Ferret brain possesses young interneuron collections equivalent to human postnatal migratory streams

Abstract The human early postnatal brain contains late migratory streams of immature interneurons that are directed to cortex and other focal brain regions. However, such migration is not observed in rodent brain, and whether other small animal models capture this aspect of human brain development is unclear. Here, we investigated whether the gyrencephalic ferret cortex possesses human‐equivalent postnatal streams of doublecortin positive (DCX+) young neurons. We mapped DCX+ cells in the brains of ferrets at P20 (analogous to human term gestation), P40, P65, and P90. In addition to the rostral migratory stream, we identified three populations of young neurons with migratory morphology at P20 oriented toward: (a) prefrontal cortex, (b) dorsal posterior sigmoid gyrus, and (c) occipital lobe. These three neuronal collections were all present at P20 and became extinguished by P90 (equivalent to human postnatal age 2 years). DCX+ cells in such collections all expressed GAD67, identifying them as interneurons, and they variously expressed the subtype markers SP8 and secretagogin (SCGN). SCGN+ interneurons appeared in thick sections to be oriented from white matter toward multiple cortical regions, and persistent SCGN‐expressing cells were observed in cortex. These findings indicate that ferret is a suitable animal model to study the human‐relevant process of late postnatal cortical interneuron integration into multiple regions of cortex.

babes, the brain is vulnerable to environmental insults, such as hypoxia-ischemia and infection, that can result in persistent cognitive deficits and increased risk of psychiatric disease (Cannon, Yolken, Buka, Torrey,, & Collaborative Study Group on the Perinatal Origins of Severe Psychiatric Disorders, 2008;Mittal, Ellman, & Cannon, 2008). Given the importance of this epoch, faithful animal models that mimic key processes of early postnatal human brain development are needed to investigate mechanisms of postnatal cortical circuit formation and the impact of injury.
Recent findings have revealed persistent postnatal streams of late migrating interneurons in human brain up to 12 months of age that are oriented toward specific cortical regions. Humans possess at least three postnatal streams of migrating interneurons; a rostral migratory stream (RMS) that supplies multiple subtypes of interneurons to the olfactory bulbs, which disappears by 12 months postnatally (Sanai et al., 2011), a medial migratory stream (MMS) that branches off of the RMS, putatively supplies calretinin+ interneurons to the ventromedial prefrontal cortex (VMPFC), which disappears by 12 months postnatally (Sanai et al., 2011), and a dorsal migratory "ARC" of cells that is present at the rostral tip of the lateral ventricles, putatively supplies multiple subtypes of interneurons to the anterior cingulate cortex (ACC) and superior frontal gyrus (SFG), and disappears by 5 months postnatally (Paredes et al., 2016).
Although embryonic migration of interneurons to cortex and postnatal migration of interneurons to olfactory bulb (RMA) has been characterized extensively in the mouse, the evidence for postnatal migration of interneurons to cortex is limited. Mice possess a population of postnatal migrating 5HT3a+ cells that exit the RMS up to postnatal Day 10 and terminate in layer six of the orbitofrontal cortex (Inta et al., 2008;Le Magueresse et al., 2011). However, the number of postnatal migrating cells is not as proportionally robust as seen in human and does not reflect human subtype diversity. Furthermore, there is no evidence that mice possess a dorsal migratory stream, seen in human. Finally, the mouse equivalency of human term is considered to be approximately postnatal Day 10, which is when these cells cease migration into the cortex. Therefore, it is unclear if what is observed in mice constitutes true postnatal migration in comparison with human as the phenomenon does not continue past the equivalency of term.
Recent evidence suggests that larger mammalian species, such as rabbits, pigs, sheep, and dolphins, may more closely resemble humans in terms of postnatal migration of interneurons (La Rosa et al., 2018;Luzzati et al., 2003;Morton et al., 2017, but whether this occurs in some small animal species is unclear.). Ferrets have been used as an animal model to study brain development for the past 30 years. Their brain size is significantly increased compared to rodents (Barnette et al., 2009;Raju et al., 2018). Like humans, and unlike rodents, they possess a gyrencephalic cortex (Barnette et al., 2009;Duque & McCormick, 2010;Johnson et al., 2018). Furthermore, they undergo extensive postnatal brain development (McConnell, 1985(McConnell, , 1988Morton et al., 2017). The ferret also captures other features specific to human brain, such as the presence of an outer sub-ventricular zone (oSVZ; Gertz, Lui, LaMonica, Wang, & Kriegstein, 2014; Martínez-Martínez et al., 2016;Reillo & Borrell, 2012). Based on these features, we hypothesized that ferrets would possess postnatal stream of interneurons that were more robust than those seen in rodents.

| Tissue processing
Ferret kits were terminally anesthetized using an isoflurane vaporizer and transcardially perfused with 0.1 M phosphate-buffered saline (PBS) at pH 7.4 followed by 4% paraformaldehyde (PFA) in PBS at pH 7.4. Brains were dissected from the skull and fixed in 4% PFA overnight at 4 C. After fixation, brains were cryoprotected in 30% sucrose in 0.1 M PBS until the brains sunk. For sectioning, brains were frozen to the stage of a sliding microtome (Leica SM2010R). Coronal, sagittal, and horizontal floating sections (50 μm) were cut. Cut sections were stored in 6 well cell culture plates in 0.1 M PBS with azide at 4 C and were used for free floating immunohistochemistry.

| Antibody characterization
All relevant antibody information and conditions can be found in Table 1.
The rabbit polyclonal anti-doublecortin antibody (Cell Signaling Technology, RRID:AB_561007) was characterized by the manufacturer in a Western blot analysis of fetal rat brain tissue lysate, yielding a single protein band at~45 kDa. We confirmed this in ferret by performing a Western blot analysis of postnatal Day 20 whole brain lysate and observing a similar band. Furthermore, in the current study this antibody specifically stained the cytoplasm of cells with a classic migratory morphology as expected.
The guinea pig polyclonal anti-doublecortin antibody (Millipore Bioscience Research, RRID:AB_1586992) was characterized by the manufacturer in a Western blot analysis of rat brain tissue lysate, yielding a single protein band at~45 kDa. We confirmed this in ferret by performing a Western blot analysis of postnatal Day 20 whole brain lysate and observing a similar band. Furthermore, in the current study this antibody specifically stained the cytoplasm of cells with a classic migratory morphology as expected.
The goat polyclonal anti-GAD67 antibody (Abcam, RRID: AB_ 1640532) was characterized by the manufacturer in Western blot analyses of both mouse and human brain lysate, yielding a single protein band of 70 kDa. We confirmed this in ferret by performing a Western blot analysis of postnatal Day 20 whole brain lysate and observing a similar band.
The mouse monoclonal anti-GFAP antibody (Sigma Aldrich, RRID: AB_477010) was characterized by the manufacturer by performing a Western blot analysis of whole extract of rat B35 cells, yielding a band of 50 kDa. We confirmed this in ferret by performing a Western blot analysis of postnatal Day 20 whole brain lysate and observing a similar band.
The mouse monoclonal anti-Olig2 antibody (gift from Chuck Stiles lab) was previously characterized by Lu et al., Neuron, 2000 and identifies oligodendrocytes. We further characterized this antibody in ferret by performing a Western blot analysis of postnatal Day 20 whole brain lysate and observing a band at 57 kDa, as expected.
The goat polyclonal anti-Sox10 antibody (Santa Cruz Biotechnology, RRID:AB_2195374) was characterized by the manufacturer by performing a Western blot analysis on both nontransfected and mouse and human recombinant transfected 293T whole cell lysates, yielding a single protein band at 55 kDa. We characterized this antibody in ferret by performing a Western blot analysis of postnatal Day 20 whole brain lysate and observing a similar band.
The rabbit polyclonal anti-Iba1 antibody (Wako Chemicals, RRID: AB_839504) was characterized by the manufacturer by performing a Western blot on isolated rat microglia, yielding an appropriate band at 17 kDa. 2.3 | Immunohistochemistry, confocal and epifluorescent microscopy, and image processing 2.3.1 | Staining All doublecortin, GAD67, SATB2, and TBR1 immunohistochemistry was performed using tyramide signal amplification. In short, freefloating sections (50 μm) were first mounted on glass slides and allowed to dry for 1 hr. Slides were then rinsed in PBS, and antigen retrieval was performed at 95 C in 0.01 M Na Citrate buffer, pH = 6. Durations of antigen retrieval can be found in Table 1. Following antigen retrieval, slides were washed with PBS-T (0.2% TX100 in PBS) for 15 min, incubated with 0.02 N HCl in PBS for 30 min at room temperature (RT), and then blocked with TNB solution (0.1 M Tris-HCl, pH 7.5, 0.15 M NaCl, 0.5% blocking reagent from PerkinElmer) for 1 hr at RT. Slides were incubated in primary antibodies overnight at 4 C (see Table 1). All antibodies were diluted in TNB solution. The following day, slides were rinsed in PBS-T for 1 hr at RT and then incubated with biotinylated secondary antibodies (1:500; Jackson Immunoresearch Laboratories), as well as any Alexa secondaries being used (1:750), for 1 hr at room temperature. Sections were then incubated for 30 min in streptavidinhorseradish peroxidase that was diluted (1:200) with TNB. Finally, sections were incubated in tyramide-conjugated fluoroscein for 5 min at 1:100. Slides were then covered with glass cover slips using DAPI Fluoromount-G (Thermo Fisher Scientific).
Primary antibodies that did not require tyramide signal amplification were stained in the following manner. Free floating sections (50 μm) were first mounted on glass slides and allowed to dry for 1 hr. Slides were then rinsed in PBS, and antigen retrieval was performed at 95 C in 0.01 M Na Citrate buffer, pH = 6. Durations of antigen retrieval can be found in Table 1

| Confocal and epifluorescent microscopy
All confocal images were taken using a Leica Sp8 confocal microscope (Leica Microsystems, Wetzlar, Germany). Images were acquired as 2um thick optical sections.
Tile scans used for mapping doublecortin signal were taken using a Zeiss Axioimager (Carl Zeiss, Oberkochen, Germany).

| Image processing
Lif files were covered to tif files using ImageJ. All image files were imported into Photoshop CS6 and adjusted for brightness and contrast only.
Tile scans of DCX stained P20 ferret brain sections in the sagittal, coronal, and horizontal planes were uploaded into Adobe Illustrator

| DCX co-localizations
Confocal images were taken of the MMS in the sagittal plane at the indicated ages and were loaded into ImageJ. Percent co-localizations were calculated by counting the number of DCX+ cell bodies per image and dividing by the number of co-localized cells. Three sections from each of three animals were counted.

| White matter cells and caspase+ cells
Sagittal sections of the MMS were stained with either secretagogin or cleaved caspase 3 at the indicated ages and were visually inspected using a confocal microscope. SCGN+ cells in the white matter with a mature, differentiated morphology were counted manually due to low density. All SCGN+ cells in the MMS were included in each count.
Cleaved caspase 3+ cells in the white matter were similarly counted manually. All positive cells in the MMS were included in each count.
Three sections from each of three animals were counted. Student's ttests were performed to determine significance using Prism version 6, Graphpad.

| Tissue clearing and staining with iDISCO+
The iDISCO+ protocol for clearing thick tissue sections was performed as described (Renier et al., 2016  Higher magnification images were taken of each sample using the ×10 objective on a spinning disk microscope (Yokagawa CSU22).

| Image processing
All image and movie files were processed and snapshots taken using Imaris v.7.22. Brightness and contrast were the only attributes adjusted.

| RESULTS
3.1 | Postnatal DCX+ cell collections in the ferret brain diminish with increasing postnatal age To study the distribution of migratory neurons in the ferret brain, we began by staining for the immature neuronal marker doublecortin (DCX) at postnatal Day 20 (P20). At this age, the ferret brain is equivalent to the term human brain in its percentage of maximum brain volume (Figure 1a Using percent maximum brain weight, ferret brain development was aligned with human to determine comparative ages. At P20, ferrets are 30% of their maximum brain weight.  (Figure 3g,h). We also observed that as MBP staining increased from P20 to P65, individual DCX+ cells in the stream condensed into clusters in the white matter. These findings indicate that the ferret possesses postnatal young neuron collections similar to human RMS and MMS.

| A young neuron collection oriented toward postnatal ferret dorsomedial cortex
We next studied other gyri containing DCX+ migratory neurons at P20 to see if these populations disappeared at a uniform rate. At P20 there is a prominent dorsal population of DCX+ cells in the white matter of the posterior sigmoid gyrus (PSG; Figure 4). Individual cells with migratory morphology were observed in the white matter of the PSG oriented away from the lateral ventricle and toward the cortex (Figure 4a). A large DCX+ cluster was also observed adjacent to the dorsolateral ventricle ( Figure 4b). In contrast to the MMS, we did not detect discrete chains of DCX+ cells exiting the dense DCX+ cluster at the dorsolateral ventricles.
In the coronal plane, the same stream was found to occupy the white matter tracts of both the PSG and the ASG. (Figure 4c). By P40, the individual cells in this region had already become rare. A few scattered DCX+ cells were observed, but by P65 they had disappeared completely  (Figure 5f).
By P90 we no longer detected them in the posterior cortex. The DCX+ cluster began to decrease in size after P20 and had disappeared completely by P65 (Figure 5d). We next investigated the types of cells that comprise these late postnatal populations in the ferret.

| Postnatal DCX+ cells with migratory morphology are young interneurons
To further characterize the DCX+ cells, we stained them for markers of neurons and glial cells (Figure 6). At P20, cells in each stream co-expressed GAD67 and DCX to a very high degree, suggesting that they are likely young inhibitory interneurons (Figure 6a,b). DCX+ cells in each stream also expressed Sp8, further suggesting they are interneurons ( Figure 6c). Each stream did not co-express DCX and Sp8 to the same degree, with the MMS expressing the most and the DMS expressing the least (Figure 6d). Interestingly, DCX+ cells were also found to express SCGN, a calcium binding protein that is expressed in human cortical interneurons but not rodent cortical interneurons (Raju et al., 2018; Figure 6e). The degree of co-localization between DCX and SCGN was almost identical to that of DCX and Sp8 ( Figure 6f). This led us to triple stain both individual cells and clusters for further verification of their identity (Figure 6g,h). Clusters observed in P20 sagittal sections of the MMS were positive for DCX, SCGN, and Sp8 to some degree. Finally, we wanted assess whether any DCX+ cells expressed glial markers, as has been shown (Boulanger & Messier, 2017;Liu et al., 2018). DCX+ cells did not express markers of glial cells, such as GFAP (Figure 6i), a marker of

| SCGN+ cells appear to transition from the white matter to the cortex
The finding of SCGN+ interneurons oriented toward the prefrontal cortex in ferret provided the unique opportunity to investigate possible contributions to developing cortical circuits. We first stained sagittal sections from P20, P40, P65 and adult ferret with SCGN and looked specifically at the MMS, the stream with the highest degree of co-expression between DCX and SCGN (Figure 7a). In P20 sections, we were able to observe individual cells with migratory morphology appearing to transition from the white matter stream into the cortex. Cells continued to be observed in this state in P40 and P65 (Figure 7b,c), although the number decreased as predicted by both the DCX maps and quantification. However, considering the density of cells in the MMS, too few cells were observed outside the stream using single section immunohistochemistry to consider any kind of robust cortical contribution. Therefore, we next attempted to clear large tissue sections in order to more accurately visualize how many cells might be transitioning into the cortex.
Using the iDISCO protocol, we cleared 2 mm sections of P20 ferret brain tissue, stained them with SCGN, and imaged them using a light sheet microscope (Figure 7d-k). The number of SCGN+ cells in the white matter made apparent by using iDISCO was significantly greater than that seen using traditional immunohistochemistry. Furthermore, we were able to better visualize the number of SCGN+ cells in the cortex itself. Examining P20 sagittal sections (Figure 7d-

| Cortical SCGN-expressing interneurons survive into adulthood and express NeuN
In order to better characterize the fate of P20 SCGN+ cortical cells, we stained sagittal sections with SCGN at P20, P40, P65, and adult ( Figure 8a-d). Focusing on the ASG, that is, the possible destination of the MMS in the prefrontal cortex, we observed that SCGN+ cells in the cortex were more numerous, and possessed smaller soma, at P20 and P40 (Figure 8a,b). By P65, SCGN+ cells were less numerous and had larger soma, and were observed with multiple neurites (Figure 8c).

SCGN+ cortical cells continued to be observed into adulthood and
were also observed with longer and more branched neurites 3.8 | SCGN+ cells exist in the white matter into adulthood SCGN+ cells were observed in the white matter at all ages studied.
Starting at P20, but most apparent at P65, cells could be observed with a mature, nonmigratory, morphology, that is, a large soma and multiple neurites (Figure 9a). These cells also did not co-express DCX.
The number of these mature looking cells continued to increase in the white matter between P20 and P90 (Figure 9b). Many of them expressed the marker calretinin (Figure 9c). To see if these cells underwent apoptosis as part of their fate, cleaved caspase 3 staining was performed. Cleaved caspase 3 signal did not co-localize with mature white matter SCGN+ cells (Figure 9d). Furthermore, as the number of white matter SCGN+ cells increased significantly from P20 to P90, the number of observed Caspase+ cells decreased significantly ( Figure 9e), suggesting that the majority of these cells will survive in the white matter.

| DISCUSSION
The ferret has been used as a model for studying CNS development for many years due to its gyrencephalic cortex and extended postnatal brain development (Chapman, Zahs, & Stryker, 1991); (Finney, Stone, & Shatz, 1998); (Penn, Riquelme, Feller, & Shatz, 1998  2016), the postnatal ferret brain contains robust streams of cortically oriented young interneurons. We also find evidence for a novel stream oriented toward the posterior cortex that has not been described in human. Finally, we find persistent expression of SCGN, both in interneurons oriented toward, as well as in, the cortex, a finding which has been observed in human, but not rodents. Persistent SCGN expression supports the view that such cells permanently invest the cortex and potentially make functional contributions.

| Ferrets possess multiple postnatal collections of young interneurons
Ferrets contain a rostromedial collection of young interneurons that branches off the RMS and is oriented toward the anterior sigmoid gyrus and the orbital gyrus, that is, the ferret prefrontal cortex.
Between P20 and P40, the ferret RMS-like structure condenses, and cells cluster together. This stream then continues past P65, until disappearing by P90. All of these characteristics are similar to the reported MMS in human (Sanai et al., 2011). In human between term and 12 months of age, the equivalency of P90 ferret, a MMS of neurons branches off the RMS and supplies calretinin+ interneurons to the ventromedial prefrontal cortex. The stream also transitions from being widely dispersed at earlier ages to being tightly condensed clusters at later stages.
Ferrets also contain a dorsal collection of young interneurons that potentially originate at a cluster of Dcx+ cells surrounding the midpoint of the lateral ventricles. Their origin is inferred based on their proximity to these clusters, as well as similar timing in the decrease in number of cells in both areas. The stream is contained within the white matter and is oriented radially toward the posterior sigmoid gyrus. The posterior sigmoid gyrus is thought to be equivalent to human anterior cingulate gyrus, a midline region that is a part of the default mode network and is correlated with an individual's ability to mentalize (Apps & Ramnani, 2014). It was recently shown that the human ACC continued to receive migrating interneurons until the first 6 months of postnatal life (Paredes et al., 2016). These interneurons differentiated into a variety of subtypes with the majority of them expressing neuropeptide Y and calretinin. In the ferret brain, we did not observe the expression of neuropeptide Y, calretinin, PV, or SST in DCX+ cells in this stream (data not shown). We did observe Sp8 and SCGN expression, which may provide clues as to the subtypes these cells may ultimately become.
Finally, we found that ferrets contain a posterior collection of young interneurons that appears to originate at a cluster of

| Characteristics of DCX+ cells in putative migratory streams
Young interneurons in the putative ferret postnatal streams express several markers associated with an origin in the ventral pallium (Ma et al., 2012;Raju et al., 2018). We found they express Sp8 and SCGN to varying degrees, two markers of interneurons shown to be expressed in cells originating in the caudal ganglionic eminence (CGE) and lateral ganglionic eminence (LGE). The DCX+/SCGN+ cells do not express markers or transcription factors associated with the medial ganglionic eminence, such as PV, SST, or Nkx2.1 (data not shown).
Interestingly, we found that nearly all of the late migrating DCX+ cells expressed SCGN, a calcium binding protein (Mulder et al., 2009;Rogstam et al., 2007). Previously, secretagogin had been observed in neurons in multiple structures in the rodent brain, such as the RMS, olfactory bulb, hypothalamus, dentate gyrus and amygdala, but not cortex (Kosaka & Kosaka, 2013;Mulder et al., 2009Mulder et al., , 2010Romanov et al., 2015). Recently, it was shown that human cortex contains SCGN+ interneurons, and forced expression in mice increases neurite outgrowth (Raju et al., 2018). The discovery that the ferret also contains SCGN+ interneurons, both in the white matter and cortex, that persist into adulthood, implies that SCGN+ interneurons may represent a change in cortical development found only in gyrencephalic brains whose function and circuit contributions can be further studied using ferret as a model.
Future work is needed to determine the fate of postnatal interneurons in ferret. In the human, the DMS (ARC) and MMS appear to contribute neuropeptide Y and calretinin+ interneurons to developing cortical circuits in the ACC and VMPFC, respectively. In the ferret, interneurons in postnatal streams are cortically oriented. We found that they persist until various time points depending on region and appear to be transitioning from white matter to cortex. Furthermore, persistent cortical SCGN expression, observed into adulthood, suggests that SCGN+ cells from the streams that appear to be exiting the white matter could potentially invest the cortex. In human, SCGN+ cortical interneurons are also observed in the upper layers. We observed SCGN+ cells in both upper and lower cortical layers. However, the cells in the upper layers were more organized into a laminar distribution and co-localized with NeuN. In contrast, SCGN+ cells in the lower layers were more randomly distributed and did not colocalize with NeuN. This could imply that cells in the lower layers are migrating into the upper layers. It could also imply that these cells are late-arriving and have yet to mature and express NeuN.
SCGN+cells were also observed throughout development in the white matter, where they develop extensive neurite branching and express calretinin. Persistent, functional, white matter interneurons have been reported in rodents (von Engelhardt, Khrulev, Eliava, Wahlster, & Monyer, 2011) and both young and older rhesus monkeys (Mortazavi, Wang, Rosene, & Rockland, 2016). Some of the interneurons in the white matter may undergo apoptosis. It has previously been reported that up to 30% of interneurons undergo apoptosis (Southwell et al., 2012). However, caspase staining did not correlate with SCGN as the number of mature SCGN+ white matter interneurons increased while the number of caspase+ white matter cells decreased. If the majority of SCGN+ cells were undergoing apoptosis, it would be expected that the number would better correlated with the pattern of caspase+ cells.
The finding of putative streams of DCX+ interneurons throughout the developing ferret brain suggests that this may be a common developmental phenomenon in some species, possibly related to brain size or gyrification. Perinatal injuries might therefore disproportionately affect the circuits that may be supplied by these cells. A continued addition of late postnatal interneurons would undoubtedly affect the excitatory-inhibitory balance of those circuits. We conclude that the extended postnatal development of the ferret CNS offers a suitable model to study postnatal interneurons for possible migration and integration into existing circuits. Further studies are needed to test roles for late postnatal interneurons and how their migration and or circuit integration could be impacted by brain injury.