. Improved specificity can also mean more therapeutic benefit from fewer injected cells and potentially better cell retention after delivery, since delivered cells will reside in the correct niche. The improved specificity of cell delivery to the TM region using the point magnet was judged by several outcomes. First, histological assessment of en face images of the ACs containing fluorescently labeled MSCs showed a clear-cut ring at the limbus with little observable signal outside this ring. In line with these observations, we discovered that the point magnet gave a nearly 20-fold higher cell concentration in the limbal region vs. other methods, with delivered cells positioned 5 times closer to the TM, as indicated by the NFPC and OTI values, respectively. As mentioned earlier, Wang et al. reported an increase in the percentage of cells delivered to the “anterior chamber angle” using a ring magnet vs. no magnet (Wang et al., 2022), while we saw no difference between these two methods using any of our cell delivery specificity parameters. However, their study did not specify important injection parameters (e.g. flow rate) or details of image processing and quantification, particularly the exact selection criteria for “anterior chamber angle” and its boundaries, which preclude a direct comparison with our results.

Our study has some limitations. In quantifying delivery specificity, we used sagittal images of the AC in which the lens had been removed, so that potential off-target delivery to the lens was not included in the calculations of specificity. We did, however, inspect the lenses of dissected eyes and observed minimal fluorescent signal (data not shown). Additionally, we report normalized parameters in all our results which, while useful for removing the effect of biological variations (such as iris length), dissection artifacts, cell fluorescent labeling efficiency variations etc., does not give much information about the total number of cells delivered using each method. One of the major sources of difference in the number of delivered cells may be the back-flush of cells through the corneal puncture site at the time of needle removal; such cell loss can undesirably reduce the therapeutic benefits of the treatment. Even though we have not quantified the total number of delivered cells, the point magnet method allowed for an extended cell incubation time inside the anterior chamber before removing the needle and thus is expected to experience the least cell loss of all three methods. Of course, the different needle retention times in the eye (15 minutes for the point magnet vs. < 1m for the other two methods) could itself be considered a limitation of our work due to its impact on cell retention inside the anterior chamber. However, this difference was an unavoidable aspect of the different delivery methods: in the no magnet case, we attempted to keep the cannula in the eye for an extended duration, but most of the cells fell onto the lens (results not shown) and so this approach was abandoned in favor of the approach developed by previous studies as described earlier. In the ring magnet case, attempting the injection with the magnet in place was extremely challenging due to limited maneuvering space on the cornea as well clumping of the cells inside the needle due to magnetic attraction. Thus, we were forced to withdraw the needle before placing the ring magnet over the eye.

In summary, we have established a protocol for a new magnetic TM cell delivery method that is potentially safer, more effective, and more repeatable than previously reported methods. It is noteworthy that the point magnet delivery approach is relatively easy to carry out and cell placement is under direct control of the operator, suggesting potential for future translatability.