All animal experimentation procedures were approved by the Ethics Committee for Animal Experimentation (CEEA) of the University of Barcelona. All protocols were accepted by the Department of Environment and Housing of the Generalitat de Catalunya with the license number 11126, date of approval 24/5/2021, and the procedure CEEA number OB 340/19 SJD. The method of IUGR induction was previously described in Eixarch et al. (2009). Briefly, IUGR was induced at the 25th gestational day (GD) in pregnant New Zealand rabbits by surgical ligature of 40–50% of the uteroplacental vessels of each gestational sac of one uterine horn, while the contralateral horn was left for normal growth. Caesarean section was carried out at GD30 to obtain IUGR and control pups.

On the day of IUGR induction pregnant rabbit mothers were assigned to 4 different groups: without (w/o) administration, or with administration of MEL, DHA, or LF. The therapies were daily administered to the pregnant rabbit by releasing the solution with a syringe in the throat from the day of IUGR induction (GD 25) until the day of caesarean section (GD 30). Specific doses were determined as followed: MEL (10 mg /kg bw/day), DHA (37 mg/kg bw/day) and Lf (166 mg/kg bw/day). We refer to Kühne et al. (2022) for a detailed description about selection of in vivo doses, calculations and suppliers. For all treatment groups, the inclusion criteria for postnatal day 0 (PND0) IUGR pups was a birth weight lower and for control pups higher than the 25th percentile (39.7 g, Barenys et al., 2021). The number and birth weight of PND0 rabbit pups from each group was equal to the animals described in Kühne et al. (2022). Briefly 12 control and 10 IUGR pups from the group w/o were included from 8 rabbit mothers, 2 control and 2 IUGR rabbit pups were included from two different rabbit mothers for each treatment group.

From one rabbit pup’s whole brain, at least four independent experiments were performed.

The in vitro neurosphere culture was generated directly after decapitation at PND0, as described in Pla et al. (2022). Briefly, neural progenitor cells (NPCs) were isolated from rabbits’ whole brains by dissection, mechanical dissociation, digestion (20 min incubation with 20 U/ml papain [Worthington #LS003124] at 37°C), mechanical homogenization into a cell suspension, and centrifugation (10 min at 1200 rpm). The cell pellet obtained was resuspended in 1 ml of freezing medium (1:1; volume of pellet: volume of freezing medium [consisting in 70% (v/v) proliferation medium, 20% (v/v) fetal calf serum (FCS [Serva #1192002]), and 10% (v/v) DMSO]) and immediately stored at −80°C. Each cryo-vial was thawed by brief immersion in a 37°C water bath, and cells were transferred to 15 ml of proliferation medium preconditioned at 37°C and 5% CO₂ for 2 h, and gentle resuspension. The cell suspension was centrifuged (10 min, 1,200 rpm), supernatant discarded and cells transferred to Poly-HEMA [Sigma #192066] coated dishes filled with proliferation medium [consisting in DMEM [Gibco #10569010] and Hams F12 [Gibco #31765027] 3:1 supplemented with 2% B27 [Gibco #17504044], and 20 ng/ml human recombinant epidermal growth factor (EGF [Gibco #PHG0313]) and recombinant human fibroblast growth factor 2 (FGF2 [R&D systems #233-FB]), Penicillin–Streptomycin (10,000 U/ml) [Gibco #15140122] supplemented with Rho kinase (ROCK) inhibitor Y-276322 [Tocris #1254] at a final concentration of 10 μM. Half of the volume of proliferation medium per petri dish was exchanged every 2–3 days by proliferation medium without ROCK inhibitor.

IUGR and control brains derived neurospheres formed for 11 days in proliferation medium were always cultured in parallel. Two days before starting experiments, proliferating neurospheres were mechanically chopped to a size of 0.2 mm (McIlwain tissue chopper) to ensure homogeneous neurosphere size and spherical shape. Neurospheres were not chopped more than once. On the experiment plating day, 0.3 mm diameter neurospheres were selected and transferred in 8-chamber slides (Falcon #354118) previously coated with laminin [Sigma #L2020] and poly-D-lysin (PDL [Sigma #P0899]) containing differentiation medium [consisting in DMEM and Hams F12 3:1 supplemented with N2 [Gibco #17502048], Penicillin–Streptomycin (10,000 U/ml)]. The medium of 7-, and 14-day experiments were supplemented with 1% FCS. Half of the medium was renewed every 2–3 days. NPCs plated on a laminin/PDL coated surface radially migrated out of the sphere core and differentiated into effector cells. Each chamber contained five (5-day experiment) or six (7-and 14-day experiment) neurospheres representing replicates within one experiment, and at least three independent experiments were performed for every endpoint and exposure condition.

Compounds for neuroprotective therapy testing were dissolved in their corresponding vehicle depending on their maximum solubility (Table 1) and subsequently in differentiation medium. The effect of the potential therapies was assessed after 5 days of differentiation.

After 5, 7, or 14 days under differentiation conditions, neurospheres were fixed with paraformaldehyd (PFA) 4% for 30 min at 37°C, washed twice with PBS and stored in PBS until immunostained.

The endpoints “% of neurons,” “number of dendrites per neuron,” and “neurite length” were analyzed after 5, 7, and 14 days of differentiation, whereas “pre-synaptic formation” was only assessed after at time point 14 days (Figure 1). EGF [20 ng/ml] was used as a positive control for neuronal differentiation. Neurospheres were fixed, immunocytostained, and image analysis was carried out by taking two images of each migration area with a BX61 microscope (Olympus, Japan) with “UPlanFl 10x/0.30 Ph1” objective lens and using the ImageJ/Fiji 1.53q software. The number of nuclei (Hoechst staining) representing the total amount of cells was automatically counted using ImageJ/Fiji 1.53q software and neurons (βIII-tubulin+ cells) were manually counted. To determine the % of neurons the number of neurons was normalized to the number of nuclei.

The number of dendrites/neuron and their distances from the soma (neurite length) were manually measured by using ImageJ 1.53q after 5 days of differentiation. After 7 and 14 days, the number of dendrites/neuron and neurite length were assessed with the “Sholl analysis” of the ImageJ/Fiji 1.53q blinded to the experimental groups (L.G.V. and B.A.K.). The tool “Sholl analysis” was used to trace manually the different paths of dendrites calculating dendritic branching and length, as described in detail in Pla et al. (2020). The 8-bit tracing was constructed by using the Fiji plugins “Segmentation” and “Simple Neurite Tracer” as described in Binley et al. (2014). The total amount of dendrites was classified in primary, secondary, or tertiary dendrites depending on their point of division. The primary dendrites are born from the soma, the secondary ones from the primaries and so on. After 14 days the ability to generate pre-synapses was analyzed. Pre-synaptic puncta were defined as Synapsin-1 (pre-synaptic marker) co-localized with βIII-Tubulin+ cells. Pre-synaptic formation (synapsin-1+ neurons [%]) was determined by the number of neurons with synapsin-1+ puncta normalized by the total number of neurons. Analysis was evaluated in 5–6 neurospheres/condition, minimum 10 neurons/neurosphere in at least 3 independent experiments.

The cell viability was assessed with the CellTiter-Blue® cell viability assay (Promega #G8081). This assay is based on the measurement of mitochondrial reductase activity of living cells by conversion of resazurin to the fluorescent product resorufin. After 2 h of incubation with the reagent (1:3), medium was placed in a 96-well plate and read with FLUOstar Optima microplate reader. Neurospheres exposed to 10% DMSO (2 h) were used as lysis control.

Statistical analysis was performed using GraphPad Prism v9. The difference between two samples was calculated with a two-tailed paired student’s t-test. Concentration-dependent effects were analyzed using one-way ANOVA. Time-course experiments including the comparisons of more than two groups were assessed by performing a two-way ANOVA. One-way and two-way ANOVA analysis was always followed by post-hoc test Bonferroni’s multiple comparison test. The respective statistical analysis is mentioned in each figure legend. The significance threshold was established at *p ≤ 0.05.

Neurospheres were prepared from 12 control and 10 IUGR rabbit pups with a significantly reduced body weight in the IUGR group as described in Kühne et al. (2022). Previous results from our group testing the impact of IUGR in the rabbit neurosphere model after 3 days in culture determined no difference between control and IUGR on the endpoint “% neurons” (Barenys et al., 2021). In the current study, we investigated the impact of IUGR on neuronal endpoints after 5 days in vitro to unravel if changes may occur at a later time point but we confirmed that the percentage of neurons was also not significantly different between control and IUGR at this time point (Figures 2A,B, Control: 2.25 ± 0.39% vs. IUGR: 2.34 ± 0.25%, p = 0.800). The positive control EGF significantly decreased the % of neurons in both groups (Figure 2B, Control: 0.21 ± 0.08%; IUGR: 0.15 ± 0.04%), and significantly increased the metabolic activity in comparison to the solvent control indicating a proliferative effect as expected for this growth factor (Supplementary Figure S1A) proving that the system is flexible and can react to external stimuli known to keep cells in a proliferating instead of in a differentiating status. The endpoint “neurite length” was measured by the distance from the soma to the neurite end (Figure 2C), and the “number of dendrites per neuron” revealed the degree of dendritic arborization (Figure 2D). After 5 days in culture, neurons of control, and IUGR neurospheres developed mainly primary dendrites and significantly fewer secondary dendrites per neuron (1.39 control primary vs. 0.12 control secondary dendrites/neuron, p < 0.0001; Figure 2D). Importantly, the total neurite length was significantly increased in IUGR neurospheres compared to the respective control value (Figure 2C, total control: 29.82 ± 2.84 vs. IUGR: 36.03 ± 3.46 μm, p = 0.011). This difference was due to the significantly larger primary neurites in IUGR neurospheres (Figure 2C, primary control: 28.50 ± 2.71 vs. IUGR: 34.81 ± 3.48 μm, p = 0.006), and not to differences in the secondary neurite length (Figure 2C, secondary control: 1.36 ± 0.35 vs. IUGR: 1.19 ± 0.16 μm, p = 0.943). The number of dendrites per neuron did not vary between control and IUGR after 5 days, neither the total number nor the number of primary or secondary dendrites (Figure 2D, total control: 1.51 ± 0.09 vs. IUGR: 1.44 ± 0.05 dendrites/neuron, p = 0.353).

We established for the very first time the maintenance of rabbit control and IUGR neurospheres under long-term differentiation conditions [maximum time in culture previously described was 5 days in vitro (Barenys et al., 2021)] and examined the changes in neuronal morphology or network formation over time (Figure 3). Representative pictures of neurospheres cultured for 7 and 14 days in vitro revealed that both groups (IUGR and control) continued developing βIII-tubulin+ cells. After 7 days both neurosphere groups, control and IUGR, did not develop the pre-synaptic marker synapsin-1 (Supplementary Figures S2), but after 14 days they presented pre-synaptic formation (Figures 3A,B). In a time-course experiment including the time points 5, 7, and 14 days, a significant interaction between time and cases (control and IUGR) was observed (p = 0.0143). The percentage of βIII-tubulin+ cells was only slightly increased in the IUGR group compared to the control after 7 days. But remarkably, after 14 days, the percentage of neurons in IUGR neurospheres exceeded to a significant extent the percentage of neurons in the control group (Figure 3C, Control: 4.43 ± 0.98% vs. IUGR: 8.82 ± 2.61%, p = 0.005). By using this time-course approach we discovered that the neuronal differentiation rate is significantly faster in IUGR compared to control (Figure 3C, difference between the slopes of control and IUGR: p = 0.013), without affecting the viability at any time point indicating a specific effect on neuronal development by excluding a cytotoxic effect (Supplementary Figures S1B,C). The total neurite length and number of dendrites per neuron increased over time in both groups, control and IUGR, demonstrating a more complex neuronal morphology with extended neurites and more branched dendrites from 5 to 14 days (Figures 3D,E). After 14 days, the percentage of neurons developing pre-synaptic puncta (synapsin-1+ neurons) was measured (Figure 3B IV–V arrows, Figure 3F). However, no significant difference in the presynaptic formation between control and IUGR was discovered (Figure 3F, Control: 15.23 ± 9.97% vs. IUGR: 27.45 ± 11.36%, p = 0.277).

To investigate the development of neurite length and dendritic arborization in more detail, we measured not only their total number and length but also the number and length of primary, secondary, and tertiary dendrites after 7 and 14 days in vitro (Figure 4). While NPCs cultured for 5 days only formed neurons with primary and secondary dendrites (Figures 2C,D), NPCs cultured for 7 and 14 days established a more advanced neuronal phenotype including tertiary dendrites (Figure 4). Representative pictures display the measurement of neurite length and the number of primary, secondary, and tertiary dendrites of control and IUGR neurons after 7 and 14 days using the “Sholl analysis” (Figures 4A,C). Primary dendrites from neurons of both groups (control and IUGR) significantly extended their length over time from 7 to 14 days (Figure 4B), whereas only the IUGR group significantly increased the number of primary dendrites/neuron over time (Figure 4D). Secondary dendrites of neurons of both groups increased their number from 7 to 14 days, while only the control group significantly increased the length of secondary dendrites (Figures 4B,D). Tertiary dendrites from neurons of both groups did not significantly expand their length or number over time.

We selected the time point 5 days in vitro for further assessments of potential therapies, because our results at this time point correlate very well with the situation described in vivo in a previous study investigating structural brain changes in a rabbit model of IUGR (Pla et al., 2020). This study observed a more advanced dendritic morphology in the frontal cortex of IUGR compared to control animals (Pla et al., 2020).

With the aim to revert adverse effects on neurogenesis induced by IUGR, we evaluated the safety and efficacy of 3 potential therapies MEL, DHA, and SA on the neuronal endpoints “% of neurons,” “neurite length,” and “number of dendrites per neuron,” as well as cell viability. In a previous study of our group using rabbit neurospheres, the maximum tolerated concentration (MTC) from MEL, DHA, and SA was determined in control neurospheres with the following criteria: viability was not lower than 70% of the solvent control (SC), migration distance and % of oligodendrocytes were not significantly reduced (Kühne et al., 2022). In the current approach, the additional criteria to set the MTC was “no significant adverse effect on any on the tested neuronal endpoints” in control neurospheres. Control neurospheres were exposed to potential therapies in a concentration-dependent manner for 5 days under differentiation conditions in vitro (Figure 5). None of the tested therapies adversely disturbed the tested endpoints (Figure 5, Supplementary Figure S3) and the MTC of each compound was set at the highest tested concentration for each compound, which was in accordance with the results in Kühne et al. (2022). A summary of these concentrations is presented in Table 1.

The main interest was to find a concentration of the selected therapies which reverts the effects of IUGR. The cell viability determined by metabolic activity was always performed in the same experiments to distinguish between a specific effect and a general cytotoxic effect (Supplementary Figure S3). IUGR neurospheres were exposed to increasing concentrations of the selected therapies up to their MTC (Table 1). MEL, DHA, and SA did not significantly interfere with the % of neurons at any of the tested concentrations neither in control nor in IUGR neurospheres (Figures 5A, 6A). Our focus lied on decreasing the neurite length of the IUGR group because the total neurite length and the length of primary dendrites were significantly increased by IUGR after 5 days in vitro. MEL did not significantly reduce the total neurite length in IUGR neurospheres in none of the tested concentrations (Figure 6B). However, the lowest (0.1 μM) and the highest (3 μM) concentration of MEL significantly reduced the length of primary dendrites by presenting a stronger effect in the lowest concentration (Figure 6C, 0.1 μM MEL 20.70 ± 2.86 μm vs. SC 34.81 ± 3.48 μm, p = 0.004). MEL induced a non-monotonic response without showing a concentration-dependent effect or impact on the total neurite length, that is why MEL was not considered as the most favorable therapy against IUGR induced adverse effects on neurite length. Likewise, DHA did not present a significant reduction in the total neurite length at any of the tested concentrations (Figure 6B). Nevertheless, DHA showed a concentration-dependent effect on primary dendrites decreasing their length significantly (p = 0.001, Figure 6C). Even though DHA showed a positive effect on primary dendrites, the total length of neurites could not be improved. On the contrary, the exposure of SA to IUGR neurospheres prompted a concentration-dependent effect on total neurites and primary dendrites by significantly reducing their length. 10 μM SA significantly decreased the total neurite length (21.03 ± 0.75 μm vs. SC 36.03 ± 3.46 μm, p = 0.05), and 1 and 10 μM SA significantly reduced the length of primary dendrites (1 μM SA 22.43 ± 0.93 μm, p = 0.022; 10 μM SA 19.64 ± 0.5, p = 0.003). The total number of dendrites per neuron was not altered by any of the tested therapies (Figure 6D), neither the number of primary nor secondary dendrites per neuron (Figure 6E). Based on these findings, SA was selected to be the best candidate in reverting neurite extension induced by IUGR in vitro.

To investigate the transferability of the in vitro results to the in vivo situation, we randomly assigned pregnant rabbits to different groups and administered MEL, DHA, or LF daily from the day of IUGR induction until C-section. SA is the main metabolite of LF, and therefore not SA but the parent compound LF was selected for the treatment in vivo. The body weight of the PND0-IUGR pups from all groups (with or without treatment; w/o) was significantly lower than the body weight of the respective control group, indicating that the treatments had no effect on the body weight [see Table 1 in Kühne et al. (2022)].

Neurospheres were obtained from control and IUGR pup’s whole brain from the different treatment groups and analyzed for neuronal endpoints after 5 days under differentiation condition in vitro. The cell viability determined by metabolic activity was not significantly reduced compared to the control value in neurospheres of any treatment group (Supplementary Figure S4). The % of neurons was not significantly different between control and IUGR neurospheres obtained from pups from all treatment groups, which is in accordance with the effect observed in vitro (Figure 7A). Neurospheres obtained from IUGR pups prenatally administered to MEL did not display any improvement in neurite length, neither the total neurite length nor the length of primary or secondary dendrites (Figure 7B). Likewise, the prenatal administration of DHA to the pregnant rabbit could not prevent the adverse effect of IUGR on total neurite length, primary, or secondary dendrites (Figure 7B). However, neurospheres from IUGR pups of LF-treated rabbits presented a significant reduction in the total neurite length compared to the non-treated IUGR group (LF total: 23.71 ± 0.64 μm vs. w/o total: 36.03 ± 3.46 μm, p = 0.049, Figure 7B). The length of primary dendrites was also significantly reduced in IUGR neurospheres of the LF group compared to primary dendrites of the non-treated IUGR group (LF primary: 22.23 ± 1.10 μm vs. w/o primary: 34.81 ± 3.48 μm, p = 0.0045), while the length of secondary dendrites remained unaffected. In control neurospheres from the LF group, the neurite length of total and primary dendrites stayed on the level of w/o control neurospheres. Remarkably, these results are in good agreement with the results of the in vitro exposure to SA, the main metabolite of LF (Figures 6B,C). Finally, the total number of dendrites per neuron did not differ between any treatment group and w/o group (Figure 7C). None of the in vivo tested therapies influenced the number of primary or secondary dendrites per neuron, which was also in line with our in vitro results (Figures 6D,E).

Taking into account all results presented, the in vitro neurosphere assay correctly predicted the outcome of both, positive and negative results of the in vivo administration for the endpoints “% of neurons,” “total neurite length,” and “total number of dendrites” or “number of primary and secondary dendrites.” Merely the in vitro results of “length of primary and secondary dendrites” after exposure to MEL and DHA was not in accordance with the results of the prenatal in vivo treatment. Our findings revealed LF as the most promising therapy to prevent increased neurite length caused by IUGR.