Growth Differentiation Factor-15 Knockout Mice Are Protected from Neonatal Hypoxic-Ischemic Injury in a Sex-Dependent Manner

Neuroinflammation has beneficial and detrimental effects on brain tissue survival in resuscitated newborns with hypoxic-ischemic (HI) neonatal encephalopathy (NE) [1]. Sex assigned at birth robustly impacts the balance between the opposing effects of neuroinflammation on the brain. Specifically, males have a greater number of activated microglia (MG) and infiltrating leukocytes and macrophages in the injured brain and increased markers of histological damage versus females in animal models of HI NE [1, 2]. These findings agree with clinical observations that male newborns with birth asphyxia have increased risk of unfavorable outcome [3].

Experimental immune-modulating therapies in NE models also affect neurological outcomes differently in males versus females. For instance, MG depletion increased brain tissue volume loss in males but not in females [4]. Gene knockout (KO) studies in mice are still required to elaborate on this finding by elucidating the role that individual neuroinflammatory protein mediators have on brain tissue injury in both sexes, particularly in neonatal HI models that recapitulate the unique molecular and physiological state of newborns.

The hormone growth differentiation factor-15 (GDF-15, a.k.a. macrophage inhibitory cytokine-1) is released from macrophages during inflammation and promotes survival in animal models of severe infection – a condition in which enhanced inflammation can be beneficial [5‒7]. Focal and global ischemia can also increase GDF-15 levels in the brain and organs outside of the CNS [8, 9]. Further, serum GDF-15 levels are markedly increased after asphyxial cardiac arrest in infants and children and associated with unfavorable outcome [10, 11]. But whether GDF-15 is detrimental or beneficial in pediatric brain injury remains to be defined. In adult mice, blocking systemic GDF-15 activity with genetic KO did not affect infarct volume after middle cerebral artery occlusion stroke, although female mice were not investigated [9]. Much remains to be learned about GDF-15 in the adult or developing female brain; the recent successful translation of GDF-15 inhibitors for clinical use highlights the timeliness of understanding GDF-15’s role in brain injury [12].

Here, we examined the effect of GDF-15 KO on 14-day post-insult hemispheric and hippocampal lesion volume, and on ionized calcium-binding adaptor molecule 1 (Iba-1) staining, in male and female postnatal day 9–11 mice in the Rice-Vannucci model of developmental unilateral forebrain HI, with strict temperature control to normothermia [13]. Our findings reveal a detrimental effect of GDF-15 in murine neonatal HI in males but not females. A concordant enhanced MG/macrophage response to HI in male versus female wild-type (WT) mice suggests a sex-dependent deleterious neuroinflammatory role of GDF-15.

C57BL/6-Gdf15^{tm1(cre/ERT2)Amc}/J mice (a.k.a. Gdf15^nuGFP−CE) were purchased from Jackson Laboratory (Stock #034497). The design and validation of these mice were previously described [14]. Mice were bred at the University of Pittsburgh in a pathogen-free housing facility managed by the Division of Laboratory Animal Resources. All studies were approved by the Institutional Animal Care and Use Committee of the University of Pittsburgh (Protocol #22071487). Euthanasia protocols followed American Veterinary Medical Association guidelines. Mice were maintained on a 12-h light cycle and given free access to food and water. To maximize rigor of the experimental design, mice were genotyped after they recovered from injury by a second technician blinded to treatment group. Tissue genotyping was performed by standard PCR in two different reactions using JAX Protocol 37350 with adaptations. The mutant Cre amplicon (200 bp band) was detected using forward and reverse primers: 5′-GCG​GTC​TGG​CAG​TAA​AAA​CTA​TC-3′ and 5′-GTG​AAA​CAG​CAT​TGC​TGT​CAC​TT-3′. The WT GDF15 amplicon (352 bp band) was detected using forward and reverse primers: 5′-TGT​TAG​GCC​TGG​TTT​GAA​GG-3′ and 5′-CTG​GGT​GGA​TGG​ATT​GTG​T-3′. WT, heterozygote (Het), and KO mice were included in this analysis. Because investigators were blinded to genotype before injury, group sizes were unevenly distributed across genotypes. Investigators were not blinded to sex. We collected and stored brains in formaldehyde and targeted n = 6 for all groups, based on our prior work in male mice [13]. We processed up to the first 6 brains in the order that they were collected for histology for each group. In total, 70 pups were used for this study.

Postnatal day 9–11 mice underwent sham or HI procedures as we reported [13]. In brief, isoflurane-anesthetized pups (3% induction and 1.5–3% maintenance in a 2:1 mixture of N₂O/O₂) were weighed and core body temperature (Tb) was monitored during surgery (sham or common carotid artery [CCA] occlusion) via a neonatal rectal probe (Physitemp). Tb was maintained to ∼36°C, which is normothermic at this age in mice [13]. Postsurgical temperature was maintained by intermittent measurement with a pediatric infrared thermometer (Braun Healthcare US). The right CCA was ligated and cut. The incision was closed with 3M Vetbond Tissue Adhesive (Maplewood) and bupivacaine (0.25%) applied to the wound. Mice advanced through the protocol in blocks of 4. A second technician working in tandem maintained pup temperatures at ∼36°C post-surgery. Pups were then returned to the foster dam for 1 h. They were then placed inside an acrylic hypoxia chamber glove box (Coy; Grass Lake, MI, USA). The hypoxia exposure time was 25 min (8% O₂/92% N₂). Tb was monitored via infrared thermometer during hypoxia and maintained at ∼36°C. Shams received surgery without CCA ligation and were maintained at normoxia and normothermia for 25 min. After hypoxia or normoxia, Tb was maintained at ∼36°C in room air for 2 h 5 min on a benchtop with lamps and heating blankets, and temperature recorded every 15 min.

At 14 d post-insult, mice were transcardially perfused with heparinized saline followed by 10% neutral buffered formalin. Whole fixed brains were embedded in paraffin. Sections were taken every 0.5 mm from posterior to anterior and stained with hematoxylin and eosin. Hemispheric area was measured on each slide by a technician blinded to group and added for total volume (MCID Analysis software, St. Catherines, ON, Canada). Hippocampal area was also quantified on slides with a visible hippocampus (∼2-mm distance in total), and the volume was calculated. To normalize for brain volume changes of technical (e.g., embedding) or genetic (e.g., GDF-15 expression) origin, lesion volume loss was calculated as a percentage of the contralateral side. In our prior work, we demonstrated consistent lesion volume at 14 d with this HI injury model, and differences at this timepoint are thought to reflect durable neuroprotection [13, 15].

Sections from formalin-fixed paraffin-embedded tissue were deparaffinized and rehydrated and underwent an antigen retrieval process (Antigen Decloaker; Biocare Medical, Pacheco, CA, USA) as previously described [16]. Endogenous peroxidases were blocked in 0.3% H₂O₂ for 30 min, and then tissue was blocked with 3% normal goat serum (Vector, Burlingame, CA, USA) in TBS containing 0.25% Triton X for 1 h at room temperature. Sections were incubated overnight containing an antibody for the MG/macrophage marker Iba-1 (Abcam, Cat. 178846, 1:1,000) [17]. The sections were then incubated for 1 h with biotinylated anti-rabbit secondary antibody followed by the avidin-biotin complex for 1 h (ABC Kit, Vector) and stained with 3,3′-diaminobenzidine (Vector). Slides were imaged using a Nikon Eclipse 90i microscope [16]. The number of Iba-1-positive (Iba-1+) objects was counted using QuPath [18]. Regions of interest were identified in scanned slides, and the threshold function was used to identify Iba-1+ staining. A size and signal intensity threshold was established to discriminate between artifact and cellular staining.

Original or transformed data (Box-Cox or aligned rank transformation [ART] method [19]) that were normally distributed and homoscedastic were analyzed using two-way ANOVA. Post hoc testing was done using Tukey’s multiple comparison test. For pairwise comparisons on ART-transformed datasets, the original data were again transformed but with the ART-Contrasts tool (ART-C [20]). Then, the transformed data were analyzed with a standard one-way ANOVA and post hoc significance detected with the Tukey’s multiple comparison test. Data were significant at p < 0.05. Cubic regression was performed to assess the correlation between Iba-1+ counts and lesion volume. Statistical analysis was performed using Prism10 (GraphPad Software, LLC).

In males, there was a significant interaction between insult and genotype on hemispheric volume at 14 d post-insult (p = 0.014). In addition, insult alone had a significant effect on hemispheric volume (p < 0.001) (shown in Fig. 1). Specifically, on post hoc testing male WT mice had significant volume loss in the injured hemisphere post-HI versus male shams (WT HI median 21.31% interquartile range [3.42–55.83] vs. WT sham 0.85% [−0.28 to 2.58], p < 0.01). In contrast, hemispheric volume loss in injured male KO mice did not differ versus male KO sham (KO HI 1.44% [−0.17 to 4.59] vs. KO sham 1.70% [−0.89 to 4.13], p > 0.99). Het males had an intermediate phenotype after injury (Het HI 5.62% [2.68–12.68] vs. Het sham 1.80% [−2.99 to 3.09], p = 0.07).

In females, there was no effect of genotype (p = 0.516) or interaction between genotype and injury (p = 0.588). But insult alone had a significant effect on hemispheric volume (p < 0.0001). HI in females produced a milder insult as determined by the magnitude of hemispheric tissue loss. On post hoc testing, injured female WT and KO mice still had significant tissue loss after HI versus sham groups (WT HI 4.61% [4.41–7.59] vs. WT sham 1.89% [0.44–2.84], p < 0.01, KO HI 5.16% [4.12–6.72] vs. KO sham 2.55% [0.55–3.29], p < 0.05). And there was no significant hemispheric volume loss in female Het mice versus shams after HI (Het HI 3.13% [2.16–8.15] vs. Het sham 1.77% [−0.17 to 4.49], p = 0.54).

Regarding hippocampal volume loss in males, there was a significant effect of insult (p < 0.0001) and a trend with genotype (p = 0.070), but no interaction between genotype and insult (p = 0.215) (shown in Fig. 1). On post hoc testing, injured WT, Het, and KO males had significant hippocampal volume loss versus their respective sham groups (WT HI 65.53% [17.02–100.00] vs. WT sham 4.50% [−3.54 to 8.95], p < 0.0001, Het HI 24.98% [18.79–59.27] vs. Het sham 0.44% [−4.36 to 6.33], p < 0.001, KO HI 31.85% [12.32–35.69] vs. KO sham −3.73% [−9.74 to 8.11], p < 0.001). In females, there was an effect of insult (p < 0.0001) but not genotype (p = 0.882) or the interaction between genotype and insult (p = 0.585) (shown in Fig. 1). On post hoc testing, injured WT, Het, and KO also had significant hippocampal volume loss versus respective sham groups (WT HI 18.53% [15.06–26.01] vs. WT sham 5.47% [−2.87 to 9.93], p < 0.05, Het HI 23.07% [10.12–44.94] vs. Het sham 1.35% [−0.85 to 7.65], p < 0.01, KO HI 32.26% [12.19–36.38] vs. KO sham 2.75% [0.24–6.80], p < 0.01).

Next, we examined Iba-1+ cell counts in the injured hemisphere and hippocampus. In males, there was a trend for insult (p = 0.081) to increase and GDF-15 KO genotype to decrease (p = 0.078) the number of Iba-1+ cells in the injured hemisphere, but no interaction between the 2 factors (p = 0.929) (shown in Fig. 2). In females, there was no significant effect of injury (p = 0.474), genotype (p = 0.590) or for their interaction (p = 0.898) (shown in Fig. 2).

Regarding the hippocampus, brain atrophy was so extensive in 2 WT mice that it was missing and thus unable to be included for Iba 1+ analysis. In males, there was a significant effect of insult (p < 0.001), but not genotype (p = 0.237) or for the interaction between genotype and insult (p = 0.196) (shown in Fig. 2). On post hoc testing, injured male KOs had a significant increase in Iba-1+ cells in the hippocampus after injury versus KO shams which was not seen in WTs or Hets (WT HI 0.56 Iba-1+/mm² [0.41–1.57] vs. WT sham 0.51 [0.46–0.54], p = 0.959, Het HI 0.52 [0.49–0.91] vs. Het sham 0.42 [0.33–0.52], p = 0.204, KO HI 0.61 [0.48–0.83] vs. KO sham 0.35 [0.29–0.40], p < 0.05). In females, there was a trend for an effect of genotype on Iba-1+ cells (p = 0.077), but not for insult (p = 0.706) or for the interaction between genotype and insult (p = 0.399) (shown in Fig. 2).

In males, there was a strong correlation between hemispheric and hippocampal volume loss and Iba-1+ cell counts (R² hemisphere = 0.976, R² hippocampus = 0.853, shown in Fig. 3). In females, correlations between hemispheric and hippocampal lesion volume and Iba-1+ cell counts were not seen (R² hemisphere = 0.365, R² hippocampus = 0.275, shown in Fig. 3).

We report that total GDF-15 KO decreased hemispheric and hippocampal volume loss in male neonatal mice after HI brain injury and was associated with decreased Iba-1+ cells in both brain regions versus injured male WT mice. Trends toward reduced histological markers of damage were also seen in injured male Het mice, but our study was not powered to test for a possible gene dose effect. In contrast, neither hemispheric nor hippocampal volume, nor Iba-1 staining, differed in injured female GDF-15 KO, Het, or WT mice. Our data suggest a detrimental role of GDF-15 for the male brain in neonatal HI and the possible involvement of MG/macrophages linked with that sex-dependent effect.

Prior studies on sex differences in the Rice-Vannucci model show that male mice have larger infarct volume versus females given equivalent insult parameters [1, 21]. Greater tissue loss in males is unlikely to be explained by the levels of sex hormones since there are no differences in testosterone and estrogen levels post-injury [1, 21]. Intrinsic sex differences in mediators of neuroinflammation could, however, underlie the divergence in tissue loss. Prior studies showed that male pups had higher numbers of MG at early postnatal ages and that male rodents had more activated MG versus females at 3 d post-neonatal HI [1]. Here, we found that male WT shams had higher Iba-1+ counts versus WT females, and that injured WT males had higher Iba-1+ counts versus injured females. Moreover, hemispheric tissue loss was greater in injured males. Thus, both baseline sex differences in Iba-1 and the response to injury were recapitulated here in the Rice-Vannucci model.

GDF-15 KO was neuroprotective in males but not in females, consistent with other studies exploring other immune-modulatory approaches to increase brain tissue survival in this model in both sexes. For instance, peripheral myeloid cell depletion decreased brain tissue loss in males, but not females at 14 d post-insult [22]. In contrast, supporting beneficial effects of MG in this model, pre-injury MG depletion increased damage, lesion volume, and number of TUNEL-positive cells in brain in males but not females [4]. It is unclear why immune modulation in females provides less benefit, or conversely does not aggravate damage. It could be that neuroinflammation is less pronounced after neonatal HI in females – providing less of a target to inhibit in settings with detrimental effects, or greater opportunities to enhance benefit. Furthermore, GDF-15 requires the GDNF Family Receptor Alpha Like (GFRAL) receptor to trigger downstream pathways [23, 24]. Sex differences have also been reported in GFRAL KO mice – such as impairing mitochondrial stress-induced nighttime food intake in males but not in females [25].

A variety of non-CNS conditions also increase GDF-15 expression [6, 14, 26‒28]. Septic patients have increased serum GDF-15 levels regardless of the infectious agent, and in models of viral or bacterial sepsis, blockade of GDF-15 signaling increases mortality [6]. This supports a beneficial effect of GDF-15 on survival in sepsis – contrasting our findings in neonatal HI in the brain without infection. Benefit from greater inflammation might be expected in models of infection. Age-related differences might also contribute to the divergent responses. Brain microvascular endothelial cells increase GDF-15 expression in Escherichia coli meningitis, but the role of GDF-15 in CNS infection remains unclear [29].

GDF-15 increases in multiple tissues post-ischemia, including the brain, myocardium, and kidney [9, 14, 26]. Regarding brain, in middle cerebral artery occlusion in adult mice, GDF-15 expression increases after injury in the cortex and hippocampus, but KO of GDF-15 expression did not impact neurologic score or infarct volume versus WT [9]. This contrasts our results suggesting that age may be critical to understand the role of GDF-15 in mediating HI brain injury. It is unclear if there is a GDF-15 switch in the neonatal period, as seen in several other key mechanisms that have divergent effects on brain tissue survival after HI in adults versus newborns [30‒32].

GDF-15 is a member of the TGF-β cytokine family and impacts the immune response to injury [6]. In a traumatic spinal cord injury (SCI) model in adult female mice, consistent with our data, GDF-15 KO reduced infiltrating macrophages. GDF-15 overexpression increased macrophage accumulation post-injury; however, it improved functional motor outcome [33]. GDF-15 is secreted by Schwann cells, and naïve GDF-15 KO mice have progressive loss of motor neurons in the spinal cord and motor deficits as they age [34]. Thus, it is unclear if GDF-15 overexpression improved functional outcomes in SCI by targeting neuroinflammation or rather neurotropic support of spinal axons. In contrast, in myocardial ischemia in adult male mice, GDF-15 inhibits β2 integrin activation on polymorphonuclear leukocytes, reducing their recruitment into the myocardium. In that model, GDF-15 KO increased mortality from 35% to 81% [26]. A gene dose effect on mortality was also seen. Thus, there are important but conflicting roles of the effect of GDF-15 on inflammatory cell accumulation that depend on the tissue, the inflammatory cells involved, the type of insult, developmental age, and sex.

Given GDF-15’s implicated role in the immune response, we examined the number of Iba-1+ cells as a marker of combined activated MG/macrophage accumulation in the injured hemisphere and hippocampus [6, 35, 36]. In males, we found higher hippocampal Iba-1+ counts after HI with a similar trend in hemispheric data. Although sample size limited our statistical power, we saw a trend toward high Iba-1+ counts post-HI that differed across genotype in males. Iba-1+ counts were not affected after HI or by genotype in females. This suggests that the impact of GDF-15 on neuroinflammation varies by sex, with an increased MG/macrophage response and exacerbation of injury in male neonatal WT post-HI. Future studies should test whether targeted GDF-15 blockade is a therapeutic opportunity in males in neonatal HI models [6].

In addition to the impact on the inflammatory response, GDF-15 KO may have more systemic effects on animal development. However, prior reports examining GDF-15 KO versus WT mice did not demonstrate any alterations in cardiovascular physiology or other major abnormalities [8, 34]. Although one report noted that adult female GDF-15 KO animals had significant weight gain, in our study there were no differences in pre-injury weights across genotypes in neonates [34]. Nevertheless, there remains a possibility that genotypic differences in extra-cerebral or noninflammatory factors may have influenced the response to HI.

Once animals mature into adulthood, GDF-15 may transition to a neuroprotectant. In a SCI model, GDF-15 inhibited ferroptotic neuronal death in adult male mice via effects on Keap1-Nrf2 [37]. GDF-15 is also directly neuroprotective in dopaminergic neurons after application of the neurotoxin 6-hydroxydopamine in vitro [38]. Age-dependent expression of GFRAL could impact effects of GDF-15, and based on our data, this question merits investigation. While GFRAL was originally thought to be restricted to a small area of the brainstem, recent work suggests that it may be expressed more widely in brain, particularly in developmental animals [23, 39, 40]. Finally, activation of GFRAL-expressing neurons in adult animals can induce hypothermia and alter neuronal metabolism [41]. This may also differ in developmental animals.

Limitations of our study include relatively small group sizes. To minimize the risk of bias, we did not genotype mice until after injury, limiting our ability to maximize group sizes within any given litter. Iba-1 staining also does not characterize the phenotype of MG activation. MG have pleiotropic and sex-dependent effects in the brain [35, 42], and future studies should seek to characterize whether GDF-15 impacts MG phenotype. We could not meaningfully characterize effects of GDF-15 KO on accumulation of polymorphonuclear leukocytes in our study, given that outcome assessment was limited to 14 d post-HI. Future studies should thus examine temporal effects on acute and chronic neuroinflammation and the functional and behavioral impacts of our histological findings.

We conclude that GDF-15 exerts a sex-dependent increase in lesion volume in an NE model. The increased injury seen in WT versus KO males is associated with an increased MG/macrophage response. Future work should aim to identify how GDF-15 mediates the increase in injury and the differing inflammatory response between the sexes and test whether GDF-15 blockade, as used clinically for other indications, is a novel therapeutic target in NE in males [12].

All studies were approved by the Institutional Animal Care and Use Committee of the University of Pittsburgh (Protocol #22071487).

The authors report no competing interests.

This work was supported by NIH/NINDS grant R01NS105721 to T.C.J., by a Lloyd Reback Family Gift, Laerdal Foundation for Acute Medicine grant, Zoll Foundation grant, and T32 (2T32HD040686) to J.R.H., and by the Ake N. Grenvik Chair in Critical Care Medicine to P.M.K. None of the funders had a role in the study design, execution, analysis, manuscript conception, planning, writing, or decision to publish.

J.R.H. analyzed the data, drafted the main manuscript, created the figures, and funded experiments. P.M.K. and T.C.J. analyzed data, funded experiments, and substantially edited the manuscript. V.A.V., K.A.J.-F., and J.P.S. performed the experiments.

All data generated or analyzed during this study are included in this published article. Further inquiries can be directed to the corresponding author.