Directed Growth of Adult Human White Matter Stem Cell–Derived Neurons on Aligned Fibrillar Collagen

Cell culture

Adult human subcortical white matter tissue was obtained from routine epilepsy surgery (anterior temporal lobectomy or resection of focal cortical dysplasias; six samples; individuals 28–49 years old; five women and one man) following informed consent of the patients. All procedures were in accordance with the Helsinki convention and approved by the ethical committee of the Dresden University of Technology (EK 47032006). Specimens were taken from the removed tissue and stored in ice-cold Hank's balanced salt solution supplemented with 11 mM glucose and 1% penicillin/streptomycin while being transported from the operating theater. All patients underwent high-resolution magnetic resonance imaging to exclude tumors and were screened for the presence of infectious disease. In all cases, tissue and neuropathological examination did not reveal evidence of tumor formation. For expansion of neurospheres consisting of hNSCs, white matter was isolated from tissue samples and then cut into small pieces with a scalpel, incubated in 0.1% trypsin (Sigma Aldrich, St. Louis, MO) for 30 min at room temperature, incubated in DNase (40 mg/mL; Sigma, Steinheim, Germany) for 10 min at 37°C, and homogenized to a quasi single-cell suspension by gentle triturating.^30,31,34 The cells were added to 25-cm² flasks (2–3 × 10⁶ viable cells per flask) in knockout Dulbecco's modified Eagle medium (DMEM) (Gibco BRL–Life Technologies, Tulsa, OK), supplemented with 10% serum replacement (Gibco), 0.5 mM glutamine, 1% penicillin/streptomycin, and 20 ng/mL of both epidermal growth factor (EGF) and fibroblast growth factor-2 (both from Sigma) at 5% CO₂, 92% N₂, and 3% O₂ using an incubator equipped with an O₂-sensitive electrode system (Heraeus, Hanau, Germany). After 10–20 days, neurosphere formation was observed, and these spheres were expanded for additional 5–8 weeks (in total, 7–12 weeks) before differentiation was initiated. The medium was changed once a week and the growth factors were added twice a week.

Differentiation conditions

Induction of neural differentiation was initiated by plating the cells on poly-d-lysine (PDL; 5 mg/100 mL Aqua Dest; Sigma)-coated glass coverslips in knockout DMEM, supplemented with 10% serum replacement, 0.5 mM glutamine, 1% penicillin/streptomycin, 10 ng/mL rh-BDNF (Promega, Madison, WI), and 100 μM di-butyryl-(db) cAMP (Sigma). Three days later, 5% fetal calf serum (Gibco) was added. The cells were differentiated for 7 days. Identical conditions were used to plate cells on various collagen matrices.

Preparation of glass substrates

For experiments with glass substrates coated with PDL and poly(octadecene-alt-maleic acid) (POMA; Polysciences, Warrington, PA), freshly cleaned 22 × 22 mm glass coverslips (Menzel Gläser, Braunschweig, Germany) were oxidized in a 1:1 mixture of aqueous ammonia solution (Acros Organics, Geel, Belgium) and hydrogen peroxide (Merck). PDL solution was pipetted onto the glass substrate, kept at room temperature for 30 min, and rinsed with phosphate-buffered saline (PBS). For POMA coating, coverslips were functionalized by reaction with 3-aminopropyl-dimethylethoxy-silane (ABCR, Karlsruhe, Germany). POMA was dissolved in tetrahydrofuran (Fluka, Deisenhofen, Germany) in a concentration of 0.08 wt% and spin coated (RC 5; Suess Microtec, Garching, Germany) onto the amine-modified coverslips at 4000 rpm for 30 s. Stable covalent binding of the polymer films was achieved by annealing at 120°C to form imide bonds with the amino-silane on the glass substrate.

Microfluidic system and flow conditions

Aligned collagen matrices were prepared by a novel approach using a microfluidic setup (for further information about the microfluidic system, please refer to Refs.^29,35). The flow conditions were performed under conditions of laminar flow. In all experiments, the Reynolds number (as a criterion for laminar flow) was smaller than 1, and a flow rate of 11 μL/min corresponding to a wall shear rate of 203.1 s⁻¹ was applied.

Collagen solution

Bovine dermal collagen I solution (purified and pepsin-solubilized in 0.012 N HCl; PureCol; Inamed, Fremont, CA) was brought to physiological pH by mixing eight parts of acidic collagen solution (3.0 mg/mL) with one part of 10-fold concentrated PBS (Sigma) and one part 0.1 M NaOH. All components were kept in an ice bath before and after mixing. Appropriate volumes of chilled 1× PBS were added to adjust the final concentration of the collagen solution.

Nonaligned collagen matrices

POMA-coated glass slides were coated by contacting the substrates to the collagen solution and then subsequently raising the temperature. ³⁶ To precisely apply collagen coating in designated areas, a silicone-based culture insert (Ibidi, Martinsried, Germany) consisting of two wells was placed onto a POMA-coated glass slide. A 100 μL drop of chilled nonfibrillar collagen solution (0.4 mg/mL) was pipetted into each well and kept in a CO₂-free incubator at 37°C to initiate fibril formation. After 90 min the resulting fibrillar collagen gel and the culture insert was removed from the surface, leaving two 0.22-cm² rectangular areas covered with a thin collagen layer on the substrate. The modified surface was rinsed with PBS for at least three times and subsequently with Milli-Q water (Millipore, Molsheim, France).

Aligned collagen matrices

To produce collagen matrices containing long fibrils at a low density, a low-concentration collagen solution (3 μg/mL) was used as described previously. ²⁹ Briefly, collagen solution (0.8 mg/mL) was prepared at 4°C in a centrifuge tube (Roth, Karlsruhe, Germany), placed in a CO₂-free incubator (Sanyo, Bensenville, IL) at 37°C, and allowed to form fibrils for 24 h. The resulting gel was homogenized for 4 min (T-8 Ultra Turrax Tube Drive from IRA, Staufen, Germany) and centrifuged (Heraeus) at 1000 g for 6 min. The supernatant (2 mL, 3 μg/mL) was drawn into a syringe that was subsequently installed into the syringe pump. The solution was then pumped for 1 h with a flow rate of 11 μL/min through the microfluidic channel. Afterward, the channel was rinsed with PBS for 20 min to flush away loose fibrils, and the coated coverslip was removed from the poly(dimethylsilotane) (PDMS) channel plate and kept in PBS for cell culture. Aligned collagen matrices containing short fibrils at low density were prepared with a solution concentration of 0.2 mg/mL. ²⁹ Briefly, the microfluidic system was placed in a CO₂-free incubator (Sanyo) at 37°C. The collagen solution was prepared at 4°C and drawn into a chilled syringe (Roth), which was quickly installed into the syringe pump to avoid early fibrillogenesis. The streaming process was started immediately by pumping the chilled collagen solution from the cooled storage reservoir through heated tubing, followed by streaming through the microfluidic channel for 1 h. The sample was then rinsed with PBS as described earlier. Aligned collagen matrices containing long fibrils at high density were prepared according to the same protocol for collagen matrices containing short fibrils at a low density using a solution concentration of 0.8 mg/mL. Samples were sterilized under UV light for 5 min.

Confocal reflection microscopy

Confocal reflection microscopy (CRM; TCS SP; Leica, Bensheim, Germany) analysis was performed at a wavelength of 488 nm (Ar laser) to visualize unstained collagen fibrils as described elsewhere. ³⁷ CRM pictures of fibril matrices were obtained using a 40× oil-immersion objective.

Atomic force microscopy

Surface topography of air-dried collagen matrices was investigated via intermittent contact scanning force microscopy with a PicoSPM (Molecular Imaging, Phoenix, AZ) using silicon cantilevers (Tap300; BudgetSensors, Sofia, Bulgaria) with a 300 ± 100 kHz resonant frequency, 40 N/m force constant, and tip radius <10 nm.

Scanning electron microscopy

Samples for scanning electron microscopy (SEM) analysis were fixed with 2% glutaraldehyde, dehydrated, and critical point dried (CPD 030; Baltec, Schalksmühle, Germany). The samples were gold coated with a sputter coater (SCD 050; Baltec) to allow sample analysis under high vacuum by SEM (XL 30 ESEM FEG; FEI-Phillips, Eindhoven, The Netherlands) equipped with a scanning electron detector and operating at 7.5–10 kV, Spot Size 3. Micrographs of hNSCs on the area covered with aligned collagen fibrils or the corresponding area (5 mm × 1 mm) on the controls were taken. For each condition (long or short fibrils at high or low density), three independent donors were analyzed.

Image analysis to quantify collagen fibril alignment

The quantification of fibril alignment was performed as described previously. ²⁹ Briefly, CRM images of three samples containing two channels with collagen fibrils were analyzed for each condition and quantified using the NIH ImageJ 1.37v software. One sample contained a minimum of 200 collagen fibrils. A threshold value of the reflection intensity was defined to isolate fibrils from background. By fitting an ellipse to the major axis of each collagen fibril, the angle of the fibrils to the flow direction was determined. The orientation of fibrils parallel to the alignment direction corresponded to an angle of 0°. The angular distribution of collagen fibrils was determined based on the relative frequency of orientation angles (classified into bins of 10° angles) and by a fit to a Gaussian full width at half maximum (FWHM).

Image analysis to quantify hNSC extension alignment and length

SEM images were analyzed to quantify the extent of alignment and length of cell extensions. The quantification of extension alignment was performed as described for quantification of collagen fibril alignment. The angular distribution of extensions was determined based on the relative frequency of extension angles (classified into bins of 10° angles) of three donors. Only cells/extensions on the area covered with aligned collagen fibrils or the corresponding area (5 mm × 1 mm) on the controls were measured. The length of each extension extending from neurospheres was measured using NIH ImageJ 1.37v (plugin NeuronJ) software. Each extension was measured from its origin at the edge of the neurosphere to its end. Changes in direction along its course were taken into account by tracing along the entire length of each extension. Extensions that were solitary and clearly isolated were measured only to exclude the possibility of mix-up with other extensions.

Immunocytochemistry

Cell cultures were fixed with Accustain (Sigma). Immunocytochemistry was carried out using standard protocols described previously.^30,38 Cell nuclei were counterstained with 4,6-diamidino-2-phenylindole (DAPI). The following primary antibodies were used: mouse anti-GalC 1:750 (Chemicon International, Temecula, CA), rabbit anti-Tuj1 1:2000, mouse anti-Tuj1 1:500 (both from Covance, Richmond, CA), chicken anti-glial fibrillary acidic protein (GFAP) 1:1000 (Abcam, Cambridge, UK), rabbit antiactivated caspase3 1 μg/mL (BD Pharmingen, San Diego, CA), and secondary antibodies conjugated to Alexa 488, 568, or 647 1:500 (all from Invitrogen–Molecular Probes, Carlsbad, CA). Images were captured using a fluorescence microscope (Leica DM IRE2; Leica Microsystems, Wetzlar, Germany). Living cells were marked for analysis using fluorescein diacetate (FDA) and dead cells were stained with propidium iodide (PI; Sigma Aldrich). A solution of 100 μg/mL FDA was prepared by dissolving in acetone. A working stock solution containing 100 ng/mL of FDA was prepared in PBS and was added directly to the cell cultures (i.e., no change of culture medium) at a 1:1 volume ratio, and PI was added at a concentration of 1:50 (1 μg/mL). Samples were incubated for ∼5 min and visualized using epifluorescent microscopy (Leica DM IRE2; Leica Microsystems) with excitation filters at 490/20 nm and emission at 527/30 nm for fluorescein. As neither PI nor activated caspase3 staining worked on the hNSCs on the substrates, dead cells were quantified by counting condensed DNA using DAPI staining correlated to all nuclei.

Cell counting and statistics

To quantify the percentage of cells producing a given marker in a given experiment, the number of positive cells on the coated well surface was determined relative to the total number of DAPI-labeled nuclei. In a typical experiment, a total of 500–1000 cells were counted per marker. Statistical comparisons were made by ANOVA with post hoc t-test and Bonferroni adjustment or Dunnett's t-test as appropriate. All data are presented as mean values ± standard error of the mean. All quantifications were done by an independent blinded investigator.

Alignment of collagen fibrils

As part of neuronal tissue regeneration in both central and peripheral nervous tissue, targeted axodendritic outgrowth is a major challenge for reconstructing neuronal tracts and circuitries. We thus tested the ability of aligned collagen fibrils generated using a microfluidic system ²⁹ to guide neurites during the differentiation process of adult white matter hNSCs (see Fig. 1 for schematic representation of the technology). Collagen fibrils were deposited on planar POMA-coated substrates from collagen solutions streaming through a microfluidic channel. By varying collagen solution concentrations and gelation times, fibril length and density were tuned and varied. In brief, lower concentrations generated shorter fibrils at lower density, whereas higher concentrations resulted in aligned matrices containing long fibrils at high density. Atomic force microscopy was used to observe the resultant aligned collagen matrices: long fibrils at high density (Fig. 1A), long fibrils at low density (Fig. 1B), and short collagen fibrils at low density (Fig. 1C).

OPEN IN VIEWER

FIG. 1.

Scheme of the approach to prepare collagen type I fibril matrices that are adjustable in orientation, density, and morphology by varying collagen solution concentration, gelling time, shear rate, and substrate. Atomic force microscopy images show the resultant aligned collagen matrices with long fibrils at high density (A), long fibrils at low density (B), and short fibrils at low density (C). Scale bar = 2 μm. Color images available online at www.liebertonline.com/ten.

Influence of aligned collagen with varying structural characteristics on extension alignment

To investigate the significance of the newly developed alignment approach for biomedical application, we attempted to prove its potential in a human nerve cell system. As human neurons are not routinely available without harming the patient or the cells themselves, we used neurons differentiated from adult hNSCs. Adult hNSCs were isolated from cortical white matter tissue obtained from neurosurgery interventions as recently described.^30,32 Triple immunofluorescence staining of neuroectodermal markers confirmed typical trilineage neuroectodermal differentiation capacity of the adult hNSCs with 30% ± 5% Tuj1⁺ neurons, 10% ± 4% GFAP⁺ astrocytes, and 4% ± 2% GalC⁺ oligodendrocytes after 7 days of differentiation on the standard differentiation surface PDL (data from three independent experiments from three different donors). Adult hNSCs are thus a valuable human cell model for investigating axodendritic outgrowth of human Tuj1⁺ neurons. We used cells from at least three different donors for all further experiments.

To assess the effect of aligned collagen matrices with varying structural properties on neurite outgrowth and extension alignment (Fig. 1), hNSCs were differentiated for 7 days on rectangular areas (1 mm × 6.5 mm) with aligned collagen fibrils (long fibrils at high density, Fig. 2A; long fibrils at low density, Fig. 2B; short fibrils at low density, Fig. 2C) deposited on a POMA-coated glass surface. Nonaligned collagen resulting from static fibril formation and immobilization on a POMA-coated surface served as the control surface (Fig. 2D). SEM and immunocytochemistry were used to analyze hNSC morphology. Fibril and cell alignment was quantified by determining the angular distribution of collagen fibrils and cell extensions. Based on this analysis, the orientation angle distribution was plotted as a histogram, and the corresponding FWHM was determined by a Gaussian fit (see also Lanfer et al. ²⁹ ). Fibril alignment was highest on matrices containing long fibrils at low density (FWHM, 18° ± 1°; Fig. 2B, F), followed by long fibrils at high density (FWHM, 30° ± 3°; Fig. 2A, E), and short fibrils at low density (FWHM, 41° ± 5°; Fig. 2C, G). Nonaligned collagen showed random distribution of fiber angles (Fig. 2D, H). On all substrates, differentiated hNSCs displayed typical unipolar, bipolar, and multipolar neuronal morphologies with axodendritic outgrowth (Fig. 2I–L). We measured the orientation of a minimum of 166 extensions per independent experiment. On substrates containing aligned fibrils (Fig. 2I–K), aligned extension outgrowth in the direction of the underlying collagen fibrils was observed, whereas extensions on nonaligned collagen exhibited no preferred orientation (Fig. 2L). In contrast to fibril alignment degree, extension alignment was highest on long collagen fibrils at high density (FWHM, 38° ± 3°; Fig. 2I, M), then successively decreased on long fibrils at low density (FWHM, 50° ± 19°; Fig. 2J, N) and short fibrils at low density (FWHM, 53° ± 11°; Fig. 2K, O). No preferred extension alignment direction was observed on nonaligned collagen (Fig. 2L, P). These findings emphasize the importance of both fibril alignment and fibril density for the orientation of axodendritic outgrowth of neurons derived from hNSCs. hNSC differentiation was confirmed using immunofluorescence staining for the neuronal marker Tuj1 and the astroglial marker GFAP. On all substrates, hNSCs acquired morphologic and phenotypic characteristics of Tuj1⁺ neurons and GFAP⁺ astrocytes (Fig. 2Q–T).

OPEN IN VIEWER

FIG. 2.

Influence of fibril alignment, length, and density on extension alignment of adult white matter-derived hNSCs. (A–D) Confocal reflection microscopy images of aligned collagen matrices with long fibrils at high density (A), long fibrils at low density (B), short fibrils at low density (C), and randomly oriented fibrils (D). Scale bar = 10 μm. (E–H) Histograms representing the relative frequency of fibril orientation angles with respect to the horizontal axis corresponding to the images shown in (A)–(D). (I–L) SEM images of hNSCs grown for 7 days on matrices of identical alignment and morphology as depicted in (A)–(D). Scale bar = 100 μm. (M–P) Histograms representing the relative frequency of extension orientation angles with respect to the horizontal axis corresponding to the images shown in (I)–(L). (Q–T) Microphotographs of immunofluorescence staining of differentiated hNSCs showing Tuj1⁺ neurons (green) and glial fibrillary acidic protein (GFAP)⁺ astrocytes (blue) grown for 7 days on matrices of identical alignment and morphology as depicted in (A)–(D). Scale bar = 100 μm. hNSCs, human neural stem cells; SEM, scanning electron microscopy; GFAP, glial fibrillary acidic protein. Color images available online at www.liebertonline.com/ten.

Directional outgrowth of hNSC extensions on aligned collagen matrices

Quantification of cell alignment revealed that the highest extension alignment occurs on matrices containing long collagen fibrils at high density. More detailed SEM studies were performed on these substrates to investigate extension morphology and cell–matrix interactions. Extensions on aligned collagen matrices containing long fibrils at high density were highly oriented, and extension alignment was consistent along the whole width and length of the collagen-coated area (Fig. 3A–C, E, dotted lines mark the collagen matrix border). In contrast, on the adjacent POMA-coated glass surface, extensions showed no preferred direction of orientation, emanating in radial fashion (Fig. 3A, D). Some neurospheres adhered at the border of the aligned collagen layer, extending processes on both the collagen surface and the adjacent POMA-coated glass. While extensions in contact with the POMA surface grew without any bias in alignment, those on the side with aligned collagen followed the direction of the underlying fibrils (Fig. 3E). Once in contact with the fibrils, the processes extended parallel to the fibrils, always in the direction of the aligned collagen (Fig. 3F, including inset). Immunocytochemistry staining demonstrated that the alignment cues from collagen fibrils acted almost exclusively on Tuj1⁺ neurites, while GFAP⁺ astroglial extensions were rather uninfluenced by collagen fibril alignment (Fig. 3G–J).

OPEN IN VIEWER

FIG. 3.

Aligned collagen fibrils guide and direct extensions of adult white matter-derived hNSCs. (A) SEM image displays differentiated hNSCs adhering on top of aligned collagen fibrils (long fibrils at high density) and the surrounding POMA surface (dotted lines mark the border between the aligned collagen matrix areas and POMA-coated areas; double-headed arrows show collagen fibril orientation) after 7 days in culture. Differentiated hNSCs adopt characteristics of glial and neuronal cells with axodendritic outgrowth. (B) Atomic force microscopy micrograph of aligned collagen with long fibrils at high density. (C–F) Higher magnification SEM photographs of hNSCs growing on aligned collagen fibrils (C, F), POMA-coated surface (D), or both surfaces (E) show that extension outgrowth was highly aligned and guided by the underlying collagen fibrils. On the adjacent POMA surface, extensions grew in radial fashion without a preferred direction of alignment. (F) High-resolution image of single extensions demonstrating that neurites in contact with single collagen fibrils extended their processes parallel to the fibrils (inset). (G–J) Immunocytochemistry staining of Tuj1, a neuronal marker, and GFAP, an astroglial marker, of differentiated hNSCs on aligned collagen fibrils (G–I) and on POMA-coated surfaces (J). The staining clearly shows orientation along the collagen fibrils for most Tuj1⁺ neurites (arrowheads), but no significant orientation of GFAP⁺ glial cell processes (arrows). On POMA, no oriented extension growth could be observed. Note the GFAP⁺ flat cells on POMA-coated surfaces representing astroglial cells (A, D, J). Scale bars = 200 μm (A, G), 80 μm (C–E), 50 μm (F), and 100 μm (H–J). POMA, poly(octadecene-alt-maleic acid). Color images available online at www.liebertonline.com/ten.

Influence of collagen fibril alignment on hNSC behavior

To further characterize the synergistic influence of topographical and biochemical cues from aligned collagen on hNSC behavior, we compared neuroectodermal differentiation on substrates with the highest cell alignment (long fibrils at high density) with nonaligned collagen and the standard differentiation surface PDL: Trilineage neural differentiation capacity was equivalent on nonaligned collagen and aligned collagen (long fibrils at high density), and no significant differences were observed when compared with PDL (p = 0.068, F-value = 4.06 for Tuj1⁺ cells; p = 0.90, F-value = 0.11 for GFAP⁺ cells; p = 0.37, F-value = 1.15 for GalC⁺ cells; ANOVA; n = 3; Fig. 4A, C). Interestingly, on POMA-coated surfaces, we found additional flat GFAP⁺ astrocytes that were not detectable by immunocytochemistry on PDL, nonaligned, or aligned collagen fibrils (Figs. 2 and 3). The influence of aligned structures on cell survival was quantified by counting condensed and inhomogeneous DNA using DAPI staining correlated to all nuclei. There were no differences in cell survival, with 5% ± 2%, 5% ± 2%, and 6% ± 1% pycnotic cell nuclei on PDL, nonaligned, and aligned collagen fibrils, respectively (p = 0.85, F-value = 0.17; ANOVA; n = 3; Fig. 4B, D). Thus both hNSC differentiation and survival showed no difference with respect to growth surface. This contrasts to the highly specific influence of aligned collagen type I on the direction of neurite outgrowth.

OPEN IN VIEWER

FIG. 4.

Influence of aligned collagen matrices on hNSC differentiation and survival. (A, C) Representative microphotographs of quadruple immunofluorescence staining of differentiated hNSCs showing Tuj1⁺ neurons, GFAP⁺ astrocytes, and GalC⁺ oligodendrocytes on PDL, nonaligned, and aligned collagen type I (long fibrils at high density). Cells were allowed to differentiate for 7 days. Note that Tuj1⁺ neurite outgrowth appeared completely randomly and radially on PDL and nonaligned collagen. Nuclei were counterstained with DAPI. Scale bar = 100 μm. (C) Quantitative measurement of cell fate decision of hNSCs on PDL, nonaligned, and aligned (long fibrils at high density) collagen type I by quadruple immunofluorescence staining of Tuj1, GFAP, and GalC in differentiated hNSCs. Nuclei were counterstained with DAPI. (B, D) Cell survival of hNSCs measured by counting condensed nuclei in DAPI staining was unaffected by different substrates (PDL, nonaligned, and aligned [long fibrils at high density] collagen type I) after 7 days of differentiation. Magnified images are shown in insets. Arrowheads indicate fragmented cell nuclei using DAPI staining. Scale bars = 200 and 25 μm (insets). Statistical analyses using ANOVA revealed no significant difference in cell types with respect to seeding surfaces (data presented are means ± standard error of the mean of at least three independent experiments from three donors). (E) Extension length of differentiated hNSCs was measured after 7 days of culture on PDL, POMA, nonaligned, and aligned collagen using SEM pictures and NeuronJ software. Mean values ± standard error of the mean of three independent are shown. *p < 0.05 and ***p < 0.001 compared with PDL, ^###p < 0.001 compared with POMA, ⁺⁺p < 0.01 compared with nonaligned collagen fibrils (post hoc t-test with Bonferroni adjustment). PDL, poly-d-lysine; DAPI, 4,6-diamidino-2-phenylindole. Color images available online at www.liebertonline.com/ten.

To determine the effect of fibril alignment on neurite outgrowth, extension length was quantified in SEM photographs by measuring each extension along its entire length from the neurosphere to its end. We measured 109–422 extensions per independent experiment. hNSC extension length significantly varied between seeding surfaces (p = 0.0014, F-value = 17.01; ANOVA; n = 3 independent experiments; Figs. 2 and 4E). Extensions on aligned collagen fibrils (long fibrils at high density) were significantly longer (145 ± 8 μm) compared with those on PDL (86 ± 8 μm; p = 0.0003; post hoc t-test), POMA (88 ± 4 μm; p = 0.0008; post hoc t-test), and nonaligned collagen (111 ± 3 μm; p = 0.0073; post hoc t-test). On nonaligned collagen, extension outgrowth was significantly higher compared with PDL surfaces (p = 0.029). No significant differences were observed between PDL and POMA substrates.