Abstract

In humans, the developmental origins of interneurons in the third trimester of pregnancy and the timing of completion of interneuron neurogenesis have remained unknown. Here, we show that the total and cycling Nkx2.1+ and Dlx2+ interneuron progenitors as well as Sox2+ precursor cells were higher in density in the medial ganglionic eminence (MGE) compared with the lateral ganglionic eminence and cortical ventricular/subventricular zone (VZ/SVZ) of 16–35 gw subjects. The proliferation of these progenitors reduced as a function of gestational age, almost terminating by 35 gw. Proliferating Dlx2+ cells were higher in density in the caudal ganglionic eminence (CGE) compared with the MGE, and persisted beyond 35 gw. Consistent with these findings, Sox2, Nkx2.1, Dlx2, and Mash1 protein levels were higher in the ganglionic eminences relative to the cortical VZ/SVZ. The density of gamma-aminobutyric acid-positive (GABA+) interneurons was higher in the cortical VZ/SVZ relative to MGE, but Nkx2.1 or Dlx2-expressing GABA+ cells were more dense in the MGE compared with the cortical VZ/SVZ. The data suggest that the MGE and CGE are the primary source of cortical interneurons. Moreover, their generation continues nearly to the end of pregnancy, which may predispose premature infants to neurobehavioral disorders.

Introduction

Several neuropsychiatric and behavioral disorders seen in children and adults have developmental origins (Clowry et al. 2010). Autism, attention deficits, hyperactivity, schizophrenia, and depressive psychosis are attributed to defects in the development and function of interneurons (Marin 2012). Information on the location of interneuron neurogenesis and timing of its completion during human gestation is fundamental to understanding these disorders. However, the developmental origin of cortical interneurons has been debated in humans, and the timing of cessation of interneuron neurogenesis during human pregnancy also remains unknown (Sanai et al. 2011; Feliciano and Bordey 2013; Hladnik et al. 2014). If generation of interneurons continues in the third trimester, this might have important clinical implications. Perinatal insults, including hypoxia–ischemia, intraventricular hemorrhage, sepsis, and maternal chorioamnionitis, could adversely affect protracted generation of interneurons during late pregnancy, which might predispose survivors to neurobehavioral disorders. Therefore, we asked where and when interneuron neurogenesis was occurring and when it ceased during human development.

Gamma-aminobutyric acid (GABA)-releasing interneurons constitute the major class of inhibitory neurons of the brain. They exhibit essential and powerful roles in modulating the electrical activity of the excitatory pyramidal cells onto which they synapse (Marin 2012). By these interactions, interneurons control pyramidal cell firing, modulate cortical rhythms, and regulate cortical function as well as plasticity (Kepecs and Fishell 2014). Interneurons are subdivided into subtypes, including parvalbumin+, calretinin+, somatostatin+, or vasointestinal peptide+, based on the calcium-binding proteins or neuropeptides they produce (Wonders and Anderson 2006). Whereas somatostatin+ and parvalbumin+ interneurons originate from the medial ganglionic eminence (MGE), calretinin+ and vasointestinal peptide+ interneurons are derived from caudal ganglionic eminence (CGE) in rodents (Wonders and Anderson 2006).

In rodents, approximately 70% of cortical neurons originate from the MGE, about 30% from the CGE, and 5% from the preoptic area (Wonders and Anderson 2006). In primates as well, it is now accepted that most of the cortical interneurons are generated in the ventral ganglionic eminence (GE). However, how much the cortical ventricular and subventricular zones (VZ/SVZ) contribute to the production of cortical interneurons has sparked a debate. A number of studies in humans and non-human primates have reported that the VZ and SVZ of the dorsal forebrain are important source of interneurons (Letinic et al. 2002; Petanjek et al. 2009; Jakovcevski et al. 2011; Clowry 2014; Radonjic et al. 2014; Al-Jaberi et al. 2015). However, 2 recent studies of human fetuses (<24 gw), on a relatively small sample size, suggest that the majority of interneurons in the first and second trimester of pregnancy are produced in the GE, and that the cortical VZ/SVZ is just a minor contributor of interneurons (Hansen et al. 2013; Ma et al. 2013; Molnar and Butt 2013). The reason for this discrepancy between studies on the origin of cortical interneurons from the cortical VZ/SVZ can be attributed to the lack of specificity of markers, different fixation protocols, selection of cortical area, small sample size, or variability between samples owing to prenatal influences. Hence, it is important to assess the density of interneuron progenitors in the cortical VZ/SVZ relative to the GE on a relatively large sample size and evaluate interneuronal production during the third trimester.

Interneuronal development is regulated by key transcription factors, including Nkx2.1, Dlx1/2, Mash1, Lhx6/8, Gsh2, and Sox6 (Kelsom and Lu 2013). Importantly, the Sox2 transcription factor, which is active in neural stem cells, directly regulates Nkx2.1 expression and is critical for embryonic development of the MGE (Ferri et al. 2013). Nkx2.1+ progenitors are localized to the MGE in rodents and give rise to parvalbumin- and somatostatin-expressing interneurons in both rodents and monkeys (Wonders and Anderson 2006; Ma et al. 2013). The Dlx family of genes plays key roles in specification of both MGE and CGE interneurons (Corbin et al. 2000; Long et al. 2009). The CGE harbors Dlx2+ and Coup-TFII+ progenitors which generate the calretinin- and vasointestinal peptide-expressing interneurons (Nery et al. 2002; Kanatani et al. 2008). Based on these considerations, we hypothesized that the Nkx2.1+ and Dlx2+ interneuronal progenitors were primarily produced in the MGE in humans, and that generation of these progenitors would continue until the end of pregnancy. We also postulated that the CGE would exhibit an abundance of Dlx2+ and Coup-TFII+ progenitors, which might diminish in number as a function of increasing gestational age.

Materials and Methods

Human Subjects

The Institutional Review Board at New York Medical College (Valhalla, NY) approved the use of autopsy materials from fetuses and premature infants for this study. The study materials included brain tissue samples taken from abortuses of 16–22 gw and autopsies of premature infants of 23–40 gw. The autopsy samples were obtained at a postmortem interval of <18 h. All infants included in the study were of <5 days of postnatal age. This was intended to minimize the impact of postnatal events occurring in the Neonatal Units on neurogenesis. We excluded premature infants with moderate-to-severe intraventricular hemorrhage, major congenital anomalies, history of neurogenetic disorder, chromosomal defects, culture-proven sepsis, meningitis, hypoxic–ischemic encephalopathy, and infants receiving extracorporeal membrane oxygenator treatment from the study. Any brain tissue showing autolysis or necrosis on hematoxylin and eosin staining was also excluded from the study. Brain samples to assess MGE, lateral ganglionic eminence (LGE), and cortical SVZ were classified into 4 groups (Table 1): (1) Fetuses of 16–22 gw (n = 5), (2) premature infants of 23–25 gw (n = 5), (3) premature infants of 26–28 gw (n = 5), and (4) premature (near-term) infants of 29–35 gw (n = 5). Of these, 8 were females and 12 were males. This classification enabled equal distribution of subjects in each group. For the evaluation of CGE, we included 18 brains, which were classified into 5 groups (Table 1): (1) fetuses of 16–22 gw (n = 4), (2) premature infants of 23–25 gw (n = 5), (3) premature infants of 26–28 gw (n = 3), (4) premature infants of 29–35 gw (n = 3), and (5) near-term infants of 36–40 gw (n = 3). Of these, 10 were females and 8 males. These samples were collected during the last 12 years (2002–2014) at Westchester Medical Center-New York Medical College in Valhalla, NY.

Table 1

Characteristics of human fetuses and premature infants (postmortem)

Case Gestational age (weeks) Weight (g) Sex Clinical diagnosis and cause of death 
16 — Female Undetermineda 
16 — Male Cervical incompetencea 
17 — Female Spontaneous abortionb 
19 — Female Spontaneous terminationa,b 
21 — Female Preterm laborb 
22 500 Male Preterm labor, respiratory failurea,b 
22 510 Male Immaturity, respiratory failurea 
23 570 Female Clinical sepsisa 
23 560 Male Immaturity, respiratory failurea 
10 23 630 Male Immaturity and respiratory failureb 
11 23 590 Female Immaturity and respiratory failureb 
12 24 680 Female Pulmonary hypertensionb 
13 24 720 Male Necrotizing enterocolitisb 
14 24 390 Male Immaturity and respiratory failurea 
15 25 710 Male Clinical sepsisa 
16 25 730 Female Metabolic acidosis, cardioresp failurea,b 
17 26 920 Female Clinical sepsis and shocka 
18 27 650 Female twin  Respiratory failurea,b 
19 27 890 Male PPHN, PIE, respiratory failurea,b 
20 27 680 Male Clinical sepsisa,b 
21 28 900 Male Respiratory failurea 
22 29 980 Female Respiratory failureb 
23 31 1690 Male Necrotizing enterocolitesa 
24 32 1392 Male Necrotizing enterocolitesa 
25 33 1940 Male Respiratory failureb 
26 33 1830 Male Non-immune hydropsb 
27 33 2380 Male Non-immune hydropsa 
28 33 2305 Female Diaphragmatic herniaa 
29 35 2900 Female Oligohydramnios, hypoplastic lunga 
30 36 2750 Male Transposition of great arteriesb 
31 37 3210 Female Hypoplastic lung, respiratory failureb 
32 40 3100 Female Hypoplastic left heartb 
Case Gestational age (weeks) Weight (g) Sex Clinical diagnosis and cause of death 
16 — Female Undetermineda 
16 — Male Cervical incompetencea 
17 — Female Spontaneous abortionb 
19 — Female Spontaneous terminationa,b 
21 — Female Preterm laborb 
22 500 Male Preterm labor, respiratory failurea,b 
22 510 Male Immaturity, respiratory failurea 
23 570 Female Clinical sepsisa 
23 560 Male Immaturity, respiratory failurea 
10 23 630 Male Immaturity and respiratory failureb 
11 23 590 Female Immaturity and respiratory failureb 
12 24 680 Female Pulmonary hypertensionb 
13 24 720 Male Necrotizing enterocolitisb 
14 24 390 Male Immaturity and respiratory failurea 
15 25 710 Male Clinical sepsisa 
16 25 730 Female Metabolic acidosis, cardioresp failurea,b 
17 26 920 Female Clinical sepsis and shocka 
18 27 650 Female twin  Respiratory failurea,b 
19 27 890 Male PPHN, PIE, respiratory failurea,b 
20 27 680 Male Clinical sepsisa,b 
21 28 900 Male Respiratory failurea 
22 29 980 Female Respiratory failureb 
23 31 1690 Male Necrotizing enterocolitesa 
24 32 1392 Male Necrotizing enterocolitesa 
25 33 1940 Male Respiratory failureb 
26 33 1830 Male Non-immune hydropsb 
27 33 2380 Male Non-immune hydropsa 
28 33 2305 Female Diaphragmatic herniaa 
29 35 2900 Female Oligohydramnios, hypoplastic lunga 
30 36 2750 Male Transposition of great arteriesb 
31 37 3210 Female Hypoplastic lung, respiratory failureb 
32 40 3100 Female Hypoplastic left heartb 

aCoronal slice was used for immunohistochemistry to assess the MGE, LGE, and cortical SVZ.

bCoronal slice was used for immunohistochemistry to assess the CGE.

Human Tissue Collection, Immunohistochemistry, Quantification of Neuronal Progenitors Under Confocal Microscope, Western Blot Analyses, Quantitative Real-Time Polymerase Chain Reaction

These techniques are described in Supplementary method section.

Statistics and Analysis

Data are expressed as mean ± one standard error of the mean (SEM). To determine differences in the density of interneuron subtype or interneuron progenitors and their proliferation in the cortical SVZ, MGE and LGE in 4 age groups of human subjects, including 16–22, 23–25, 26–28, and 29–35 gw, a two-way analysis of variance (ANOVA) with repeated measures was used. The repeated factor was applied to the 3 brain regions, MGE, LGE, and cortical SVZ. Optical densities obtained by western blot analyses and real-time qPCR data were analyzed by a two-way ANOVA with repeated measures. All post hoc comparisons to test for differences between means were done using the Tukey multiple comparison test at the 0.05 significance level. To compare progenitors in the 5 gestational groups of the CGE, a one-way ANOVA was used. For 2 group comparisons, either a T-test or a Mann–Whitney U-test was performed, as applicable.

Results

VZ and SVZ in the GE Consistently Reduced in Thickness with Advancing Gestational Age

To determine the origin of cortical interneurons, we assessed Sox2+ precursor cells, Dlx2+ and Nkx2.1+ intermediate progenitors, and GABA+ interneurons in the MGE, LGE, and cortical SVZ of human fetuses and premature infants (16–35 gw, n = 20 brains, Table 1). Immunolabeling of coronal sections with Sox2, Ki67, and Dlx2 antibodies delineated the VZ and SVZ around the cerebral ventricle (Fig. 1a). The VZ was comprised of primarily Sox2+ progenitor cells, whereas the SVZ was enriched with both Dlx2+ and Sox2+ progenitors. Proliferating cells (Ki67+) were abundant in the VZ and SVZ, whereas the adjacent intermediate zone (white matter) exhibited a relative paucity of cycling neural cells. Dlx2+ cells were present in the SVZ, but almost absent in the VZ. We measured thickness and cell density of the VZ and SVZ in coronal sections of 4 subsets of fetuses and premature infants: 16–22, 23–25, 26–28, and 29–35 gw (n = 5 each). We found that the thickness of VZ and SVZ consistently decreased in all the 3 germinal regions—MGE, LGE, and cortical SVZ (P < 0.001 all)—with advancing gestational age (Fig. 1a). The widths of the SVZ were similar between the 3 germinal zones for all gestational groups, except being higher in the MGE and LGE relative to the cortical SVZ for 16–22 gw fetuses (P = 0.001 and 0.025, respectively). The width of the VZ was also comparable in the 3 germinal regions for all gestational age categories, except for 26–28 gw, in which VZ was thinner in the cortical SVZ compared with the LGE and MGE (P = 0.02 both). Cell density (DAPI+ nuclear count) of the VZ and SVZ was comparable between germinal zones and decreased with increasing gestational age in all 3 regions (Supplementary Fig. 1a). Taken together, the attenuation of VZ and SVZ thickness and cell density with increasing gestational age suggests that interneuron neurogenesis slows down with fetal maturation in the third trimester (>24 gw). In addition, the large volume of the SVZ in the 3 germinal regions, enriched with stem cells and progenitors, suggests that SVZ is the major contributor of interneuron neurogenesis in humans.

Figure 1.

VZ and SVZ in the GE diminished in width as gestational age increased. (a) Representative immunofluorescence of coronal sections from the MGE of a 20-gw fetus labeled with antibodies specific to Sox2, Dlx2, and Ki67. Scale bar, 50 µm. Bar graph shows mean ± SEM (n = 4 brains each group). The VZ and SVZ thickness reduced as a function of gestational age. *P < 0.05 and **P < 0.01 for MGE versus cortical SVZ; #P < 0.05 for LGE versus cortical SVZ. (b) Cresyl violet staining of coronal section through the right cerebral hemisphere at the level of the head of caudate nucleus showing cortical SVZ, MGE, and LGE of 20 gw fetus (left). Scale bar, 0.25 cm. A typical triple immunolabeling of coronal brain section from 22 gw human fetus with Tbr2, Sox2, and Ki67 antibodies. Note the presence of Tbr2 immunoreactivity (arrows) in the cortical SVZ, but absence in the LGE. Scale bar, 50 µm. D, dorsal; V, ventral; M, medial; L, lateral. (c) Nkx2.1 immunolabeling of the MGE of 22 gw fetus showing Nkx2.1+ cells are densely packed in the MGE, but sparse in the LGE. Scale bar, 100 µm. (d) Representative immunofluorescence of cryosections from the CGE of a 19-gw fetus labeled with Coup-TFII- and Dlx2-specific antibodies. Coup-TFII+ cells and Dlx2+ cells are densely arranged in the CGE (both single- and double-labeled images are from the same field of view). Scale bar, 100 µm. Inset shows that coronal section of this 19 gw fetus stained with cresyl violet depicting the CGE. cSVZ, cortical SVZ.

Figure 1.

VZ and SVZ in the GE diminished in width as gestational age increased. (a) Representative immunofluorescence of coronal sections from the MGE of a 20-gw fetus labeled with antibodies specific to Sox2, Dlx2, and Ki67. Scale bar, 50 µm. Bar graph shows mean ± SEM (n = 4 brains each group). The VZ and SVZ thickness reduced as a function of gestational age. *P < 0.05 and **P < 0.01 for MGE versus cortical SVZ; #P < 0.05 for LGE versus cortical SVZ. (b) Cresyl violet staining of coronal section through the right cerebral hemisphere at the level of the head of caudate nucleus showing cortical SVZ, MGE, and LGE of 20 gw fetus (left). Scale bar, 0.25 cm. A typical triple immunolabeling of coronal brain section from 22 gw human fetus with Tbr2, Sox2, and Ki67 antibodies. Note the presence of Tbr2 immunoreactivity (arrows) in the cortical SVZ, but absence in the LGE. Scale bar, 50 µm. D, dorsal; V, ventral; M, medial; L, lateral. (c) Nkx2.1 immunolabeling of the MGE of 22 gw fetus showing Nkx2.1+ cells are densely packed in the MGE, but sparse in the LGE. Scale bar, 100 µm. (d) Representative immunofluorescence of cryosections from the CGE of a 19-gw fetus labeled with Coup-TFII- and Dlx2-specific antibodies. Coup-TFII+ cells and Dlx2+ cells are densely arranged in the CGE (both single- and double-labeled images are from the same field of view). Scale bar, 100 µm. Inset shows that coronal section of this 19 gw fetus stained with cresyl violet depicting the CGE. cSVZ, cortical SVZ.

The identification of MGE, LGE, and cortical SVZ was performed on the basis of their anatomical location and immunoreactivity to specific makers. The pallial–subpallial boundary, separating cortical SVZ and GE, was unclear in fetuses >20 gw. However, the presence of Tbr2 (progenitors of glutamatergic neurons) immunolabeling in the cortical SVZ and its absence in the GE demarcated the 2 regions (Fig. 1b). The MGE was identified by its ventro-medial location and by an abundance of Nkx2.1+ immunoreactivity, whereas LGE was recognized by its dorso-lateral position and by a relative paucity of Nkx2.1+ cells (Fig. 1c). The CGE displayed an abundance of Coup-TFII+ cells, and a relative absence of Nk2.1+ cells (Fig. 1d).

Nkx2.1+ Progenitors Were Abundant in the MGE, Moderate in the LGE, and Scarce in the Cortical SVZ

Nkx2.1+ progenitors primarily originate from the MGE in both rodents and primates; however, their extent of origin from the cortical SVZ in humans has been debated (Letinic et al. 2002; Jakovcevski et al. 2011; Hansen et al. 2013; Ma et al. 2013; Radonjic et al. 2014). Therefore, we evaluated Nkx2.1+ progenitors in the cortical SVZ (VZ/SVZ) compared with MGE and LGE in four subsets of fetuses and premature infants: 16–22, 23–25, 26–28, and 29–35 gw (n = 5 each; Table 1). Triple labeling of coronal sections with Nkx2.1 and Ki67 (proliferation marker) antibodies along with DAPI (nuclear stain) revealed that the Nkx2.1+ cells were densely packed in both VZ and SVZ of the MGE, sparse in the LGE, and few in the cortical SVZ. Moreover, their densities in the MGE were higher in subjects born in the second trimester (≤24 gw) compared with infants born in the third trimester (>24 gw, P < 0.001; Fig. 2a,b and Supplementary Fig. 1b,c). Of all cells in the MGE (DAPI+), Nkx2.1+ progenitors constituted on average 39.6 ± 6.7, 31.6 ± 5.9, 27.9 ± 6.5, and 4.6 ± 2.8% for 16–22, 23–25, 26–28, and 29–35 gw subjects, respectively. In the LGE and cortical SVZ, Nkx2.1+ cells were few (<2 and 1% of DAPI+ cells, respectively) and were heterogeneously dispersed throughout the SVZ in all gestational age categories.

Figure 2.

Nk2.1+ progenitors densely packed in the MGE of 16–28 gw fetuses and markedly reduced in 29–35 gw. (a) Representative immunofluorescence of coronal sections from a 20-gw fetus and a 25-gw premature infant labeled with Nkx2.1- and Ki67-specific antibodies. Inset shows a high power view of the boxed region in the image. Note abundance of both non-proliferating (arrows) and proliferating (arrowheads) Nkx2.1+ cells in the MGE compared with the LGE and cortical SVZ. Scale bar, 50 μm. Shown in the lower panel, above and to the right of the image, are orthogonal views in xz and yz planes of a composite z-stack of a series of confocal images taken 0.8 µm apart; the images depict Ki67-stained nucleus embedded within Nkx2.1+-stained cells (arrowhead) in a 25-gw infant. Scale bars, 20 µm. (b) A typical triple immunolabeling of coronal brain section from 20 and 25 gw subjects labeled with Nkx2.1- and Ki67-specific antibodies and DAPI. The non-proliferating (arrows) and proliferating (arrowheads) Nkx2.1+ cells are shown. Note the percentage of Nkx2.1+ cells (Nkx2.1+/DAPI+) was higher in the MGE relative to the LGE and cortical SVZ. Scale bar, 20 µm. (c) Bar diagram shows mean ± SEM (n = 5 each gestational group). Total Nkx2.1+ cells were higher in number in the MGE compared with the LGE and cortical SVZ for all ages, except for 29–25 gw. *P < 0.001 MGE versus LGE and #P < 0.001 MGE versus cortical SVZ within gestational group comparison. P < 0.001, P < 0.001, and ψP < 0.001 16–22, 23–25, and 26–28 versus 29–35 gw in the MGE, respectively. (d) Bar chart shows mean ± SEM (n = 5 each group). Nkx2.1+Ki67+ cells were almost absent in the LGE and cortical SVZ and reduced with advancing gestation. *P < 0.001 MGE versus LGE and #P < 0.001 for MGE versus cortical SVZ for within gestational group comparison. P < 0.001, P < 0.001 16–22 versus 26–28 and 29–35 gw in the MGE, respectively. ψP < 0.001 23–25 versus 29–35 gw in the MGE. cSVZ, cortical SVZ.

Figure 2.

Nk2.1+ progenitors densely packed in the MGE of 16–28 gw fetuses and markedly reduced in 29–35 gw. (a) Representative immunofluorescence of coronal sections from a 20-gw fetus and a 25-gw premature infant labeled with Nkx2.1- and Ki67-specific antibodies. Inset shows a high power view of the boxed region in the image. Note abundance of both non-proliferating (arrows) and proliferating (arrowheads) Nkx2.1+ cells in the MGE compared with the LGE and cortical SVZ. Scale bar, 50 μm. Shown in the lower panel, above and to the right of the image, are orthogonal views in xz and yz planes of a composite z-stack of a series of confocal images taken 0.8 µm apart; the images depict Ki67-stained nucleus embedded within Nkx2.1+-stained cells (arrowhead) in a 25-gw infant. Scale bars, 20 µm. (b) A typical triple immunolabeling of coronal brain section from 20 and 25 gw subjects labeled with Nkx2.1- and Ki67-specific antibodies and DAPI. The non-proliferating (arrows) and proliferating (arrowheads) Nkx2.1+ cells are shown. Note the percentage of Nkx2.1+ cells (Nkx2.1+/DAPI+) was higher in the MGE relative to the LGE and cortical SVZ. Scale bar, 20 µm. (c) Bar diagram shows mean ± SEM (n = 5 each gestational group). Total Nkx2.1+ cells were higher in number in the MGE compared with the LGE and cortical SVZ for all ages, except for 29–25 gw. *P < 0.001 MGE versus LGE and #P < 0.001 MGE versus cortical SVZ within gestational group comparison. P < 0.001, P < 0.001, and ψP < 0.001 16–22, 23–25, and 26–28 versus 29–35 gw in the MGE, respectively. (d) Bar chart shows mean ± SEM (n = 5 each group). Nkx2.1+Ki67+ cells were almost absent in the LGE and cortical SVZ and reduced with advancing gestation. *P < 0.001 MGE versus LGE and #P < 0.001 for MGE versus cortical SVZ for within gestational group comparison. P < 0.001, P < 0.001 16–22 versus 26–28 and 29–35 gw in the MGE, respectively. ψP < 0.001 23–25 versus 29–35 gw in the MGE. cSVZ, cortical SVZ.

A comparisons between gestational groups demonstrated that the density of total Nkx2.1+ cells (both Ki67 and Ki67+) in the MGE was higher in 16–22, 23–25, and 26–28 gw subjects relative to 29–35 gw infants (P < 0.001 all; Fig. 2c), but was similar between 16–22, 23–25, and 26–28 gw subjects. We then compared between germinal regions and found that the Nkx2.1+ cells were more dense in the MGE compared with LGE and cortical SVZ for each gestational group (P < 0.001 for all comparisons), except for 29–35 gw infants.

We next analyzed proliferation of Nkx2.1+ cells in these germinal zones and found that the number of cycling Nkx2.1+ cells in the MGE consistently decreased as a function of gestational age. Accordingly, cycling Nkx2.1+ cells in the MGE were higher in density in 16–22 gw compared with 26–28 and 29–36 gw infants (P < 0.001, Fig. 2d). In addition, proliferating Nkx2.1+ cells in the MGE were more abundant in 23–25 gw relative to 29–35 gw infants (P < 0.001). A comparison between the 3 germinal regions revealed that the density of cycling Nkx2.1+ cells was higher in the MGE relative to the LGE and cortical SVZ for 16–22 and 23–25 gw subjects (P < 0.001 all).

Since vimentin, an intermediate filament expressed by progenitors in the VZ and SVZ, is phosphorylated during the M-phase of mitosis (Kamei et al. 1998), we used a phospho-vimentin (p-vimentin) antibody to evaluate the population of Nkx2.1+ cells expressing p-vimentin. We found that Nkx2.1+p-vimentin+ cells were most abundant in the MGE, but nearly absent in the LGE and cortical SVZ (Supplementary Fig. 2). These mitotic progenitors in the MGE reduced with increasing gestational age (P < 0.001) and were largely absent by 29–35 gw. The percentages of Nkx2.1+p-vimentin+ cells (ratio of Nkx2.1+p-vimentin+ and total Nkx2.1+ cells) in the MGE were higher in 16–22 and 23–25 gw subjects, compared with 26–28 and 29–35 gw infants (P < 0.001 all). In addition, these subset of cells were higher in the MGE relative to the LGE and cortical SVZ for 16–22 and 23–25 gw subjects (P < 0.001 all), but not for 26–28 and 29–35 gw infants.

We also noted the presence of Nkx2.1+ cells in the neostriatum. Nkx2.1+ cells seemed to be denser in the caudate nucleus compared with the globus pallidus and putamen. Moreover, similar to what was observed in the MGE, their density appeared to reduce with advancing gestational age. In summary, our data suggest that Nkx2.1+ interneuronal progenitors are predominantly located in the MGE and that their proliferation markedly slows down after 28 weeks of gestational age.

Dlx2+ Progenitors Were Widespread in the MGE, Moderate in the LGE, and Scarce in the Cortical SVZ

Since Dlx2 is expressed on the interneuron progenitors in the subpallium (Panganiban and Rubenstein 2002; Petanjek et al. 2009), we triple-labeled the brain sections with Dlx2- and Ki67-specific antibodies along with DAPI to further assess interneuron progenitors (Fig. 3a,b). We found that Dlx2+ progenitors were densely packed in the inner regions of the SVZ and were sparse in the outer SVZ of the MGE for subjects <25 gw. Dlx2+ cells were arranged in columns and clusters in the outer SVZ of the MGE, consistent with previous studies (Hansen et al. 2013; Figs 2 and 3a). Intriguingly, Dlx2+ cells in the SVZ were aligned around nestin+ radial glia fibers; this pattern was clearer in the MGE compared with the LGE (Supplementary Fig. 3). Dlx2+ cells were almost absent in the VZ, whereas Nkx2.1+ cells were densely packed in this region (Fig. 2). This suggests that Dlx2+ cells in the subpallium mainly originate in the SVZ, whereas Nkx2.1+ progenitors are produced in both the VZ and SVZ.

Figure 3.

Dlx2+ progenitors were higher in density in the MGE relative to the LGE and cortical SVZ, and disappeared by 35 gw. (a) Representative immunolabeling of coronal sections from a 20-gw fetus and a 25-gw preterm infant stained with Dlx2- and Ki67-specific antibodies. Dlx2+ progenitors were more abundant in the MGE compared with the LGE and cortical SVZ. Inset shows a high power view of the boxed region in the image. Note total (arrow) and cycling (arrowhead) Dlx2+ cells. Shown in the lower right panel, above and to the right of the image, are orthogonal views in xz and yz planes of a composite z-stack of a series of confocal images; the images depict Ki67-stained nucleus embedded within Dlx2-stained cells (arrowhead) in a 25-gw infant. Scale bar, 20 µm. (b) Representative immunofluorescence of coronal sections from 20 and 25 gw fetuses labeled with Dlx2- and Ki67-specific antibodies and DAPI. Note the percentage of Dlx2+ cells (Dlx2+/DAPI+) was higher in the MGE relative to the LGE and cortical SVZ. Proliferating (arrowhead) and non-proliferating (arrow) cells shown. Scale bar, 20 µm. (c) Bar chart shows mean ± SEM (n = 5 each group). The density of Dlx2+ cells was reduced in 29–35 gw relative to other groups. *P < 0.001 for MGE versus LGE and #P < 0.001 for MGE versus cortical SVZ within gestational group comparisons. P < 0.001, P < 0.001 and ψP < 0.001 for 29–35 versus 16–22, 23–25, and 26–28 gw in the MGE, respectively. (d) Data are mean ± SEM (n = 5 each). Cycling Dlx2+ cells were more numerous in the MGE relative to the LGE and cortical SVZ. *P < 0.001 for MGE versus LGE and #P < 0.001 for MGE versus cortical SVZ within gestational group comparisons. P < 0.001, P < 0.001, and ψP < 0.01 for 29–35 versus 16–22, 23–25, and 26–28 gw within the MGE, respectively. (e) Bar chart shows mean ± SEM (n = 5 each group). The percentage of Dlx2+ cells (Dlx2+/DAPI+ cells) was higher in the MGE compared with the LGE and cortical SVZ. *P < 0.001 for MGE versus LGE and #P < 0.001 for MGE versus cortical SVZ within gestational group comparisons. P < 0.001, P < 0.001 and ψP < 0.01 for 29–35 versus 16–22, 23–25, and 26–28 gw in the MGE, respectively. φP < 0.01 for 23–25 versus 26–28 in the MGE.

Figure 3.

Dlx2+ progenitors were higher in density in the MGE relative to the LGE and cortical SVZ, and disappeared by 35 gw. (a) Representative immunolabeling of coronal sections from a 20-gw fetus and a 25-gw preterm infant stained with Dlx2- and Ki67-specific antibodies. Dlx2+ progenitors were more abundant in the MGE compared with the LGE and cortical SVZ. Inset shows a high power view of the boxed region in the image. Note total (arrow) and cycling (arrowhead) Dlx2+ cells. Shown in the lower right panel, above and to the right of the image, are orthogonal views in xz and yz planes of a composite z-stack of a series of confocal images; the images depict Ki67-stained nucleus embedded within Dlx2-stained cells (arrowhead) in a 25-gw infant. Scale bar, 20 µm. (b) Representative immunofluorescence of coronal sections from 20 and 25 gw fetuses labeled with Dlx2- and Ki67-specific antibodies and DAPI. Note the percentage of Dlx2+ cells (Dlx2+/DAPI+) was higher in the MGE relative to the LGE and cortical SVZ. Proliferating (arrowhead) and non-proliferating (arrow) cells shown. Scale bar, 20 µm. (c) Bar chart shows mean ± SEM (n = 5 each group). The density of Dlx2+ cells was reduced in 29–35 gw relative to other groups. *P < 0.001 for MGE versus LGE and #P < 0.001 for MGE versus cortical SVZ within gestational group comparisons. P < 0.001, P < 0.001 and ψP < 0.001 for 29–35 versus 16–22, 23–25, and 26–28 gw in the MGE, respectively. (d) Data are mean ± SEM (n = 5 each). Cycling Dlx2+ cells were more numerous in the MGE relative to the LGE and cortical SVZ. *P < 0.001 for MGE versus LGE and #P < 0.001 for MGE versus cortical SVZ within gestational group comparisons. P < 0.001, P < 0.001, and ψP < 0.01 for 29–35 versus 16–22, 23–25, and 26–28 gw within the MGE, respectively. (e) Bar chart shows mean ± SEM (n = 5 each group). The percentage of Dlx2+ cells (Dlx2+/DAPI+ cells) was higher in the MGE compared with the LGE and cortical SVZ. *P < 0.001 for MGE versus LGE and #P < 0.001 for MGE versus cortical SVZ within gestational group comparisons. P < 0.001, P < 0.001 and ψP < 0.01 for 29–35 versus 16–22, 23–25, and 26–28 gw in the MGE, respectively. φP < 0.01 for 23–25 versus 26–28 in the MGE.

Our analyses revealed that the total Dlx2+ cells in the MGE were higher in density in the subjects delivered in the second compared with infants born in the third trimester (P < 0.001), and that they were almost absent by 35 gw (Fig. 3c). Accordingly, the total Dlx2+ population (cycling and non-cycling) was higher in 16–22, 23–25, and 26–28 gw subjects compared with 29–35 gw infants (P < 0.001 all, Fig. 3c). Of all neural cells (DAPI+) in the MGE, Dlx2+ progenitors comprised 20% at 16–22 gw, diminishing to 7% in 29–35 gw infants (Fig. 3e). In the LGE and cortical SVZ, the Dlx2+ population was relatively smaller compared with the MGE. Comparisons between the 3 germinal regions showed that Dlx2+ cells were higher in density in the MGE compared with the LGE and cortical SVZ for all gestational groups (P < 0.001 all, Fig. 3b), except for 29–35 gw infants, where there was no difference between regions.

We next evaluated cycling Dlx2+ cells in the 3 germinal zones (Fig. 3a,b,d). Similar to total Dlx2+ cells, proliferating Dlx2+ cells were abundant in the MGE in both fetuses (16–22 gw) and premature infants of 23–25 gw; their population steadily declined with advancing gestational age (P < 0.001), becoming almost absent by 35 gw. Accordingly, the density of cycling Dlx2+ cells in the MGE was higher in 16–22 and 23–25 gw infants compared with both 26–28 and 29–35 gw infants (P < 0.001 all). Moreover, the number of Dlx2+ cells was also higher in the MGE of 26–28 gw compared with 29–35 gw infants (P = 0.002). Comparison between germinal regions showed that cycling Dlx2+ cells were more abundant in the MGE compared with the LGE and cortical SVZ for all gestational groups (P < 0.001 all), except for 29–35 gw infants.

We next examined mitotic Dlx2+ cells by double-labeling brain sections with Dlx2 and p-vimentin antibodies (Fig. 3). We found that a substantial number of Dlx2+ cells in the MGE co-labeled for p-vimentin, and that the density of Dlx2+p-vimentin+ cells reduced as a function of gestational age (P < 0.001). Accordingly, the percentages of Dlx2+p-vimentin+ cells (ratio of Dlx2+p-vimentin+ and total Dlx2+ cells) in the MGE were higher in 16–22 and 23–25 gw subjects compared with those of 26–28 (P = 0.02, both) and 29–35 gw (P < 0.01, both) infants.

We observed a substantial number of Dlx2+ cells in the striatum in subjects ≤25 gw; however, they were relatively scant in subjects older than 25 gw. They were sparsely present in the cortical mantle and a few of them were also observed in the subcortical area. These Dlx2+ cells were postmitotic (Ki67) in both the striatum and cortical plate. Taken together, total and cycling Dlx2+ cells are primarily located in the SVZ of the MGE and their population consistently reduces with advancing gestational age, almost disappearing by 35 gw.

Majority of Nkx2.1+ or Dlx2+ Cells Were Interneuronal Progenitors, But Not Glial Precursors

Since switch from neuronal to glial production occurs during the second and third trimester (Gallo and Deneen 2014), it was important to determine whether Nkx2.1+ or Dlx2+ cells in the germinal zones were predominantly glial or neuronal progenitors. To this end, we triple-labeled the forebrain sections from fetuses and preterm infants with Nkx2.1, Dlx2, and doublecortin (migratory neuron marker) antibodies. While the majority of Nkx2.1+ or Dlx2+ cells in the MGE expressed doublecortin, doublecortin+ cells were still more abundant in the cortical SVZ and LGE compared with the MGE (Supplementary Fig. 4a,b). Moreover, the doublecortin+ interneurons in the MGE were fewer in subjects born in the second trimester (<24 gw) compared with preterm infants born in the third trimester (>24 gw). A relative paucity of doublecortin+ cells in the MGE relative to the LGE and cortical SVZ reinforces the notion that interneuronal progenitors in the MGE are more immature relative to the other germinal regions.

To determine whether Nkx2.1+ cells in the germinal regions also give rise to oligodendrocyte precursors, we triple-labeled the sections with PDGFRα (early oligodendrocyte progenitor), Ki67-, and Nkx2.1-specific antibodies. We found few PDGFRα+ (<1% of all Nkx2.1 cells) cells in the 3 germinal regions (Supplementary Fig. 4c). These cells were postmitotic (Ki67-negative) and only a few of them weakly expressed Nkx2.1. Taken together, an abundance of doublecortin-expressing Nkx2.1 or Dlx2+ progenitors and a relative paucity of PDGFRα+ cells in the MGE suggest that the majority of Nkx2.1+ or Dlx2+ cells in the MGE are interneuronal precursors.

Sox2+ Neural Progenitors Were More Abundant in the MGE Relative to the LGE and Cortical SVZ

Conditional deletion of Sox2 impairs expression of Nkx2.1 and sonic hedgehog in the MGE (Ferri et al. 2013). This led us to evaluate Sox2+ progenitor cells in the coronal sections from the MGE, LGE, and cortical SVZ. To this end, we triple-labeled the coronal sections with Sox2, Ki67, and Nkx2.1 or Dlx2 antibodies (Figs 4 and 5). Sox2+ cells were densely packed in the VZ, whereas these cells were arranged in large clusters and streaks in the SVZ of the MGE (Supplementary Fig. 5a). Merged images from the MGE of 16- to 22-week subjects showed that Sox2+ cells were interspersed with Nkx2.1+ or Dlx2+ cells throughout the SVZ without forming a definite pattern. However, in the MGE of 23–25 gw subjects, both mitotic and postmitotic Sox2+ cells were assembled to form a rim around clusters of postmitotic Nkx2.1+ or Dlx2+ cells (Fig. 4b and Supplementary Fig. 5b,c). In infants who were older than 25-week gestational age, Sox2+ cells were sparse and heterogeneously distributed in the MGE. Sox2+ cells were relatively thinly distributed in the LGE and cortical SVZ, compared with the MGE for subjects of all gestational ages (Supplementary Fig. 5a).

Figure 4.

Sox2+ and Sox2+Nkx2.1+ cells were more abundant in the MGE relative to other germinal zones. (a) Representative triple labeling of coronal sections from the MGE of 20 and 25 gw subjects stained with Sox2-, Nkx2.1-, and Ki67-specific antibodies. Note abundance of Sox2+Ki67+ and Sox2+Nkx2.1+ cells in the VZ and SVZ of the MGE (lower magnification). Scale bar, 50 µm. (b) Typical triple immunolabeling of coronal sections from 20 and 25 gw subjects (higher magnification) using Sox2-, Nkx2.1-, and Ki67-specific antibodies. Note fewer Sox2+Nkx2.1+ (arrow) and Sox2+Ki67+ (arrowheads) cells in the LGE and cortical SVZ relative to the MGE. In 25 gw infants, clusters of non-proliferating Nkx2.1+Sox2 cells (shown by an oval outline) were surrounded by both proliferating and non-proliferating Sox2+ cells. Scale bar, 20 µm. (c) Data are mean ± SEM (n = 5 each group). The total Sox2+ cells were reduced in 29–35 gw relative to other groups in all 3 regions. *P < 0.05 for 29–35 versus 16–22, 23–25, and 26–28 gw in the MGE; #P < 0.05 for 29–35 versus 16–22, 23–25, and 26–28 gw in the LGE; P < 0.05 for 29–35 versus 16–22, 23–25, and 26–28 gw in the cortical SVZ. (d) Bar charts are mean ± SEM (n = 5 each group). P < 0.001, P < 0.001 for 16–22 versus 26–28 and 29–35 gw, respectively, in the MGE; αP < 0.001, βP < 0.001 for 16–22 versus 26–28 and 29–35 gw, respectively, in the LGE; φP < 0.001, ψP < 0.001 for 16–22 versus 26–28 and 29–35 gw, respectively, in the cortical SVZ. *P < 0.05, #P < 0.01 for MGE versus cortical SVZ for 16–22 and 23–25 gw, respectively. (e) Bar charts are mean ± SEM (n = 3–5 brains each group). Sox2+Nkx2.1+ were abundant in the MGE and relatively few in the LGE and cortical SVZ. *P < 0.001 MGE versus LGE and #P < 0.001 for MGE versus cortical SVZ for within gestational group comparison. P < 0.001 P < 0.001 for 16–22 gw versus 26–28 and 29–35 gw within the MGE; φP < 0.001, ψP < 0.001 for 29–35 gw versus 23–25 and 26–28 gw within the MGE.

Figure 4.

Sox2+ and Sox2+Nkx2.1+ cells were more abundant in the MGE relative to other germinal zones. (a) Representative triple labeling of coronal sections from the MGE of 20 and 25 gw subjects stained with Sox2-, Nkx2.1-, and Ki67-specific antibodies. Note abundance of Sox2+Ki67+ and Sox2+Nkx2.1+ cells in the VZ and SVZ of the MGE (lower magnification). Scale bar, 50 µm. (b) Typical triple immunolabeling of coronal sections from 20 and 25 gw subjects (higher magnification) using Sox2-, Nkx2.1-, and Ki67-specific antibodies. Note fewer Sox2+Nkx2.1+ (arrow) and Sox2+Ki67+ (arrowheads) cells in the LGE and cortical SVZ relative to the MGE. In 25 gw infants, clusters of non-proliferating Nkx2.1+Sox2 cells (shown by an oval outline) were surrounded by both proliferating and non-proliferating Sox2+ cells. Scale bar, 20 µm. (c) Data are mean ± SEM (n = 5 each group). The total Sox2+ cells were reduced in 29–35 gw relative to other groups in all 3 regions. *P < 0.05 for 29–35 versus 16–22, 23–25, and 26–28 gw in the MGE; #P < 0.05 for 29–35 versus 16–22, 23–25, and 26–28 gw in the LGE; P < 0.05 for 29–35 versus 16–22, 23–25, and 26–28 gw in the cortical SVZ. (d) Bar charts are mean ± SEM (n = 5 each group). P < 0.001, P < 0.001 for 16–22 versus 26–28 and 29–35 gw, respectively, in the MGE; αP < 0.001, βP < 0.001 for 16–22 versus 26–28 and 29–35 gw, respectively, in the LGE; φP < 0.001, ψP < 0.001 for 16–22 versus 26–28 and 29–35 gw, respectively, in the cortical SVZ. *P < 0.05, #P < 0.01 for MGE versus cortical SVZ for 16–22 and 23–25 gw, respectively. (e) Bar charts are mean ± SEM (n = 3–5 brains each group). Sox2+Nkx2.1+ were abundant in the MGE and relatively few in the LGE and cortical SVZ. *P < 0.001 MGE versus LGE and #P < 0.001 for MGE versus cortical SVZ for within gestational group comparison. P < 0.001 P < 0.001 for 16–22 gw versus 26–28 and 29–35 gw within the MGE; φP < 0.001, ψP < 0.001 for 29–35 gw versus 23–25 and 26–28 gw within the MGE.

Figure 5.

Sox2+Dlx2+ cells were more abundant in the MGE relative to other germinal zones, and Sox2 protein was elevated in the GE relative to the cortical SVZ. (a) Representative immunofluorescence of coronal sections from the MGE of a 20- and 25-gw subject labeled with antibodies to Sox2, Dlx2, and Ki67. Note abundance of Sox2+Ki67+ and Sox2+Dlx2+ cells and their distribution in the SVZ of the MGE (lower magnification). Scale bar, 50 µm. (b) Typical triple immunolabeling of coronal sections from the MGE, LGE, and cortical SVZ (higher magnification) for 20 and 25 gw subjects using Sox2-, Dlx2-, and Ki67-specific antibodies. Note fewer Sox2+Dlx2+ in the LGE and cortical SVZ relative to the MGE. Scale bar, 20 µm. (c) Bar charts are mean ± SEM (n = 5 each group). Sox2+ Dlx2+ cells were fewer in the LGE and cortical SVZ relative to the MGE. *P < 0.05 MGE versus cortical SVZ for 23–25 gw. P < 0.01, P < 0.01 for 16–22 and 23–25 gw versus 29–35 gw in the MGE. (d) Representative western blot analyses for Sox2 performed on homogenates of tissues from cortical SVZ and GE. Each lane represents one brain region from one brain. Bar charts are mean ± SEM (n = 7 brains each group). Data were normalized to β-actin. Sox2 protein levels were higher in the GE compared with the cortical SVZ. *P < 0.05 for cortical SVZ versus GE.

Figure 5.

Sox2+Dlx2+ cells were more abundant in the MGE relative to other germinal zones, and Sox2 protein was elevated in the GE relative to the cortical SVZ. (a) Representative immunofluorescence of coronal sections from the MGE of a 20- and 25-gw subject labeled with antibodies to Sox2, Dlx2, and Ki67. Note abundance of Sox2+Ki67+ and Sox2+Dlx2+ cells and their distribution in the SVZ of the MGE (lower magnification). Scale bar, 50 µm. (b) Typical triple immunolabeling of coronal sections from the MGE, LGE, and cortical SVZ (higher magnification) for 20 and 25 gw subjects using Sox2-, Dlx2-, and Ki67-specific antibodies. Note fewer Sox2+Dlx2+ in the LGE and cortical SVZ relative to the MGE. Scale bar, 20 µm. (c) Bar charts are mean ± SEM (n = 5 each group). Sox2+ Dlx2+ cells were fewer in the LGE and cortical SVZ relative to the MGE. *P < 0.05 MGE versus cortical SVZ for 23–25 gw. P < 0.01, P < 0.01 for 16–22 and 23–25 gw versus 29–35 gw in the MGE. (d) Representative western blot analyses for Sox2 performed on homogenates of tissues from cortical SVZ and GE. Each lane represents one brain region from one brain. Bar charts are mean ± SEM (n = 7 brains each group). Data were normalized to β-actin. Sox2 protein levels were higher in the GE compared with the cortical SVZ. *P < 0.05 for cortical SVZ versus GE.

Our analyses revealed that the Sox2+ cells were more abundant in subjects born before 28 gw compared with infants born after 28 gw in the VZ and SVZ of 3 brain regions—MGE, LGE, and cortical SVZ. Accordingly, the density of total (cycling and non-cycling) Sox2+ cells was reduced in 29–35 gw compared with 16–22, 23–25, and 26–28 gw subjects in 3 germinal regions—MGE (P < 0.01 all), LGE (P < 0.001, 0.046, and 0.05, respectively), and cortical SVZ (P < 0.001, 0.03, and 0.01, respectively; Fig. 4c). Total Sox2+ cells were higher in density in the MGE compared with the LGE and cortical SVZ, when subjects of all gestational groups were combined together (16–35 gw; P < 0.001, both).

We next evaluated cycling Sox2+ progenitor cells (Figs 4 and 5). Comparisons among gestational groups showed that cycling Sox2+ cells were more abundant in 16–22 gw fetuses compared with 26–28 and 29–35 gw preterm infants in the MGE, LGE, or cortical SVZ (P < 0.001 all, Fig. 4d). The proliferating Sox2+ cells were also higher in density in 16–22 gw fetuses relative to 23–25 gw infants in the LGE and cortical SVZ (P < 0.01 and0.02), but not in the MGE. Comparisons between germinal regions revealed that cycling Sox2+ cells were higher in density in the MGE relative to the LGE and cortical SVZ (P < 0.001 and 0.002), when analyses were performed after combining subjects of all gestational groups.

To assess Sox2 protein levels in the germinal regions, we performed western blot analyses on homogenates made of tissues dissected from the cortical SVZ and GE. The GE samples were a composite of MGE and LGE. We found that Sox2 protein levels were higher in the GE relative to the cortical SVZ (P = 0.013; n = 7 each group, Fig. 5d). Taken together, Sox2 precursor cells exhibit diverse distribution patterns in the 3 germinal regions; they are higher in density in the MGE relative to the LGE and cortical SVZ; and their abundance reduces with increasing gestational age in 3 germinal regions.

Sox2+ Progenitor Cells Co-expressing Nkx2.1 and Dlx2 Were Predominantly Located in the MGE

Sox2+, Nkx2.1+, and Dlx2+ progenitors were predominantly located in the MGE. Hence, we next asked whether Sox2+ cells co-express Nkx2.1 and Dlx2 transcription factors in the germinal regions and what was the comparative density of Nkx2.1+Sox2+ and Dlx2+Sox2+ progenitors. We found that Nkx2.1+Sox2+ cells were abundant in the MGE, but relatively few in the LGE and cortical SVZ (Fig. 4a,b,e). The total number of Nkx2.1+Sox2+ cells (/mm2) in the MGE was significantly higher in 16–22 gw fetuses compared with 26–28 and 29–35 gw infants (P < 0.001 both). These progenitors were also higher in 23–25 and 26–28 gw relative to 29–35 gw infants in the MGE (P < 0.001 both, Fig. 4e). Comparison between the 3 germinal regions showed that Nkx2.1+Sox2+ progenitors were higher in density in the MGE relative to the LGE and cortical SVZ for all gestational groups (P < 0.001), except for 29–35 gw.

We next quantified Dlx2+Sox2+ cells in the VZ and SVZ of the 3 germinal zones. Similar to Nkx2.1+Sox2+ progenitors, Dlx2+Sox2+ cells were abundant in the MGE and reduced in density as a function of increasing gestational age (P < 0.001, Fig. 5c). The density of Dlx2+Sox2+ cells in the MGE was higher in 16–22 and 23–25 gw subjects compared with 29–35 gw infants (P = 0.009 and 0.007, respectively). In addition, these progenitors were more plentiful in the MGE relative to the cortical SVZ for 23–25 gw infants (P = 0.02), but not for other gestational groups. Comparisons between the MGE and LGE were insignificant for individual gestational groups. However, a comparison after combining subjects of all gestational groups showed that Dlx2+Sox2+ cells were more abundant in the MGE relative to the LGE and cortical SVZ (P = 0.017 and <0.001). We next compared the density of Nkx2.1+Sox2+ precursors versus Dlx2+Sox2+ progenitors in the 3 germinal regions. Dlx2+Sox2+ cells were higher in number compared with Nkx2.1+Sox2+ cells in the LGE and cortical SVZ (P < 0.001 and 0.01, respectively), but not in the MGE.

Collectively, both Nkx2.1+Sox2+ and Dlx2+Sox2+ progenitors were predominantly located in the MGE, suggesting that the MGE is the principal source of interneuron progenitors and that Nkx2.1+ and Dlx2+ progenitors are derived from Sox2+ precursor cells. Moreover, greater abundance of Dlx2+Sox2+ cells, compared with Nkx2.1+Sox2+ progenitors in the LGE and cortical SVZ, indicates that the expression of Nkx2.1 transcription factor in Sox2+ cells are likely downregulated earlier than that of Dlx2.

Coup-TFII+ and Dlx2+ Progenitors Were Abundant in the CGE

The Coup-TFII transcription factor is important for migration of cortical interneurons in human fetal brains from the CGE to the neocortex and thus, is highly expressed in the CGE (Reinchisi et al. 2012). Here, we quantified total and cycling Coup-TFII+ and Dlx2+ cells in the ventral proliferative zone of the CGE in 5 groups of human subjects including 16–22, 23–25, 26–28, 29–35, and 36–40 gw (n = 3–5 each; total 18 subjects; Table 1 and Fig. 6). Our analyses revealed that both total and proliferating Coup-TFII+ cells were abundant in the VZ and SVZ of the CGE and that they reduced in density with advancing gestational age (P < 0.05), disappearing by 40 gw (Fig. 6c). Comparison between gestational groups revealed that total (cycling plus non-cycling) Coup-TFII+ cells were higher in density in 16–22 and 23–25 gw, compared with the older 29–35 (P = 0.02 and 0.009, respectively) and 36–40 gw infants (P = 0.008 and 0.003 respectively; Fig. 6c). Likewise, cycling Coup-TFII+ cells were more numerous in 23–25 gw compared with 29–35 and 36–40 gw subjects (P = 0.009 and 0.005, respectively).

Figure 6.

CGE exhibited Coup-TFII and Dlx2 proliferation, which reduced with advancing gestational age. (a) Representative immunofluorescence of coronal sections from the CGE of 17 and 24 gw subjects labeled with antibodies to Coup-TFII and Ki67. Shown in the second and fourth image, above and to the right, are orthogonal views (as in Figs 2 and 3), depicting Ki67-stained nucleus embedded within the Coup -TFII+-stained cells (arrow) in 17 and 24 gw infants. Proliferating cells are shown by arrowheads. Scale bars, 20 µm. (b) Typical immunofluorescence of cryosections from the MGE of 17 and 24 gw subjects labeled with antibodies to Dlx2 and Coup-TFII. Shown in the second and fourth micrograph of the lower panel, above and to the right of the image, are orthogonal views depicting complete overlap of Dlx2 and Coup-TFII-stained cells (arrows) in the CGE of 17 and 24 gw subjects. Proliferating cells are shown by arrowheads. (c) Bar charts are mean ± SEM (n = 3–5 each group). Coup-TFII+ and Coup-TFII+Ki67+ cells reduced as a function of increasing gestational age. *P < 0.05, #P < 0.01 for 16–22 versus 29–35 and 35–40 gw, respectively. P < 0.01, P < 0.01 for 23–25 versus 29–35 and 36–40 gw, respectively. ψP < 0.01, φP < 0.01 for 16–22 versus 29–35 and 36–40 gw, respectively. (d) Bar charts are mean ± SEM (n = 3–5 each group). *P < 0.05, #P < 0.01 for 16–22 versus 29–35 and 35–40 gw, respectively. P < 0.01 Dlx2+ cells for 23–25 versus 36–40 gw, respectively. ψP < 0.01, φP < 0.01 for 23–25 versus 29–35 and 36–40 gw, respectively. αP < 0.05 for 16–22 versus 36–40 gw. (e) Coup-TFII+Dlx2+ cells diminish with advance in gestational age. ψP < 0.01, φP < 0.05 for 16–22 versus 29–35 and 36–40 gw, respectively. αP < 0.05 for 23–25 versus 36–40 gw.

Figure 6.

CGE exhibited Coup-TFII and Dlx2 proliferation, which reduced with advancing gestational age. (a) Representative immunofluorescence of coronal sections from the CGE of 17 and 24 gw subjects labeled with antibodies to Coup-TFII and Ki67. Shown in the second and fourth image, above and to the right, are orthogonal views (as in Figs 2 and 3), depicting Ki67-stained nucleus embedded within the Coup -TFII+-stained cells (arrow) in 17 and 24 gw infants. Proliferating cells are shown by arrowheads. Scale bars, 20 µm. (b) Typical immunofluorescence of cryosections from the MGE of 17 and 24 gw subjects labeled with antibodies to Dlx2 and Coup-TFII. Shown in the second and fourth micrograph of the lower panel, above and to the right of the image, are orthogonal views depicting complete overlap of Dlx2 and Coup-TFII-stained cells (arrows) in the CGE of 17 and 24 gw subjects. Proliferating cells are shown by arrowheads. (c) Bar charts are mean ± SEM (n = 3–5 each group). Coup-TFII+ and Coup-TFII+Ki67+ cells reduced as a function of increasing gestational age. *P < 0.05, #P < 0.01 for 16–22 versus 29–35 and 35–40 gw, respectively. P < 0.01, P < 0.01 for 23–25 versus 29–35 and 36–40 gw, respectively. ψP < 0.01, φP < 0.01 for 16–22 versus 29–35 and 36–40 gw, respectively. (d) Bar charts are mean ± SEM (n = 3–5 each group). *P < 0.05, #P < 0.01 for 16–22 versus 29–35 and 35–40 gw, respectively. P < 0.01 Dlx2+ cells for 23–25 versus 36–40 gw, respectively. ψP < 0.01, φP < 0.01 for 23–25 versus 29–35 and 36–40 gw, respectively. αP < 0.05 for 16–22 versus 36–40 gw. (e) Coup-TFII+Dlx2+ cells diminish with advance in gestational age. ψP < 0.01, φP < 0.05 for 16–22 versus 29–35 and 36–40 gw, respectively. αP < 0.05 for 23–25 versus 36–40 gw.

Similar to Coup-TFII+ cells, total and cycling Dlx2+ cells also diminished in number with increasing gestational age, disappearing by 40 gw (P < 0.05, Fig. 6d). Comparisons between groups showed that the density of total Dlx2+ cells was higher in 16–22 gw compared with 29–35 and 36–40 gw infants (P = 0.023 and 0.003; Fig. 6d). Moreover, Dlx2+ cells were also higher in density in 23–25 gw compared with 35–40 gw infants. Cycling Dlx2+ cells were more abundant in 23–25 gw compared with 29–35 and 36–40 gw infants. We next examined a subset of cells that co-expressed Coup-TFII and Dlx2. These double-labeled cells were higher in density in 16–22 gw fetuses compared with 29–35 and 36–40 gw infants (P = 0.03 and 0.06; Fig. 6e). In addition, these progenitors were also higher in number in 23–25 gw infants compared with 36–40 gw infants (P = 0.04).

We next compared total and cycling Dlx2+ cells between the 2 regions—MGE and CGE. The density of total Dlx2+ cells was comparable between these 2 brain regions. However, cycling Dlx2+ cells were higher in density in the CGE compared with MGE (P = 0.005), when subjects of all gestational groups were combined. In addition, cycling Dlx2+ cells were higher in density in the CGE compared with the MGE for 23–25 and 26–28 gw infants (P = 0.014 and 0.039), but not for other gestational groups. Our observations indicate that the CGE is a major contributor of the Dlx2+ progenitors in premature infants.

Pan-Dlx, Nkx2.1, and Mash1 Proteins Are Enriched in GE, But Not in Cortical SVZ or Other Brain Regions

GABAergic neurogenesis is regulated by the coordinated action of transcription factors, including Nkx2.1, Dlx2, Lhx6, Mash1, Gsh2, and Sox6 (Kelsom and Lu 2013). Therefore, we assayed Nkx2.1, Dlx2, Lhx6, Mash1, Gsh2, and Sox6 mRNA expression in tissue samples from the GE compared with white matter (intermediate zone) and cortex (cortical plate) by real-time qPCR (n = 4 each gestational group, Supplementary Fig. 6). In addition, Nkx2.1, Dlx2, and Mash1 protein levels were quantified by western blot analyses (n = 4 each group, Fig. 7). GE samples used were a composite of MGE and LGE as described above. We found that mRNA expression of Dlx2 was higher in the GE compared with the cortex and white matter, when comparison was performed by combining subjects in all gestational ages (P < 0.001 all; Fig. 7a and Supplementary Fig. 6a). Dlx2 mRNA was also elevated in the GE relative to the cortex and white matter for individual gestational groups (P < 0.05 all), except for 29–35 gw infants. Pan-Dlx protein expression was also higher in the GE relative to the cortex and white matter for 16–22 and 23–25 gw subjects (P < 0.001 all), but not for 29–35 gw infants (Fig. 7a).

Figure 7.

Pan-Dlx, Nkx2.1, and Mash1 protein levels were higher in the GE compared with the cortex and white matter; these molecules were more abundant in the GE compared with the cortical SVZ. (a) Representative western blot analyses of Pan-Dlx, Nkx2.1, and Mash1 performed on homogenates form tissues taken from GE, cortex (cortical plate), and white matter (WM, intermediate zone). Rat brain (P5) was used as a positive control. Pan-Dlx, Nkx2.1, and Mash1 were predominantly expressed in the GE and weakly in the cortex and white matter. Each lane represents one brain region. Bar charts are mean ± SEM (n = 4 brains each gestational group). Data normalized to β-actin. *P < 0.001 **P < 0.01 for cortex versus GE within the gestational group; #P < 0.001 ##P < 0.01 for WM versus GE within the gestational group. P < 0.001, ††P < 0.05 for 16–22 versus 29–35 gw in the GE; P < 0.001 for 23–25 versus 29–35 gw in the GE; ψP < 0.05 for 26–28 versus 29–35 gw in the GE. φP < 0.05 for 23–25 versus 26–28 gw in the GE; αP < 0.001 for 26–28 versus 23–25 gw in the GE; βP < 0.05 for 23–25 versus 29–35 gw in the GE. (b) Typical western blot analyses for Pan-Dlx, Nkx2.1, and Mash1 in the cortical SVZ and GE. Bar charts are mean ± SEM (n = 7 each group). Data normalized to β-actin.

Figure 7.

Pan-Dlx, Nkx2.1, and Mash1 protein levels were higher in the GE compared with the cortex and white matter; these molecules were more abundant in the GE compared with the cortical SVZ. (a) Representative western blot analyses of Pan-Dlx, Nkx2.1, and Mash1 performed on homogenates form tissues taken from GE, cortex (cortical plate), and white matter (WM, intermediate zone). Rat brain (P5) was used as a positive control. Pan-Dlx, Nkx2.1, and Mash1 were predominantly expressed in the GE and weakly in the cortex and white matter. Each lane represents one brain region. Bar charts are mean ± SEM (n = 4 brains each gestational group). Data normalized to β-actin. *P < 0.001 **P < 0.01 for cortex versus GE within the gestational group; #P < 0.001 ##P < 0.01 for WM versus GE within the gestational group. P < 0.001, ††P < 0.05 for 16–22 versus 29–35 gw in the GE; P < 0.001 for 23–25 versus 29–35 gw in the GE; ψP < 0.05 for 26–28 versus 29–35 gw in the GE. φP < 0.05 for 23–25 versus 26–28 gw in the GE; αP < 0.001 for 26–28 versus 23–25 gw in the GE; βP < 0.05 for 23–25 versus 29–35 gw in the GE. (b) Typical western blot analyses for Pan-Dlx, Nkx2.1, and Mash1 in the cortical SVZ and GE. Bar charts are mean ± SEM (n = 7 each group). Data normalized to β-actin.

Similar to Dlx2, both mRNA and protein expression of Nkx2.1 were significantly higher in the GE compared with the cortex and white matter on a combined analysis of subjects of all gestational ages (P < 0.001 all; Supplementary Fig. 6b and Fig. 7a). Nkx2.1 gene expression was also higher in the GE relative to the cortex and white matter for all gestational groups (P < 0.05 all), except for 16–22 gw fetuses (Supplementary Fig. 6b). Nkx2.1 protein levels were elevated in the GE relative to the cortex and white matter for 16–22 (P = 0.004 and 0.008) and 23–25 gw (P < 0.001 both, Fig. 7a) infants, but not for 26–28 and 29–35 gw neonates.

Likewise, protein and mRNA expression of Mash1 were higher in the GE compared with the cortex and white matter on a combined analysis of subjects of all gestational ages (P < 0.001 all). Mash1 gene expression was also higher in the GE relative to the cortex for 16–22 and 23–25 gw infants (P = 0.001 and 0.0.26, respectively), but not to the white matter (Supplementary Fig. 6c). Mash1 protein was elevated in the GE compared with the cortex and white matter for 16–22 (P = 0.001 both) and 23–25 gw subjects (P = 0.01 and 0.022), but not for 26–28 and 29–35 gw infants (Fig. 7a).

Similar to Nkx2.1 and Dlx2, mRNA expression of Lhx6 was higher in the GE compared with the cortex and white matter for 23–25 (P = 0.02 both) and 26–28 gw (P = 0.01 and 0.02) infants, but not for 16–22 and 29–35 gw subjects (Supplementary Fig. 6d). Likewise, Sox6 and Gsh2 mRNA expression were elevated in the GE relative to the cortex and white matter (P < 0.001 and 0.002 for Sox6; P = 0.016 and 0.03 for Gsh2; Supplementary Fig. 6e,f) on analyses of all subjects together.

Western blot analyses also revealed that Nkx2.1, pan-Dlx, and Mash1 protein levels were significantly higher in the GE compared with the cortical SVZ (P = 0.021, 0.045, and 0.031, respectively; Fig. 7b). Taken together, GE has higher expression of Nkx2.1, Dlx, and Mash1 transcription factors compared with cortical SVZ, cortical plate, and white matter. This suggests that Nkx2.1, Dlx, and Mash1 are the important players of interneuron neurogenesis.

GABAergic Interneurons Were More Numerous in the Cortical SVZ than in the LGE and MGE

Since postmitotic GABAergic interneurons migrate out of the MGE to travel tangentially through the LGE and cortical SVZ in order to settle into the cortical layers (Kelsom and Lu 2013; Marin 2013; Southwell et al. 2014), we postulated that GABA+ interneurons might be sparser in the MGE compared with the LGE and cortical SVZ. To this end, we triple-labeled brain sections with GABA- and Nkx2.1 or Dlx2-specific antibodies along with DAPI (Fig. 8a and Supplementary Fig. 7). We noted that GABA+ cells were more numerous in the cortical SVZ relative to the MGE and LGE in all 4 gestational groups (P < 0.001 all comparisons, Fig. 8b). Intriguingly, GABA+ cells were almost absent in the VZ of the MGE, while they were abundant in the VZ of the cortical SVZ and LGE. GABA+ cells exhibited unipolar and multipolar processes with substantial intercellular space in the cortical SVZ (Fig. 8a). Conversely, GABA+ cells in the MGE were densely packed, confluent, and lacking processes. GABA+ cells in the LGE were similar to those in the cortical SVZ. GABA+ cells were also present throughout the intermediate zone, subplate, and cortical plate of the neocortex, which morphologically resembled GABA+ cells of the cortical SVZ. GABA+ cells in the basal ganglia—caudate, globus pallidus, and putamen—were relatively sparse (Supplementary Fig. 8). These findings suggest that GABA+ cells are immature in the MGE, but mature and migratory in the cortical SVZ, LGE, and embryonic intermediate zone.

Figure 8.

GABAergic cells were higher in number in the cortical SVZ compared with the MGE and LGE. (a) Representative immunostaining of coronal cyrosections from 17 and 24 gw subjects with GABA-specific antibody. Lower panel images are high magnification of the boxed area in the upper panel. Note more abundant GABA+ cells in the cortical SVZ (cSVZ) relative to the MGE. GABA+ cells were almost absent in the VZ of the MGE. In the cSVZ, GABA+ cells exhibited processes (arrowheads), and there were separated by intercellular spaces. Conversely, GABA+ cells were densely-packed, confluent, and lacking process in the MGE (arrows). Scale bar, 50 µm (upper panels), 20 µm (lower panels). (b) Bar charts are mean ± SEM (n = 5 each). The density of GABA+ cells was higher in the cSVZ compared with MGE and LGE for all ages, except for 19–35 gw. *P < 0.001, #P < 0.001 for MGE versus cSVZ and LGE versus cSVZ within the gestational group, respectively. P < 0.05, P < 0.05 16–22 versus 29–35 within the cSVZ and LGE, respectively. (c) Bar charts are mean ± SEM (n = 5 each group). The percentages of GABAergic cells (density of GABA+/density of DAPI+) were also elevated in the cSVZ compared with the MGE and cSVZ. *P < 0.001, #P < 0.001 for MGE versus cSVZ and LGE versus cSVZ within the gestational group, respectively. P < 0.05 16–22 versus 29–35.

Figure 8.

GABAergic cells were higher in number in the cortical SVZ compared with the MGE and LGE. (a) Representative immunostaining of coronal cyrosections from 17 and 24 gw subjects with GABA-specific antibody. Lower panel images are high magnification of the boxed area in the upper panel. Note more abundant GABA+ cells in the cortical SVZ (cSVZ) relative to the MGE. GABA+ cells were almost absent in the VZ of the MGE. In the cSVZ, GABA+ cells exhibited processes (arrowheads), and there were separated by intercellular spaces. Conversely, GABA+ cells were densely-packed, confluent, and lacking process in the MGE (arrows). Scale bar, 50 µm (upper panels), 20 µm (lower panels). (b) Bar charts are mean ± SEM (n = 5 each). The density of GABA+ cells was higher in the cSVZ compared with MGE and LGE for all ages, except for 19–35 gw. *P < 0.001, #P < 0.001 for MGE versus cSVZ and LGE versus cSVZ within the gestational group, respectively. P < 0.05, P < 0.05 16–22 versus 29–35 within the cSVZ and LGE, respectively. (c) Bar charts are mean ± SEM (n = 5 each group). The percentages of GABAergic cells (density of GABA+/density of DAPI+) were also elevated in the cSVZ compared with the MGE and cSVZ. *P < 0.001, #P < 0.001 for MGE versus cSVZ and LGE versus cSVZ within the gestational group, respectively. P < 0.05 16–22 versus 29–35.

As the gestational age increased, GABA+ cells reduced in density within the LGE and cortical SVZ (P < 0.05 each), but not in the MGE. Consistent with these findings, GABA+ cells were fewer in 26–28 and 29–35 gw infants compared with 16–22 gw fetuses in the LGE (P < 0.05 and 0.03, respectively). Similarly in the cortical SVZ, GABA+ cells were diminished in density in 29–35 gw relative to 16–22 gw neonates (P = 0.003). The percentage of GABA+ interneurons (total GABA+ cells/nuclear count) displayed a similar trend as those of the absolute number of GABA+ cells (Fig. 8c).

We next evaluated GABA+ interneurons that co-expressed Nkx2.1 or Dlx2 transcription factors (Supplementary Fig. 7). We found that the percentage of GABA+ cells expressing Nkx2.1 (ratio of Nkx2.1+ GABA+ and total GABA+ cells) was higher in the MGE compared with the LGE and cortical SVZ for all gestational ages (P < 0.001 all), except for 29–35 gw (Supplementary Fig. 7a,c). Similarly, the percentage of GABA+ cells expressing Dlx2 (ratio of Dlx2+GABA+ and total GABA+ cells) was higher in the MGE relative to the LGE and cortical SVZ for all gestational ages (P < 0.001 all groups, except P < 0.05 for 29–35 gw; Supplementary Fig. 7b,d). Collectively, abundance of GABAergic neurons in the germinal regions reduces with fetal maturation; GABA+ interneurons are relatively immature in the MGE compared with the GABAergic neurons in the LGE and cortical SVZ.

Interneurons Generated in the MGE Migrated to the Cortical Plate

GABA+ interneurons were more immature in the MGE and relatively mature and migratory (exhibit processes) in the LGE and cortical SVZ. Thus, to determine the regional distribution of migratory GABA+ cells in the forebrain, we double-labeled the sections with GABA and doublecortin antibodies. Although, we have not quantified doublecortin- and GABA-co-labeled cells in these brain regions, our examination of several brain sections (n = 2 each group) revealed that the majority of GABA+ cells throughout the cortical SVZ and LGE co-expressed doublecortin. However, a substantial number of GABA+ cells, particularly confluent ones who were lacking processes, were not reactive for doublecortin in the MGE (Supplementary Fig. 9a). We found few doublecortin- and GABA-co-labeled cells in the striatum (Supplementary Fig. 8), but a large number of them were present in the cortical SVZ, intermediate zone, and subplate of the neocortex (Supplementary Fig. 9b). Taken together, our observations suggest that cortical interneurons are produced in the MGE which migrate through the LGE, cortical SVZ, and intermediate zone before reaching neocortical mantle.

Discussion

Determining the origin of cortical interneurons and mapping the period of GABAergic neurogenesis during human gestation are fundamental to understanding cortical development and the effect of perinatal insults and genetic mutations on the production of interneurons. This information is also crucial in elucidating the pathogenesis of a number of neuropsychiatric disorders and designing cell-based therapies. The present study suggests that the MGE and CGE were the primary source of cortical interneurons, and that GABAergic neurogenesis continued nearly to the end of human gestation. In addition, a lower density and a relative immaturity of GABA+ interneurons in the MGE compared with other germinal regions suggested that interneurons originating from the MGE progressively mature as they migrate through the LGE and cortical SVZ to finally settle in the neocortex.

Our results demonstrated that there is protracted generation of cortical interneurons during the third trimester of pregnancy. We found the presence of cycling and non-cycling interneuron progenitors in the MGE and CGE during the third trimester; however, their density was substantially reduced after the second trimester. Our finding is consistent with the previous report that cell proliferation is substantially exhausted in the VZ/SVZ of the GE by the last trimester of human gestation (Zecevic et al. 2011). In rodents as well, immature, motile, and Pax6+ interneurons have been observed in the dorsal white matter during early postnatal period (P1–7 mouse), which are likely to be derived from CGE and destined to settle in the olfactory bulb (Riccio et al. 2012). In our previous study, we showed that glutamatergic neurogenesis continues until 28 gw (Malik et al. 2013). Hence, GABAergic progenitors remain in the MGE several weeks longer than the glutamatergic progenitors in the cortical SVZ. Persistence of neurogenesis in the third trimester has enormous clinical implications. Events in the third trimester in utero (e.g., hypoxia–ischemia, chorioamnionitis, and hemorrhage) as well as neonatal complications of prematurity (e.g., intraventricular hemorrhage, neonatal hypoxia, and sepsis) might adversely affect GABAergic neurogenesis (Komitova et al. 2013), which may predispose the survivors to neurological and psychiatric disorders. Indeed, prematurity and associated complications are linked with cognitive disabilities, mental retardation, learning disabilities, neurodevelopmental delay, and psychiatric disorders (Indredavik et al. 2010; Whitaker et al. 2011). Moreover, adolescents born extremely preterm are at high risk for inattention and hyperactivity, emotional disturbance, social incompetence, poor educational achievement, and lower intellectual abilities (Ornstein et al. 1991; Taylor et al. 1995; Botting et al. 1997; Costeloe et al. 2000; Wood et al. 2000; Chan et al. 2001; Anderson and Doyle 2003; Anderson and Doyle 2004; Vanhaesebrouck et al. 2004; Delobel-Ayoub et al. 2009). Accordingly, subtle differences of in utero environment, reflected in the birth weight variation within monozygotic twins, are associated with significant alterations in brain anatomy and cognitive functions, which persist even into early adulthood (Petanjek and Kostovic 2012; Raznahan et al. 2012). Taken together, persistence of interneuron neurogenesis in the third trimester underscores the need to determine how neonatal complications of prematurity adversely impact GABAergic neurogenesis and how this can be minimized.

Our results demonstrated that the MGE is the principal source of GABAergic interneurons in humans both in the second and third trimester of pregnancy. We quantified total and cycling Nkx2.1+ and Dlx2+ progenitors in the MGE, LGE, and the cortical SVZ. We found a paucity of both cycling and non-cycling Nkx2.1+ and Dlx2+ cells in the cortical SVZ and LGE relative to the MGE, suggesting that the MGE is a principal site of GABAergic neurogenesis in humans. In addition, significantly higher protein levels of Nkx2.1, Dlx2, and Mash1 in the GE compared with the cortical SVZ (19–36 gw, Fig. 7) reinforced our conclusion that the MGE was the main locus of interneuron neurogenesis in humans, similar to rodents (Wonders and Anderson 2006; Miyoshi et al. 2010). While the GE as the primary source of interneurons in primates is becoming accepted, the proportion of interneurons contributed by the cortical VZ/SVZ has been debated. Two recent studies noted that the cortical SVZ is small contributor to the pool of cortical interneurons as these studies noted a paucity of proliferating interneuronal progenitors, including Nkx2.1+ and Dlx2+ cells, in the cortical SVZ (Hansen et al. 2013; Ma et al. 2013). These studies are limited by a relatively small sample size and lack samples from preterm infants born in the third trimester of pregnancy. In contrast, several studies in humans and primates have claimed that the cortical VZ/SVZ are important source of cortical interneurons (Letinic et al. 2002; Petanjek et al. 2009; Jakovcevski et al. 2011; Zecevic et al. 2011; Reinchisi et al. 2012; Radonjic et al. 2014). A recent study has shown that 2.6–5% of cells in the VZ and 6–9% of cells in the SVZ of the dorsal telencephalon in human fetuses (14–22 gw) were Nkx2.1+ and about half of these cells were proliferating (Radonjic et al. 2014). Moreover, higher expression of Dlx genes and several GABAergic markers have also been reported in the dorsal pallium compared with the GE in human fetuses (8–12 gw), which has also strengthened the notion that the cortical VZ/SVZ is an important source of interneurons (Al-Jaberi et al. 2015). In the present study, which examined fetuses and preterm infants of 16–35 gw, 0.4–1.1% of cells in the VZ/SVZ of the dorsal telencephalon were Nkx2.1+ and they were rarely proliferating. These results highlight the differences between the studies on the origin of human cortical interneurons from the cortical VZ/SVZ. This discrepancy in findings between 2 sets of studies assessing the origin of interneurons can be ascribed to technical limitations, including lack of specificity of markers, differences in the immunohistochemical methods, and the selection of cortical areas for the study. The merits of the present study, which supports the cortical SVZ to be a relatively minor contributor of interneurons, are large sample size, inclusion of subjects born in the third trimester, and corroborating evidence from 3 distinct methodologies—immunohistochemistry, western blot analyses, and RT-qPCR. One limitation of our study is a lack of evaluation of the cortical SVZ in brain regions other than at the level of the head of caudate nucleus (MGE, LGE, and cSVZ) and mid-thalamus (CGE). It is possible that the cortical SVZ of the anterior and posterior cerebrum might give rise to cortical interneurons, which will reinforce the findings of a number of previous studies in human and non-human primates (Letinic et al. 2002; Petanjek et al. 2009; Jakovcevski et al. 2011; Zecevic et al. 2011)

The CGE accounts for approximately 30% of GABAergic cortical neurons, and interneuron generation in the CGE outlasts the MGE in rodents (Miyoshi et al. 2010). Consistent with rodent studies, the present human study showed similar density of total Dlx2+ cells in the CGE and MGE and significantly a higher number of cycling Dlx2+ cells in the CGE relative to the MGE. In addition, these progenitors persisted beyond 35 gw in the CGE, largely disappearing by 40 gw. This suggests that both the CGE and MGE are the principal source of GABAergic interneurons in humans. In agreement with our studies, previous reports in human fetuses have estimated that the CGE is a major contributor of cortical interneurons (Hansen et al. 2013; Ma et al. 2013).

The present study demonstrated that the periventricular compartment was layered both in mid and late pregnancy. It consisted of VZ and SVZ in all 3 germinal regions—MGE, LGE, and cortical SVZ; both the VZ and SVZ regressed markedly after 25 gw. The reduction in thickness was seen in all 3 regions in almost a similar manner, which reflected that neurogenesis and the related neuronal migration in these regions slows down in all 3 regions with increased gestation. Previous studies have reported a reduction in the width of the MGE and LGE between 10 and 14 gw, which is consistent with our report (Hansen et al. 2013). The outer SVZ of the MGE exhibited clustering of progenitors around the radial glia during the second trimester. A similar patterning of progenitors has been reported in the previous study using 10–14 gw infants (Hansen et al. 2013). A number of our observations suggested that the cortical interneurons were generated in the MGE and travel through the LGE and cortical SVZ before finally settling in the cortical layers: (1) Both total and cycling interneuronal precursors were most abundant in the MGE, relatively sparse in the LGE, and almost absent in the cortical SVZ, (2) immature GABA+ interneurons (confluent cells lacking processes) were observed in the MGE, whereas mature (heading and tailing processes) GABA+ cells were noted in the LGE and cortical SVZ; GABA+ cells were more abundant in the cortical SVZ relative to the MGE, (3) Nkx2.1+ and Dlx2+ cells sparsely populated the striatum, suggesting that a small fraction of interneurons is conceivably migrating to the basal ganglia, (4) migratory GABA+ cells (doublecortin+) were almost absent in the VZ and relatively sparse in the SVZ of the MGE, whereas they were abundant throughout the LGE and cortical SVZ, and (5) mature and migratory GABA+ cells were found in the intermediate zone and subplate. Since little is known about migration of interneurons in late gestation and in postnatal development (Morozov and Freund 2003; Sonego et al. 2013), it is premature to comment on the complexity of the migratory process of interneurons through the white matter and deep layers of the cerebral cortex during the third trimester.

The present study demonstrated that Sox2+ precursor cells were abundant in all 3 germinal zones. However, they were more prevalent in the MGE relative to the LGE and cortical SVZ. Moreover, Nkx2.1+Sox2+ and Dlx2+Sox2+ progenitors were predominantly located in the MGE. This suggests that Sox2+ progenitors in the MGE are the primary source of interneuron progenitors. Importantly, Sox2+ radial glia cells in the cortical SVZ give rise to glutamatergic neurons and these precursor cells in the LGE produce striatal projection neurons and olfactory interneurons (Deacon et al. 1994; Olsson et al. 1995; Wichterle et al. 2001). The presence of Sox2 reactivity on Nkx2.1+ and Dlx2+ interneuronal progenitors in the LGE and cortical SVZ implies that the Sox2 expression persists in the intermediate progenitors harboring the germinal regions, which is consistent with previous studies (Ghashghaei et al. 2006; Ferri et al. 2013). We found Dlx2+Sox2+ cells to be more abundant compared with Nkx2.1+Sox2+ cells in the LGE, which suggests that Dlx2 expression may be retained by these progenitors for a longer time relative to Nkx2.1 expression. Indeed, the Dlx2 transcription factor plays a key role in migration of interneurons (Anderson et al. 1997) and promotes interneuron neurogenesis over oligodendrogenesis (Petryniak et al. 2007).

In conclusion, this study highlights that cortical interneurons are produced in the MGE and CGE in humans, just as in rodents and other primates. Although a small percentage of interneuron progenitors are present in the cortical SVZ, this might have roles to play in the higher cortical function. Importantly, generation of cortical interneurons continues until the end of human pregnancy, although it steadily declines with increasing gestational age. Hence, this raises concerns over in utero perinatal insults and neonatal complications of premature infants, which can potentially affect these critical biological processes that shape the human cerebral cortex.

Supplementary Material

Supplementary material can be found at: http://www.cercor.oxfordjournals.org/.

Funding

This work was supported by NIH-NINDS Grants RO1 NS071263 (P.B.), R21NS085508 (P.B.), R01NS083947 (P.B.), and a scientist development grant from the American Heart Association (G.V.).

Notes

We thank Joanne Abrahams for the assistance with images. Conflict of Interest: None declared.

References

Al-Jaberi
N
,
Lindsay
S
,
Sarma
S
,
Bayatti
N
,
Clowry
GJ
.
2015
.
The early fetal development of human neocortical GABAergic interneurons
.
Cereb Cortex
 .
25
:
631
645
.
Anderson
P
,
Doyle
LW
.
2003
.
Neurobehavioral outcomes of school-age children born extremely low birth weight or very preterm in the 1990s
.
JAMA
 .
289
:
3264
3272
.
Anderson
PJ
,
Doyle
LW
.
2004
.
Executive functioning in school-aged children who were born very preterm or with extremely low birth weight in the 1990s
.
Pediatrics
 .
114
:
50
57
.
Anderson
SA
,
Eisenstat
DD
,
Shi
L
,
Rubenstein
JL
.
1997
.
Interneuron migration from basal forebrain to neocortex: dependence on Dlx genes
.
Science
 .
278
:
474
476
.
Botting
N
,
Powls
A
,
Cooke
RW
,
Marlow
N
.
1997
.
Attention deficit hyperactivity disorders and other psychiatric outcomes in very low birthweight children at 12 years
.
J Child Psychol Psychiatry
 .
38
:
931
941
.
Chan
K
,
Ohlsson
A
,
Synnes
A
,
Lee
DS
,
Chien
LY
,
Lee
SK
.
2001
.
Survival, morbidity, and resource use of infants of 25 weeks’ gestational age or less
.
Am J Obstet Gynecol
 .
185
:
220
226
.
Clowry
GJ
.
2014
.
An enhanced role and expanded developmental origins for gamma-aminobutyric acidergic interneurons in the human cerebral cortex
.
J Anat
 .
Advance Access published May 20, doi: 10.1111/joa.12198.
Clowry
G
,
Molnar
Z
,
Rakic
P
.
2010
.
Renewed focus on the developing human neocortex
.
J Anat
 .
217
:
276
288
.
Corbin
JG
,
Gaiano
N
,
Machold
RP
,
Langston
A
,
Fishell
G
.
2000
.
The Gsh2 homeodomain gene controls multiple aspects of telencephalic development
.
Development
 .
127
:
5007
5020
.
Costeloe
K
,
Hennessy
E
,
Gibson
AT
,
Marlow
N
,
Wilkinson
AR
.
2000
.
The EPICure study: outcomes to discharge from hospital for infants born at the threshold of viability
.
Pediatrics
 .
106
:
659
671
.
Deacon
TW
,
Pakzaban
P
,
Isacson
O
.
1994
.
The lateral ganglionic eminence is the origin of cells committed to striatal phenotypes: neural transplantation and developmental evidence
.
Brain Res
 .
668
:
211
219
.
Delobel-Ayoub
M
,
Arnaud
C
,
White-Koning
M
,
Casper
C
,
Pierrat
V
,
Garel
M
,
Burguet
A
,
Roze
JC
,
Matis
J
,
Picaud
JC
et al
.
2009
.
Behavioral problems and cognitive performance at 5 years of age after very preterm birth: the EPIPAGE Study
.
Pediatrics
 .
123
:
1485
1492
.
Feliciano
DM
,
Bordey
A
.
2013
.
Newborn cortical neurons: only for neonates?
Trends Neurosci
 .
36
:
51
61
.
Ferri
A
,
Favaro
R
,
Beccari
L
,
Bertolini
J
,
Mercurio
S
,
Nieto-Lopez
F
,
Verzeroli
C
,
La Regina
F
,
De Pietri Tonelli
D
,
Ottolenghi
S
et al
.
2013
.
Sox2 is required for embryonic development of the ventral telencephalon through the activation of the ventral determinants Nkx2.1 and Shh
.
Development
 .
140
:
1250
1261
.
Gallo
V
,
Deneen
B
.
2014
.
Glial development: the crossroads of regeneration and repair in the CNS
.
Neuron
 .
83
:
283
308
.
Ghashghaei
HT
,
Weber
J
,
Pevny
L
,
Schmid
R
,
Schwab
MH
,
Lloyd
KC
,
Eisenstat
DD
,
Lai
C
,
Anton
ES
.
2006
.
The role of neuregulin-ErbB4 interactions on the proliferation and organization of cells in the subventricular zone
.
Proc Natl Acad Sci USA
 .
103
:
1930
1935
.
Hansen
DV
,
Lui
JH
,
Flandin
P
,
Yoshikawa
K
,
Rubenstein
JL
,
Alvarez-Buylla
A
,
Kriegstein
AR
.
2013
.
Non-epithelial stem cells and cortical interneuron production in the human ganglionic eminences
.
Nat Neurosci
 .
16
:
1576
1587
.
Hladnik
A
,
Dzaja
D
,
Darmopil
S
,
Jovanov-Milosevic
N
,
Petanjek
Z
.
2014
.
Spatio-temporal extension in site of origin for cortical calretinin neurons in primates
.
Front Neuroanat
 .
8
:
50
.
Indredavik
MS
,
Vik
T
,
Evensen
KA
,
Skranes
J
,
Taraldsen
G
,
Brubakk
AM
.
2010
.
Perinatal risk and psychiatric outcome in adolescents born preterm with very low birth weight or term small for gestational age
.
J Dev Behav Pediatr
 .
31
:
286
294
.
Jakovcevski
I
,
Mayer
N
,
Zecevic
N
.
2011
.
Multiple origins of human neocortical interneurons are supported by distinct expression of transcription factors
.
Cereb Cortex
 .
21
:
1771
1782
.
Kamei
Y
,
Inagaki
N
,
Nishizawa
M
,
Tsutsumi
O
,
Taketani
Y
,
Inagaki
M
.
1998
.
Visualization of mitotic radial glial lineage cells in the developing rat brain by Cdc2 kinase-phosphorylated vimentin
.
Glia
 .
23
:
191
199
.
Kanatani
S
,
Yozu
M
,
Tabata
H
,
Nakajima
K
.
2008
.
COUP-TFII is preferentially expressed in the caudal ganglionic eminence and is involved in the caudal migratory stream
.
J Neurosci
 .
28
:
13582
13591
.
Kelsom
C
,
Lu
W
.
2013
.
Development and specification of GABAergic cortical interneurons
.
Cell Biosci
 .
3
:
19
.
Kepecs
A
,
Fishell
G
.
2014
.
Interneuron cell types are fit to function
.
Nature
 .
505
:
318
326
.
Komitova
M
,
Xenos
D
,
Salmaso
N
,
Tran
KM
,
Brand
T
,
Schwartz
ML
,
Ment
L
,
Vaccarino
FM
.
2013
.
Hypoxia-induced developmental delays of inhibitory interneurons are reversed by environmental enrichment in the postnatal mouse forebrain
.
J Neurosci
 .
33
:
13375
13387
.
Letinic
K
,
Zoncu
R
,
Rakic
P
.
2002
.
Origin of GABAergic neurons in the human neocortex
.
Nature
 .
417
:
645
649
.
Long
JE
,
Cobos
I
,
Potter
GB
,
Rubenstein
JL
.
2009
.
Dlx1&2 and Mash1 transcription factors control MGE and CGE patterning and differentiation through parallel and overlapping pathways
.
Cereb Cortex
 .
19
(Suppl 1)
:
i96
106
.
Ma
T
,
Wang
C
,
Wang
L
,
Zhou
X
,
Tian
M
,
Zhang
Q
,
Zhang
Y
,
Li
J
,
Liu
Z
,
Cai
Y
et al
.
2013
.
Subcortical origins of human and monkey neocortical interneurons
.
Nat Neurosci
 .
16
:
1588
1597
.
Malik
S
,
Vinukonda
G
,
Vose
LR
,
Diamond
D
,
Bhimavarapu
BB
,
Hu
F
,
Zia
MT
,
Hevner
R
,
Zecevic
N
,
Ballabh
P
.
2013
.
Neurogenesis continues in the third trimester of pregnancy and is suppressed by premature birth
.
J Neurosci
 .
33
:
411
423
.
Marin
O
.
2013
.
Cellular and molecular mechanisms controlling the migration of neocortical interneurons
.
Eur J Neurosci
 .
38
:
2019
2029
.
Marin
O
.
2012
.
Interneuron dysfunction in psychiatric disorders
.
Nat Rev Neurosci
 .
13
:
107
120
.
Miyoshi
G
,
Hjerling-Leffler
J
,
Karayannis
T
,
Sousa
VH
,
Butt
SJ
,
Battiste
J
,
Johnson
JE
,
Machold
RP
,
Fishell
G
.
2010
.
Genetic fate mapping reveals that the caudal ganglionic eminence produces a large and diverse population of superficial cortical interneurons
.
J Neurosci
 .
30
:
1582
1594
.
Molnar
Z
,
Butt
SJ
.
2013
.
Best-laid schemes for interneuron origin of mice and men
.
Nat Neurosci
 .
16
:
1512
1514
.
Morozov
YM
,
Freund
TF
.
2003
.
Postnatal development and migration of cholecystokinin-immunoreactive interneurons in rat hippocampus
.
Neuroscience
 .
120
:
923
939
.
Nery
S
,
Fishell
G
,
Corbin
JG
.
2002
.
The caudal ganglionic eminence is a source of distinct cortical and subcortical cell populations
.
Nat Neurosci
 .
5
:
1279
1287
.
Olsson
M
,
Campbell
K
,
Wictorin
K
,
Bjorklund
A
.
1995
.
Projection neurons in fetal striatal transplants are predominantly derived from the lateral ganglionic eminence
.
Neuroscience
 .
69
:
1169
1182
.
Ornstein
M
,
Ohlsson
A
,
Edmonds
J
,
Asztalos
E
.
1991
.
Neonatal follow-up of very low birthweight/extremely low birthweight infants to school age: a critical overview
.
Acta Paediatr Scand
 .
80
:
741
748
.
Panganiban
G
,
Rubenstein
JL
.
2002
.
Developmental functions of the Distal-less/Dlx homeobox genes
.
Development
 .
129
:
4371
4386
.
Petanjek
Z
,
Berger
B
,
Esclapez
M
.
2009
.
Origins of cortical GABAergic neurons in the cynomolgus monkey
.
Cereb Cortex
 .
19
:
249
262
.
Petanjek
Z
,
Kostovic
I
.
2012
.
Epigenetic regulation of fetal brain development and neurocognitive outcome
.
Proc Natl Acad Sci USA
 .
109
:
11062
11063
.
Petryniak
MA
,
Potter
GB
,
Rowitch
DH
,
Rubenstein
JL
.
2007
.
Dlx1 and Dlx2 control neuronal versus oligodendroglial cell fate acquisition in the developing forebrain
.
Neuron
 .
55
:
417
433
.
Radonjic
NV
,
Ayoub
AE
,
Memi
F
,
Yu
X
,
Maroof
A
,
Jakovcevski
I
,
Anderson
SA
,
Rakic
P
,
Zecevic
N
.
2014
.
Diversity of cortical interneurons in primates: the role of the dorsal proliferative niche
.
Cell Rep
 .
9
:
2139
2151
.
Raznahan
A
,
Greenstein
D
,
Lee
NR
,
Clasen
LS
,
Giedd
JN
.
2012
.
Prenatal growth in humans and postnatal brain maturation into late adolescence
.
Proc Natl Acad Sci USA
 .
109
:
11366
11371
.
Reinchisi
G
,
Ijichi
K
,
Glidden
N
,
Jakovcevski
I
,
Zecevic
N
.
2012
.
COUP-TFII expressing interneurons in human fetal forebrain
.
Cereb Cortex
 .
22
:
2820
2830
.
Riccio
O
,
Murthy
S
,
Szabo
G
,
Vutskits
L
,
Kiss
JZ
,
Vitalis
T
,
Lebrand
C
,
Dayer
AG
.
2012
.
New pool of cortical interneuron precursors in the early postnatal dorsal white matter
.
Cereb Cortex
 .
22
:
86
98
.
Sanai
N
,
Nguyen
T
,
Ihrie
RA
,
Mirzadeh
Z
,
Tsai
HH
,
Wong
M
,
Gupta
N
,
Berger
MS
,
Huang
E
,
Garcia-Verdugo
JM
et al
.
2011
.
Corridors of migrating neurons in the human brain and their decline during infancy
.
Nature
 .
478
:
382
386
.
Sonego
M
,
Gajendra
S
,
Parsons
M
,
Ma
Y
,
Hobbs
C
,
Zentar
MP
,
Williams
G
,
Machesky
LM
,
Doherty
P
,
Lalli
G
.
2013
.
Fascin regulates the migration of subventricular zone-derived neuroblasts in the postnatal brain
.
J Neurosci
 .
33
:
12171
12185
.
Southwell
DG
,
Nicholas
CR
,
Basbaum
AI
,
Stryker
MP
,
Kriegstein
AR
,
Rubenstein
JL
,
Alvarez-Buylla
A
.
2014
.
Interneurons from embryonic development to cell-based therapy
.
Science
 .
344
:
1240622
.
Taylor
HG
,
Hack
M
,
Klein
N
,
Schatschneider
C
.
1995
.
Achievement in children with birth weights less than 750 grams with normal cognitive abilities: evidence for specific learning disabilities
.
J Pediatr Psychol
 .
20
:
703
719
.
Vanhaesebrouck
P
,
Allegaert
K
,
Bottu
J
,
Debauche
C
,
Devlieger
H
,
Docx
M
,
Francois
A
,
Haumont
D
,
Lombet
J
,
Rigo
J
et al
.
2004
.
The EPIBEL study: outcomes to discharge from hospital for extremely preterm infants in Belgium
.
Pediatrics
 .
114
:
663
675
.
Whitaker
AH
,
Feldman
JF
,
Lorenz
JM
,
McNicholas
F
,
Fisher
PW
,
Shen
S
,
Pinto-Martin
J
,
Shaffer
D
,
Paneth
N
.
2011
.
Neonatal head ultrasound abnormalities in preterm infants and adolescent psychiatric disorders
.
Arch Gen Psychiatry
 .
68
:
742
752
.
Wichterle
H
,
Turnbull
DH
,
Nery
S
,
Fishell
G
,
Alvarez-Buylla
A
.
2001
.
In utero fate mapping reveals distinct migratory pathways and fates of neurons born in the mammalian basal forebrain
.
Development
 .
128
:
3759
3771
.
Wonders
CP
,
Anderson
SA
.
2006
.
The origin and specification of cortical interneurons
.
Nat Rev Neurosci
 .
7
:
687
696
.
Wood
NS
,
Marlow
N
,
Costeloe
K
,
Gibson
AT
,
Wilkinson
AR
.
2000
.
Neurologic and developmental disability after extremely preterm birth. EPICure Study Group
.
N Engl J Med
 .
343
:
378
384
.
Zecevic
N
,
Hu
F
,
Jakovcevski
I
.
2011
.
Interneurons in the developing human neocortex
.
Dev Neurobiol
 .
71
:
18
33
.