The study was approved by the Danish National Committee on Animal Experimentation (license no.: 2020-15-0201-00520) and complied with ARRIVE guidelines.

Preterm pigs were delivered by cesarean section at 90% gestation (gestational age 106 days, n = 43, male/female = 24/19, from two sows of Danish Landrace × Yorkshire × Duroc crossbreed). Following delivery, all pigs were resuscitated and fitted with an orogastric feeding tube and an umbilical catheter. The pigs were stratified according to birth weight and sex and randomly allocated into two groups: one treated with rhIGF-1/rhIGFBP-1 complex (2.25 mg/kg/day; mecasermin rinfabate; Takeda, Cambridge, MA, USA) dissolved in formulation buffer as vehicle (50 nM sodium acetate, 105 mM sodium chloride, 0.005% (v/v) polysorbate 20, pH 5.5), hereafter named “IGF-1 group,” and the other group of pigs treated with vehicle (‘Control’ group). Treatment was administered by continuous intraarterial infusion on postnatal day one to eight (PND1-8) and then by three daily injections via a subcutaneous catheter on PND9-19. Throughout the study, the pigs were housed individually in temperature regulated environments and on PND1-5 in oxygenated incubators. The pigs were fed every 2 h during the day and every 3 h during the night on PND1-9, from PND10 they were fed every 3 h throughout the day. The diet consisted of increasing enteral volumes of raw bovine milk (32–180 mL/kg/day) supplemented until PND8 with parental nutrition in decreasing volumes (120–48 mL/kg/day). The raw bovine milk used for enteral nutrition was supplemented with 7 g/L vitamins (Phlexy-Vits; Nutricia, Zoetermeer, Netherlands), 25 g/L whey proteins (DI9025; Arla Foods Ingredients, Viby, Denmark) and 6 g/L electrolytes (Revolyt Nutrition; Gunnar Kjems, Copenhagen, Denmark), while the parental nutrition was based on the commercially available product (Kabiven; Fresenius Kabi, Uppsala, Sweden), modified as previously described (Sangild et al., 2013). Body weight of the pigs was recorded daily, and they were clinically assessed twice daily as previously described (Holme Nielsen et al., 2019). Within the first 48 h of life, all pigs were infused with sterile maternal sow plasma via the umbilical catheter (25 mL/kg) to provide passive immunity. All pigs were treated prophylactically with a combination of oral antibiotics: Amoxicillin/clavulanic acid (2care4, Esbjerg, Denmark), gentamicin (Gentocin Vet; Scanvet, Fredensborg, Denmark), metronidazole (Flagyl; Sanofi A/S, Paris, France) from PND8-10. On PND19, the pigs were anesthetized and then euthanized with an intracardiac injection of pentobarbital sodium (Euthanimal Vet; Alfasan, Woerden, The Netherlands). The total brain and selected regions were dissected, weighed and snap-frozen or fixed in paraformaldehyde, as described previously (Christiansen et al., 2023).

All parts of the in vivo study, including treatment administration, scoring, behavioral tests, and analyses were performed by blinded investigators. Animals who were immobile, had severe diarrhea, were severely dehydrated, emaciated, or fatigued and non-responsive, were omitted from behavioral tests and in the most severe cases, euthanized prior to assessment.

The volume of the right brain hemisphere, white matter, and grey matter were measured by MRI as previously described (Henriksen et al., 2022a). Briefly, the entire right hemisphere was immersed in 4% paraformaldehyde in phosphate buffered saline for 24 h at 4°C (PBS; 0.1 M, pH 7.4), rehydrated with PBS for 24 h at 4°C and washed twice with Fomblin (perfluoro-polyether; Solvay, Princeton, New Jersey, USA) to remove the excess fluid and resubmerged in clean Fomblin prior to scanning. It was confirmed that there was no surrounding fluid signal on the surface of each brain before high-resolution imaging. Data was acquired with a 9.4 Tesla preclinical scanner (Bruker BioSpin, Ettlingen, Germany) equipped with a 240 mT/m gradient coil (BGA-12S, Bruker) and using an 86-mm inner diameter transmit-receive volume coil. The imaging protocol used a 3D gradient-spoiled steady state free precession, and the parameters were set to: repetition time = 4.6 ms, echo time = 2.3 ms, number of signals averaged = 10, flip angle = 25°, field of view = 60 mm × 38.4 mm × 25.6 mm, matrix = 300 × 192 × 128, image resolution = 200 μm isotropic, acquisition time = 20 min. The images were bias-field corrected, and segmented as previously described (Henriksen et al., 2022a). The volume of white and grey matter was normalized to the total volume of the brain hemisphere.

The protein synthesis rate in the cerebellum was measured both in vivo and ex vivo. The in vivo measurement was performed as previously described (Rudar et al., 2021; Christiansen et al., 2023). In brief, animals were injected with a dose (10 mL/kg) of L-phenylalanine (Phe; 1.5 mmol/kg) containing L-[ring-²H₅]Phe at 30 mol % (0.45 mmol/kg; Cambridge Isotope Laboratories, Tewksbury, MA, USA) 30 min prior to euthanasia. Cerebellar tissue was collected and immediately snap-frozen in liquid nitrogen. The tissue was analyzed for tracer accumulation and the fractional rates of protein synthesis (Ks; protein mass synthesized/day) was calculated.

The protein synthesis rate was validated ex vivo utilizing the principles of surface sensing translation (SUnSET) and western blot as described previously (Schmidt et al., 2009; Goodman et al., 2011). Following dissection of cerebellar tissue, the samples were preincubated for 15 min in tissue buffer (Hibernate-A, Gibco, Waltham, MA, USA), then treated 30 min in 1 μM puromycin (Sigma-Aldrich, St. Louis, MO, USA) diluted in tissue buffer and immediately snap-frozen in liquid nitrogen. The frozen samples were disrupted with a pestle on ice and lyzed with neuronal protein extraction reagent (N-PER) supplemented with HALT Protease Inhibitor Cocktail kit and 0.5 M EDTA solution (all from Thermo Fisher Scientific, Waltham, MA, USA). Protein concentration was determined using Pierce BCA Protein Assay kit (Thermo Fisher Scientific) according to manufacturer’s protocol. Protein extracts (30 μg) were loaded onto Mini-PROTEAN TGX Precast Gels (Bio-Rad Laboratories, Hercules, CA, USA). The gels were electro-transferred to 0.2 μm PVDF membranes (Bio-Rad Laboratories). The blots were subsequently blocked in 5% milk in PBS, 0.05% Tween 20 for 1.5 h at room temperature and then incubated at 4°C overnight with IgG2ak monoclonal anti-puromycin antibody (1:2000, MABE343, Merck Millipore, Burlington, MA, USA). Bands were detected with horseradish peroxidase-conjugated anti-mouse IgG Fc 2a antibody (1:5000, 115–035-164, Jackson Immuno Research Labs, West Grove, PA, USA), developed with SuperSignal TMB reagent (Thermo Fisher Scientific) and visualized with an Amersham ImageQuant 800 imaging system (Cytiva Europe GmbH, Germany). Band intensity was quantified using ImageJ program (version 10.2, NIH, Berthesda, MD, USA).

IHC analysis of brain tissue was performed as previously described (Christiansen et al., 2023). Briefly, the entire right brain hemisphere was stored in 4% paraformaldehyde in PBS for collectively 72 h, cut in 7–8 mm tissue slabs, embedded in paraffin and the same brain region of all samples was cut in coronal sections (5 μm), at two levels of selected regions (at least 30 μm apart). Sections were collected on SuperFrost Plus microscope slides and were deparaffinized and rehydrated. Sections were then processed for heat induced epitope retrieval (HIER), using Tris-EDTA pH 9.0 (Olig-2) or Citrate buffer, pH 6.0 (all other epitopes), around 5 min with end-temperatures at 95°C. Hematoxylin–eosin (HE) staining was made to ensure collection of sections containing the same neuroanatomical brain structures from all animals, according to a pig brain stereotaxic atlas (Félix et al., 1999). Endogenous peroxidase was quenched by incubation in 0.3% H₂O₂ (in PBS) for 10 min. Blocking of unspecific binding sites was performed with 1% BSA dissolved in PBS containing 0.05% Triton X-100. Sections were then incubated with primary antibodies (Supplementary Table S1) diluted in the blocking buffer. Primary antibody incubation (see Supplementary Table S1) was performed for 60 min at room temperature (RT, anti-doublecortin (DCX) labeling) or 16 h at 8°C (all other primary antibody incubations). The sections were subsequently incubated with species-specific horseradish peroxidase-conjugated secondary antibodies against mouse, rabbit, or goat IgG (Supplementary Table S1), for 30 min at RT. All antibody incubations were performed in moisture chamber. Antibody binding sites of the secondary HRP conjugated antibodies were visualized with a diaminobenzidine (DAB)-H₂O₂ reaction. Sections were then counterstained with Mayer’s hematoxylin. Samples from both groups were treated simultaneously to minimize inconsistency. The whole labeled sections were scanned at ×40 magnification (Hamamatsu, NanozoomerS60, Hamamatsu, Japan) and the specific regions of interest (ROIs) were delineated on each digital image (NDP.view 2 software, Hamamatsu Photonics, Hamamatsu, Japan) according to the pig brain stereotaxic atlas (Félix et al., 1999). Labeling consistency (background, intensity) was checked before quantification and labeling artefacts (e.g., tissue folds) were excluded from annotation. The ROIs were analyzed with ImageJ (NIH, Berthesda, MD, USA) and the fraction of immunoreactive area (area%) was digitally quantified for labeling of synaptophysin (SYN), myelin basic protein (MBP), DCX, allograft inflammatory factor 1 (Iba1), oligodendrocyte transcription factor 2 (Olig2), and neuronal nuclei (NeuN).

Frozen hippocampal, cerebellar, periventricular white matter, and caudate nucleus tissues were freeze-fractured in a cryogenic tissue pulverizer and homogenized in QIAzol lysis reagent (Qiagen, Copenhagen, Denmark). Total RNA was extracted using the RNeasy Lipid Tissue Mini kit (Qiagen, Copenhagen, Denmark) and 1 μg of RNA was reverse transcribed to cDNA using the high-capacity cDNA reverse transcription kit (Thermo Fisher, Waltham, MA, USA). Porcine-specific primers were designed with Primer-BLAST software (National Center for Biotechnology Information, USA) and were validated for efficiency and specificity (Supplementary Table S2). Relative gene expression was estimated by performing quantitative real-time PCR analysis on cDNA samples (10 ng) using LightCycler 480 SYBR Green I Master kit with porcine-specific primers on a LightCycler 480 machine (both Roche, Basel, Switzerland). Expression of target gene was normalized to the expression of housekeeping gene hypoxanthine phosphoribosyltransferase 1 (HPRT1) and presented as relative expression to the Control group as determined with the 2^−△△CT method.

Statistical analysis was performed using the R software (version 4.1.2, Vienna, Austria). Effect of IGF-1 treatment on protein synthesis rate, immunohistochemical, gene expression, beam test, gait performance, open field, and novel object recognition data were analyzed using a linear model with litter, birth weight, and sex as covariates. Effect of IGF-1 treatment on repeated measurement of home cage activity was analyzed using a linear mixed effects model with the same covariates as mentioned above in addition to the pig as a random effect, and an interaction term between time and treatment. Time for acquisition of operant conditioning (as defined by the learning criteria) was analyzed using a generalized linear mixed effects model, while time for acquisition of basic motor skills and survival probability were analyzed using a cox proportional hazard model, both statistical models included same cofactors as the linear model. Normal distribution and homoscedasticity of the residual and fitted values were tested for model validation, and data were log transformed when required. If log transformation of the data did not result in normal distribution and homoscedasticity of residuals a nonparametric analysis was used. Probability values <0.05 were considered statistically significant, while values between 0.1 and 0.05 were reported as a tendency toward an effect. Correlation analysis was performed using a Pearson R correlation. A one-sample Student t-test was used to compare the novel object reference index to a null preference value of 0.5.

Sensitivity analyses were conducted by removing animals in the lower 25th percentile for birth weight across groups to account for possible survival bias of smaller animals due to treatment. Additional sensitivity analyses were conducted by removing animals in the lower 25th percentile for growth rate (g/kg/day) across groups to account for possible survival bias due to treatment of weaker animals as defined by slow growth.

IGF-1 is an essential factor for postnatal growth and organ development, including gain in brain volume (Hellström et al., 2016, 2022). Total body and brain weights, including absolute and relative weights of specific brain regions, and water content was not affected by IGF-1 treatment (Table 1). However, absolute, and relative weights of the caudate nucleus region were reduced in IGF-1 pigs (p < 0.05 and p < 0.01, respectively). Male compared to female pigs had larger total cerebellum weight (3.60 ± 0.34, n = 18 vs. 3.35 ± 0.30, n = 12; p < 0.05) and relative cerebellum weight (10.16 ± 0.79, n = 18 vs. 9.82 ± 0.68; n = 12; p < 0.05), while their relative cerebrum weights were smaller than females (77.40 ± 0.20, n = 17 vs. 78.97 ± 0 0.39, n = 12; p < 0.01). Birth weight positively affected the final body weight, relative brain weight (p < 0.001, both), the total brain, and total cerebrum weights (p < 0.05, both). The rate of cerebellar protein synthesis measured in vivo, and ex vivo indicated brain anabolic effects of IGF-1 (p < 0.05 and p < 0.001, respectively, Figures 2A,B). Preterm pigs like preterm infants are predisposed to several morbidities due to organ immaturity, however, there was a tendency for IGF-1 treatment to increase the overall survival rate (64% for Control and 76% for IGF-1 pigs, p = 0.06; Figure 2C). Further, there was significant effect of birth weight on survival (p < 0.001), indicating a positive correlation between survival and birth weight especially for the Control group (Supplementary Figure S1).

Preterm pigs, unlike term-born, are not immediately able to walk and show general delayed motor development (Holme Nielsen et al., 2019). Acquisition of essential neuromotor control such as ability to open eyes, stand, walk, and drink from a feeding trough were not different between the Control and IGF-1 group (Supplementary Figure S2). Pigs in both groups were able to open their eyes, stand and walk within five days (Supplementary Figure S2A–C), and the pigs could drink from a trough within three days after being presented to it (Supplementary Figure S2D). The activity of the pigs in their home cage, measured as the relative time active in 12 h intervals, was comparable between the groups (Figure 3A). For both groups the relative activity increased on PND5 when the pigs were moved from incubators into larger home cages. At PND7 and PND14, pigs were tested to assess the development of basic locomotor function over time using an open field test (Figure 1). Pigs from both groups traveled similar distances and had similar average velocity in the arena at the two timepoints (Figure 3B).

The development of fine locomotor skills including balance and coordination was assessed in the balance beam test on PND10 and PND18 (Figure 3C). Pigs in the IGF-1 group had fewer average mis-steps compared to the Control at both time points (PND10, p < 0.05; PND18, p = 0.07). Analysis of the gait parameters showed that the IGF-1 group took significantly longer steps on the beam at PND10, however the effect was less pronounced at PND18 (PND10, p < 0.05; PND18, p > 0.05, Figure 3D).

Associative learning was analyzed using the operant conditioning test on PND12-19. The percentage of pigs reaching the learning criteria over time was similar between groups (Figure 3E). There was a significant effect of sex and birth weight with males reaching the learning criteria faster than females and pigs were faster with increasing birth weight (p < 0.001, both).

In the novel object recognition test, the capability of non-spatial short-term memory of the pigs was evaluated. For both groups the recognition index was significantly above the null preference value of 0.5 (Control: p < 0.001; IGF-1: p < 0.05), indicating a preference for the new object, but no differences were observed between the groups (Figure 3F). There was a significant effect of birth weight, indicating increasing recognition index value with increasing birth weight and suggesting reduced preference for the new object among low birth weight animals (p < 0.05).

Grey and white matter volumes, measured by MRI, did not differ between groups (Figure 4A). To investigate if treatment had an effect on myelination in selected brain regions, the area of immunoreactivity (ir) of MBP, a marker of myelination and oligodendrocyte differentiation, was quantified in three regions of interest (ROIs): PvWM, intracortical white matter (IcWM), and hippocampus (Figure 4B). Comparison of MBP-ir in these ROIs showed significantly lower levels in PvWM and IcWM (p < 0.05 and p < 0.01, respectively) for the IGF-1 group, and similar levels in the hippocampus (Figure 4C). There was a significant effect of sex on MBP-ir in the hippocampus where males had higher levels compared to females (6.26 ± 0.37 vs. 4.88 ± 0.45, p < 0.05). Subdivision of IcWM into cingulate, parietal, and temporal subcortical areas showed that both the parietal and cingulate ROIs had significantly lower levels of MBP-ir in the IGF-1 group (p < 0.05, both), while the temporal ROI showed a tendency to a decreased level (p = 0.051, Figure 4D). Further analysis of the oligodendrocyte population was conducted by quantification of Olig2-ir, a pan-oligodendrocyte marker, in the same ROIs as for MBP-ir. Compared with controls, IGF-1 pigs had similar levels of Olig2-ir for all specified ROIs (Figures 4E,F).

Gene expression analyses of myelin protein related genes: MBP, myelin oligodendrocyte glycoprotein (MOG), myelin-associated glycoprotein (MAG); oligodendrocytic myelin paranodal and inner loop protein (OPALIN), showed no differences between groups in PvWM (Figure 4G). However, decreased expression of these genes was measured in the caudate nucleus for MBP (p < 0.01), MOG (p < 0.01), MAG (p < 0.01) and OPALIN (p < 0.05, Figure 4H). In cerebellum, expression of these genes was also lower for the IGF-1 group compared to Control, significantly for MAG (p < 0.05) although only with tendencies to significance for MBP (p = 0.08), MOG (p = 0.06) and OPALIN (p = 0.09; Figure 4I). There were no differences between groups in expression of the selected genes in the hippocampus (Supplementary Table S3).

Based on the specific effect of the treatment on caudate weight and expression of myelination-related genes, the effect on neuronal maturation in the caudate nucleus was also investigated. The expression of a panel of genes was quantified to reflect different aspects of neuronal system development (Supplementary Table S3). Interestingly, the ratio of sodium-potassium-chloride cotransporter 1 (NKCC1) over potassium-chloride cotransporter 2 (KCC2) expression, an indicator of maturation of the gamma-aminobutyric acid (GABA) signaling system, was significantly decreased for IGF-1 vs. Control pigs (Figure 5A, p < 0.05), suggesting progressed maturation of the GABAergic system. A significant decrease was also seen in the gene expression of GABA receptor subunit alpha 3 (GABRA3) for the IGF-1 group (Figure 5B, p < 0.05), further supporting an IGF-1 treatment effect on the GABAergic system. Decreased IGF-1 related expression was also seen for glutamate receptor 4 (GRIA4, p < 0.01) and hypoxia-inducible factor 1-alpha (HIF1A, p < 0.05), with tendencies for cadherin 8 (CDH8, p = 0.08) and S100 calcium-binding protein B (S100B, p = 0.05). The expression of genes related to IGF-1 signaling were similar between groups (Supplementary Table S3). Overall, the gene expression analyses indicated IGF-1 effects on neuronal maturation, especially with regards to the GABAergic system, but without effects on endogenous IGF-1 signaling in the caudate nucleus.

For hippocampus, previous studies show that the IGF-1 receptor is highly expressed in this region and co-localized with immature neurons in preterm pigs (Christiansen et al., 2023). However, the expression level of a panel of genes related to general neuronal maturation and development showed no differences between groups (Figure 6A; Supplementary Table S3). There were no effects on the hippocampal neuron population, as examined by NeuN-ir, a marker of mature neurons, and DCX-ir, a marker of immature neurons (Supplementary Figure S3). In contrast, quantification in hippocampal subregions (Figure 6B) of SYN-ir, a membrane protein of synaptic vesicles and marker of synaptic density, showed lower levels in the hilus subregion of the hippocampus for the IGF-1 group (Figure 6C, p < 0.05) and a tendency for this effect in the CA subregion (Figure 6C, p = 0.09). Iba1-ir (a marker of microglia) was similar between groups, measured for the same regions as SYN-ir, indicating no microgliosis here (Figure 6D).

The tendency to higher survival for IGF-1 pigs, with significant effect of birth weight, suggested a survival bias for smaller animals in the IGF-1 group. Sensitivity analyses were therefore conducted by removing the lowest 25th percentile for birth weight for both treatment groups (Supplementary Table S4), accepting lower, but more homogenous, sample size for comparison. Most findings were unchanged after this adjustment, however the previous significant effects (full sample size) of IGF-1 on MBP-ir (PvWM, cingulate subcortex), stride length (beam test) some caudal gene expressions (MOG, MAG, OPALIN, NKCC1/KCC2 ratio) disappeared after exclusion of the low birth weight pigs in the sensitivity analysis, although group means ± SEM remain similar to the primary analyses.

The increased risk of mortality for pigs with lower birth weight reflects their increased risk of severe clinical complications (Supplementary Figure S1). Animals that suffer from these complications show poor growth and sensitivity analyses were therefore also conducted by removing the lower 25th percentile for growth rate. Again, treatment effects remained mostly unchanged but after exclusion of pigs with lowest growth rate, the differences in levels of MBP-ir (PvWM, parietal and cingulate subcortex) between groups were no longer significant. Further, IGF-1 treatment failed to affect stride length (beam test) and reduced the percentage of animals reaching the learning criteria in the operant conditioning test. Although, these results were based on a lower sample size.