Neurogenesis in Postnatal Rat Spinal Cord: A Study in Primary Culture

See allHide authors and affiliations

Science  25 Apr 1997:
Vol. 276, Issue 5312, pp. 586-589
DOI: 10.1126/science.276.5312.586


Spinal cord injuries result in paralysis, because when damaged neurons die they are not replaced. Neurogenesis of electrophysiologically functional neurons occurred in spinal cord cultured from postnatal rats. In these cultures, the numbers of immunocytochemically identified neurons increased over time. Additionally, neurons identified immunocytochemically or electrophysiologically incorporated bromodeoxyuridine, confirming they had differentiated from mitotic cells in vitro. These findings suggest that postnatal spinal cord retains the capacity to generate functional neurons. The presence of neuronal precursor cells in postnatal spinal cord may offer new therapeutic approaches for restoration of function to individuals with spinal cord injuries.

Neurogenesis occurs postnatally in tissue from certain brain regions (1-3). However, generation of newly differentiated neurons in rat spinal cord is thought to cease by embryonic day 17 (4). Furthermore, although spinally transected neonatal rats (0 to 12 days old) recover substantial motor function mediated by spinal cord caudal to the site of injury, limited behavioral recovery is seen following transection on postnatal day 15 (P15) or later (5). We investigated whether postnatal spinal cord contains cells with a capacity to generate functional neurons.

Cervical spinal cord tissue was harvested from rats at P15 to P16 and cultured on laminin-coated cover slips (6, 7). Neurons and glia were identified immunocytochemically with antibodies directed against two different neuronal proteins and one glial-specific protein: class III neuron-specific β-tubulin (TuJ1), neuron-specific enolase (NSE), and glial fibrillary acidic protein (GFAP). The TuJ1 antibody is directed against class III β-tubulin, a tubulin isotype expressed only by neurons (8, 9) early in, during, or directly after the final mitotic division (10). The antibody to NSE used for these studies is directed against the γγ epitope of NSE and selectively identifies neurons (11).

Cells immunopositive for TuJ1 or NSE, but not GFAP, were identified in several cultures on various days [4 to 31 days in vitro (DIV)] after plating. The number of these presumptive neurons increased with advancing time in culture (Fig. 1). However, it is conceivable that increased expression of neuronal proteins by neurons already present at the earliest time point examined may have accounted for this result. We investigated this possibility with the following experiment.

Figure 1

Numbers of cells immunopositive for neuronal markers increase with time in vitro (27). (A) TuJ1-immunopositive cell numbers are shown at various times in vitro (mean + SEM). Analysis of variance: P< 0.005, F(3,13) = 10.0. (B) A number of NSE-immunopositive cells are shown on various DIV [mean + SEM; *P < 0.05 (Student’s t test)]. Anti-GFAP immunocytochemistry demonstrated that <3% of cells were astrocytes.

Several cultures, plated at the same density, were incubated either with or without the mitotic inhibitor 5-fluoro-2-deoxyuridine (FUDR; 15 μg/ml) from 4 to 12 DIV. At 12 DIV, cultures maintained without FUDR exhibited a marked increase in TuJ1- and NSE-immunopositive cells compared with FUDR-treated cultures (Fig. 2). This result contrasts with the mutual increase that would have been expected in FUDR-treated cultures if increased labeling was due to up-regulation of neuronal protein expression. Despite differences in neuronal numbers that we attribute to variability in composition of conditioned media, we observed this result consistently in each trial. These data confirm that cell division is necessary for the increase of TuJ1- and NSE-immunopositive cells to occur. However, this experiment does not determine that mitotic cells exclusively differentiate into neurons. The possibility exists that generation of nonneuronal cells stimulates neurons to up-regulate neuronal protein expression. Therefore, two experiments were performed to determine if mitotic cells differentiated to become neurons in culture.

Figure 2

FUDR inhibited the increase of immunocytochemically identified neurons over time (28). Shown are (A) TuJ1-immunopositive cell counts and (B) NSE-immunopositive cell counts. Stacked bar graphs represent total TuJ1- or NSE-immunopositive cell numbers at 12 DIV after an 8-day incubation in either neuronal media (no FUDR) or neuronal media with FUDR. This experiment was performed three times with different sets of cultures, plated on three different days. Comparisons were only made between cover slips plated on the same day, as indicated by bar graph segments with matching shades of gray or black. Note that in each case neuronal cell counts were greater when cultured without FUDR.

Postnatal cultures were incubated with bromodeoxyuridine (BrdU, 1 μM), a thymidine analog that is incorporated into DNA during the S phase of the cell cycle (12). Because cells incorporating BrdU during this phase of DNA replication subsequently undergo mitosis, detection of BrdU in a cell confirms it is the progeny of a mitotic division in vitro. After 7 to 20 days of incubation with BrdU, cultures were double-immunostained for BrdU content and either class III β-tubulin or NSE. Subsequent cell counts demonstrated colocalization of BrdU in 10 of 42 TuJ1-positive cells, consecutively identified on one cover slip (Fig. 3, A and B), indicating that these presumptive neurons differentiated from cells born in culture. Additionally, on average, 20% of anti-NSE–labeled cells were also immunopositive for BrdU (Fig. 3, C through E), confirming results obtained with TuJ1/BrdU double-immunocytochemistry.

Figure 3

Immunocytochemically identified neurons positive for BrdU content (29). Shown are (A) a TuJ1-immunopositive cell; (B) the cell in (A), double-immunostained for BrdU; (C) a different cell, demonstrating NSE-immunoreactivity; and (D) the cell in (C), immunopositive for BrdU. Scale bars, 20 μm. (E) The line graph shows NSE-immunopositive cell numbers increased primarily during the second week in vitro. The superimposed bar graph represents percent of NSE-immunopositive cells also immunostained for BrdU. Double-labeled cells comprised a consistent proportion of NSE-immunopositive cells throughout the culture period. These data demonstrate that presumptive neurons are produced in vitro throughout the culture period. Data are expressed as mean + SEM. NSE-immunopositive cell counts are repeated from Fig. 1B.

Collectively, these immunocytochemical studies provide evidence that postnatal spinal cord contains cells that retain the capacity for neurogenesis in vitro. However, although TuJ1 identifies a neuron-specific tubulin isotype (8, 10, 13), several studies report that reactive glia may express other neuronal proteins such as microtubule-associated protein 2 (MAP2) and NSE (14). Furthermore, because glial cells proliferate in spinal cord postnatally (15), we addressed the possibility that some NSE/BrdU double-immunostained cells were glia. To confirm that at least a subset of spinal cord cells incorporating BrdU were functional neurons, we electrophysiologically studied 60 cultured postnatal cells between 20 and 27 DIV, using whole-cell patch clamp (Fig. 4).

Figure 4

BrdU-immunoreactivity in electrophysiologically identified neurons (30). Shown are (A) a dextran fluorescein–injected neuron from which all-or-none APs were elicited and (B) BrdU-immunoreactivity of the cell in (A) (arrow). (C) All-or-none APs were elicited from the neuron in (A) and (B) by incremental 20-pA depolarizing current steps from RMP (–66 mV). (D) A second dextran fluorescein–injected cell from which graded APs were generated and (E) BrdU-immunoreactivity of the cell in (D) (arrow) are shown. (F) Graded APs were generated by the neuron in (D) and (E) to incremental 50-pA depolarizing current steps from RMP (–51 mV). Shown are (G) a dextran fluorescein–filled cell (arrow) which generated rudimentary voltage spikes and (H) BrdU-immunoreactivity of the cell in (G). The white asterisk (G and H) indicates a dead cell stained by fluorescein, released during pipette withdrawal from the filled cell. (I) Small voltage spikes were generated by the cell in (G) and (H) to incremental 150-pA depolarizing current steps from RMP (–35 mV). Scale bars, 20 μm.

Most cells exhibited some level of electrical excitability as demonstrated by the ability of depolarization to elicit voltage-gated currents. However, only 20 cells responded to depolarizing current steps by generating voltage spikes (16). These cells were iontophoretically injected with 0.2% dextran fluorescein dye and subsequently immunoreacted for BrdU content.

Application of depolarizing current evoked three types of voltage traces from the cells’ resting membrane potentials (RMPs) in 18 cells filled sufficiently for post hoc identification. First, all-or-none action potentials (APs) were generated in three BrdU-positive (Fig. 4, A through C) and three BrdU-negative neurons (16). Second, graded APs were elicited from six BrdU-positive (Fig. 4, D through F) and two BrdU-negative neurons (16). These recordings are comparable to those described in cultured small hippocampal neurons (17) and developing neurons derived from a multipotent precursor cell line (18). Finally, four BrdU-positive cells (Fig. 4, G through I) exhibited small rudimentary voltage spikes of the type reported in neonatal hypoglossal motorneurons (19), developing neurons (18, 20), and computer-modeled immature neurons (20).

AP-like responses are reported in pancake astroglia from cultured P0 spinal cord (21). However, the cells we studied electrophysiologically differed in two important ways. First, unlike flat nonprocess-bearing pancake astrocytes, cells in this study were phase-bright with two or more projections. Secondly, pancake astrocytes generate AP-like responses only after removal of Na+-channel inactivation by hyperpolarization to potentials negative to –70 mV. Because these cells rest at around –40 mV, it is unlikely that AP-like responses would occur spontaneously in vivo (22). In contrast, APs in the present study were elicited from each neuron’s RMP (–32 to –69 mV). This finding carries particular significance because it suggests that APs could be elicited in newly generated postnatal spinal cord neurons under physiologic conditions in vivo.

The generation of neurons raises the question of their source. The subventricular zone (SVZ) of mature mammalian brain contains precursor cells capable of differentiating into neurons (2). Although this area contributes primarily to gliogenesis (23), in vivo and in vitro studies have demonstrated an SVZ region containing essentially all TuJ1-immunoreactive cells (13). Like the cerebral ventricles, the spinal cord central canal derives from a cavity within the embryonic neural tube (24) and is lined with neuroepithelial cells that divide and differentiate into central nervous system neurons and glia. Conceivably, central canal cells in these postnatal cultures may have given rise to the newly generated neurons we observed.

This study confirms neuronal function (AP generation) in mammalian postnatal spinal cord cells identified with a proliferative immunomarker (BrdU). The observation of neuronal proliferation in the presence of preexisting neurons, unlike implantation of peripheral nerve or isolated neuronal precursors (25), suggests that spinal neurogenesis may be induced in situ (16). Indeed, patch-clamp studies conducted on more than 131 cultured spinal cord cells at P15 and P16 demonstrate functionality of neurons for up to 32 DIV, including the apparent capacity to form synaptic connections (6). Collectively, these findings support the utility of postnatal rat spinal cord cultures as a practical system to investigate factors involved in neurogenesis of functional postnatal spinal cord neurons. Parallel in vivo studies could lead to the design of innovative strategies for treating injured spinal cord. This would represent an important clinical advance given the severity of sequelae to cervical spinal cord injuries, including paralysis and dependence on ventilator-assisted breathing (26).

  • * To whom correspondence should be addressed. E-mail: george{at}


View Abstract

Navigate This Article