Microfluidic Platforms Promote Polarization of Human-Derived Retinal Ganglion Cells That Model Axonopathy

Abstract

Purpose: Axons depend on long-range transport of proteins and organelles which increases susceptibility to metabolic stress in disease. The axon initial segment (AIS) is particularly vulnerable due to the high bioenergetic demand of action potential generation. Here, we prepared retinal ganglion cells derived from human embryonic stem cells (hRGCs) to probe how axonal stress alters AIS morphology.

Methods: hRGCs were cultured on coverslips or microfluidic platforms. We assayed AIS specification and morphology by immunolabeling against ankyrin G (ankG), an axon-specific protein, and postsynaptic density 95 (PSD-95), a dendrite-specific protein. Using microfluidic platforms that enable fluidic isolation, we added colchicine to the axon compartment to lesion axons. We verified axonopathy by measuring the anterograde axon transport of cholera toxin subunit B and immunolabeling against cleaved caspase 3 (CC3) and phosphorylated neurofilament H (SMI-34). We determined the influence of axon injury on AIS morphology by immunolabeling samples against ankG and measuring AIS distance from soma and length.

Results: Based on measurements of ankG and PSD-95 immunolabeling, microfluidic platforms promote the formation and separation of distinct somatic-dendritic versus axonal compartments in hRGCs compared to coverslip cultures. Chemical lesioning of axons by colchicine reduced hRGC anterograde axon transport, increased varicosity density, and enhanced expression of CC3 and SMI-34. Interestingly, we found that colchicine selectively affected hRGCs with axon-carrying dendrites by reducing AIS distance from somas and increasing length, thus suggesting reduced capacity to maintain excitability.

Conclusions: Thus, microfluidic platforms promote polarized hRGCs that enable modeling of axonopathy.

Translational relevance: Microfluidic platforms may be used to assay compartmentalized degeneration that occurs during glaucoma.

Figure 1.

Mouse- and human-derived RGCs exhibit heterogeneity in AIS localization. (A, B) Representative micrographs of whole-mount retinas (WMRs) from B6.Cg-Tg(Thy1-YFP)16Jrs/J mice labeled for ankG demonstrate two distinct patterns of RGC AIS localization. (A) The majority of RGCs had a single ankG-labeled AIS (cyan) located on a process emanating directly from the YFP-labeled soma (yellow, with white dotted outline). (B) A small subset of mouse RGCs had an AIS located on a process distal to a bifurcation in a primary dendrite (white arrowhead), defined as an AcD. (C–H) Example micrographs of tdTomato (tdTom, orange) expressing RGCs derived from hRGCs cultured on coverslips demonstrate greater heterogeneity in ankG-labeled (green) AIS localization. As seen in mouse retina, hRGCs had a single AIS localized on a process emanating directly from the soma (direct, C; diagrammed in D) as well as on an AcD (E; diagrammed in F). There was also a subset of hRGCs that contained multiple AISs (multi, G; diagrammed in H).

Figure 2.

hRGCs plated on microfluidic platforms are polarized into somato–dendritic and axonal compartments. (A) Schematic of Xona XC450 microfluidic platform (not illustrated to scale). (B) After plating on the XC450 for 8 to 10 days, hRGC projections extended through the microgroove chamber and into the axon chamber. (C) When hRGCs were cultured in microfluidic platforms, ankG (green) typically localized to an axon extending directly from the soma (direct) or on an AcD within the soma chamber (D). Insets in C and D show ankG channel only. Scale bars: 10 µm. (E) ankG did not accumulate within putative axons in the axon chamber. (F, G) Representative micrographs of PSD-95–labeled (cyan) hRGCs cultured in microfluidic devices demonstrate evidence of postsynaptic specification in the soma chamber (F) and localization to putative axon growth cones in the axon chamber (G). Arrows indicate PSD-95 colocalized with tdTom-positive putative axon growth cones. (H) Quantification of PSD-95 immunofluorescence suggests significantly greater expression in the soma chamber than in the axon chamber (P = 0.0281, unpaired t-test). Integrated (Integ) density is the summed pixel values multiplied by area. (I) The soma chamber exhibited significantly greater colocalization of PSD-95 labeling and tdTom expression than did the axon chamber (P = 0.0003, unpaired t-test) (n = 4 independent devices). Error bars: ±SEM. *P < 0.05, ***P < 0.001.

Figure 3.

Systematic quantification of AIS length and distance from soma. (A, B) To determine AIS distance from soma and length, intensity profiles of ankG immunofluorescence were determined from hand traces (shown as dashed line) extending from the soma edge along the ankG-containing process for direct (A) and AcD (B) morphologies. The ankG is illustrated as a green span along the process. (C–E) Representative ankG intensity profiles from mouse RGCs (C), coverslip hRGCs (D), and hRGCs plated on microfluidic platforms (E). From raw traces (light gray lines), background fluorescence was subtracted, ankG intensity profiles were smoothed (black lines), and we defined the AIS length as the extent where the smoothed ankG intensity was greater than 50% of the difference between baseline and maximum intensity (green shaded region). au, arbitrary units; Int, mean intensity.

Figure 4.

Microfluidic platforms promote normalization of AIS morphology in hRGCs. (A) Cumulative percentages of AIS localizations observed in mouse RGCs, hRGCs plated on coverslips, and hRGCs cultured in microfluidic platforms. (B) The AIS distance from the soma is significantly longer for AcD than direct AISs (P < 0.0001), but largely the same across cell types/culture platforms (P = 0.6999). (C) hRGCs cultured on coverslips possessed direct (P = 0.0041) and AcD (P = 0.0031) AISs significantly longer than mouse RGC direct and AcD AISs, respectively. The hRGCs plated on microfluidic platforms contained AcD AISs significantly shorter than hRGCs cultured on coverslips (P = 0.0182). The hRGCs plated on microfluidics possessed AISs of similar length to mouse RGCs (P ≥ 0.083). (D) Distribution of AIS distance versus length scatterplots for each of the cell types/culture platforms as determined by kernel density estimates. Coverslip hRGC (orange) AIS dimensions appear to have a more variable distribution than microfluidic hRGCs (blue) or mouse RGCs (gray). Sample sizes (excluding multi): mouse RGC group, 135 cells, six retinas (only two retinas contained AcD); coverslip hRGC group, 109 cells from three independent samples; microfluidic hRGC group, 190 cells, seven independent devices. Statistics: two-way ANOVA, Tukey post hoc test (B, C); kernel density estimate (D). Error bars: ±SEM. *P < 0.05, **P < 0.01.

Figure 5.

Colchicine application models axonopathy in hRGCs cultured on microfluidic platforms. (A) Epifluorescence micrograph of live tdTomato-positive (black) hRGCs after 8 DIV, demonstrating extension of putative axons into the axon chamber prior to application of colchicine. (B–D) Representative images of tdTomato-positive hRGCs (orange) cultured in microfluidic devices after the addition of either vehicle (B) or colchicine (30 nM, 3 days) to the axon chamber (C, D). Colchicine caused axon retraction in some samples (C) and outright axon degeneration in other samples (D). (D, right) Example degenerative axon located within the microgroove section of the microfluidic platform after colchicine treatment. Fluorescently conjugated CTB (cyan) added to the soma chamber (left side of images) was transported anterogradely to distal putative axons in the axon chamber. Colchicine appeared to reduce the transport of CTB to spared axons in the axon chamber (C–E). (E) Higher-magnification example images of the axon chambers in vehicle- and colchicine-treated cultures. CTB fluorescence (cyan) is evident in the vehicle samples but was reduced after colchicine treatment. In colchicine-treated cultures, spared axons exhibited degenerative varicosities (E, right, white arrowheads). (F, G) After colchicine treatment, CTB fluorescence in the axon chamber was significantly reduced (P = 0.0159), and varicosity density significantly increased (P = 0.0095). The vehicle group had four independent samples, and the colchicine group had five or six independent samples. (H) Example confocal images of hRGCs (orange) after vehicle (left) or colchicine (right) treatment and immunolabeled against SMI-34 (green) and CC3 (blue). Scale bars: 10 µm (E, H). (I) We found significant positive correlations between CC3 and SMI-34 for both vehicle-treated cells (R² = 0.80, P < 0.001) and colchicine-treated cells (R² = 0.72, P < 0.001). CC3 and SMI-34 integrated density was normalized by tdTomato fluorescence integrated density. The vehicle group included 123 cells from two devices, and the colchicine group included 214 cells from four devices. Statistics: Mann–Whitney test (F, G), linear regression (I). Error bars: ±SEM. *P < 0.05, **P < 0.01.

Figure 6.

AcD axon initial segments are preferentially susceptible to colchicine-induced changes in morphology. (A, B) Example images of ankG labeling (green) in hRGCs (orange) treated with vehicle (A) or colchicine (B). (C) Cumulative percentages of AIS localizations for both treatment groups. No multi-AIS cells were seen in the colchicine group, although there was not a significant difference in the distribution of localizations between groups (P = 0.104). (D) Colchicine treatment significantly decreased AIS distance from the soma (P = 0.0063), with a greater decrease seen in AcD AISs (P = 0.0179) than in direct AISs (P = 0.3041). (E) Colchicine treatment did not significantly affect AIS length (P = 0.3066). The vehicle group included 190 cells from seven independent devices, and the colchicine group included 76 cells from four independent devices. Statistics: χ² test (C), two-way ANOVA with Šídák's multiple comparisons test (D, E). Error bars: ±SEM. *P < 0.05.