Downregulation of the GHRH/GH/IGF1 axis in a mouse model of Börjeson-Forssman-Lehman syndrome

The Börjeson-Forssman-Lehmann syndrome (BFLS; OMIM 301900) is a rare intellectual disability syndrome caused by mutations in the X-linked gene PHF6 (Börjeson et al., 1962; Lower et al., 2002). The disorder affects all known males with deleterious hemizygous PHF6 mutations, while females heterozygous for PHF6 mutations range from a full presentation of BFLS that is similar to males to apparently unaffected carriers (Berland et al., 2011; Crawford et al., 2006; Zweier et al., 2013). Neurological symptoms of BFLS include low IQ and susceptibility to seizures (Turner et al., 2004). Physical characteristics of BFLS in both males and females include deep-set eyes, large ears, tapering fingers and short toes (Turner et al., 2004). Several clinical signs of BFLS, including short stature, truncal obesity, gynecomastia and genital hypoplasia, point to disrupted endocrine control (Turner et al., 2004).

The PHF6 protein contains two plant homeodomain (PHD) zinc fingers (Lower et al., 2002), which are domains that occur in chromatin-associated proteins implicated in transcriptional regulation (Aasland et al., 1995). Mutations of PHF6 in BFLS occur along the length of the gene, indicating a likely loss or reduced function mechanism in the pathogenesis. PHF6 localizes to the nucleus (Lower et al., 2002) and associates with histone proteins, including H3 (Soto-Feliciano et al., 2017), and histone variants, including H1.2, H2B.1, H2A.Z and H3.1 (Todd and Picketts, 2012). PHF6 has been implicated in transcriptional regulation through association with the PAF1 and NuRD complexes (Todd and Picketts, 2012). PHF6 has recently been reported to function in processes other than transcription, including in non-homologous end-joining (Warmerdam et al., 2020), and as an E3 ubiquitin ligase (Oh et al., 2020). In addition to germline mutations in BFLS, somatic PHF6 mutations occur in human hematological malignancies, including T-cell acute lymphoblastic leukemia (Van Vlierberghe et al., 2010) and acute myeloid leukemia (Van Vlierberghe et al., 2011). PHF6 has recently been proven to be a tumor suppressor (McRae et al., 2019), in addition to regulating hematopoietic stem and progenitor cell function (Hsu et al., 2019; McRae et al., 2019; Miyagi et al., 2019; Wendorff et al., 2019).

A recent report characterized the neurological features of a mouse model of BFLS with a patient-specific mutation involving replacement of cysteine 99 in the first PHF6 PHD finger with phenylalanine (PHF6^C99F) (Cheng et al., 2018). These mice displayed reduced levels of PHF6 protein, evidence of deficits in contextual and social recognition memory, and increased seizure susceptibility in response to epilepsy-inducing drugs (Cheng et al., 2018). Mice with the PHF6^C99F mutation were reported to weigh less and to have reduced body length (Cheng et al., 2018), recapitulating the short stature observed in BFLS. The mechanisms underpinning this growth defect have not been investigated.

The hypothalamic-pituitary axis controls endocrine function in mammals. Transcription and secretion of the anterior pituitary hormones, growth hormone (GH), thyroid stimulating hormone (TSH), adrenocorticotrophic releasing hormone (ACTH), and the gonadotropins luteinizing hormone (LH) and follicle stimulating (FSH) are stimulated by their respective hypothalamic releasing hormones, including growth hormone-releasing hormone (GHRH), thyrotrophin-releasing hormone (TRH), corticotrophin-releasing hormone (CRH) and gonadotropin-releasing hormone (GnRH) (Chrousos et al., 2009; Ikegami and Yoshimura, 2017; Ranke and Wit, 2018; Stamatiades and Kaiser, 2018; Thackray et al., 2010), while prolactin (PRL) secretion is primarily controlled via negative regulation by dopamine (Bernard et al., 2019). The anterior pituitary hormones (GH, TSH, LH, FSH, ACTH and PRL) are secreted into the bloodstream and in turn regulate production of downstream hormones at peripheral sites. Pituitary hormones exert pleiotropic effects, including the control of growth in the case of the GH and thyroid axes (Ikegami and Yoshimura, 2017; Ranke and Wit, 2018).

Endocrine testing of individuals with BFLS has revealed alterations in pituitary hormone levels in some patients [see Table S1, adapted from a summary by Zhang et al. (2019)]. Specifically, GH was found to be lower than average in three out of four males (Birrell et al., 2003; Dereymaeker et al., 1986; Robinson et al., 1983) and three out of four females assayed (Petridou et al., 1997; Robinson et al., 1983; Zhang et al., 2019). TSH was decreased in two out of four assessed males (Birrell et al., 2003; Dereymaeker et al., 1986; Robinson et al., 1983; Turner et al., 1989; Weber et al., 1978) and increased in one out of five assessed females (Birrell et al., 2003; Carter et al., 2009; Crawford et al., 2006; Matsuo et al., 1984; Petridou et al., 1997; Robinson et al., 1983; Zhang et al., 2019). As both the GH and thyroid axis control growth (Wit et al., 2016), it is possible the reduced function of one or both of these pathways is responsible for the observed short stature in BFLS.

PRL levels were reported to be unchanged in one male tested (Robinson et al., 1983) and in three females tested (Petridou et al., 1997; Robinson et al., 1983), while ACTH deficiency was mentioned in a case report of two brothers (Birrell et al., 2003). Testosterone levels were commonly affected, with 11 out of 14 males showing a decrease (Ardinger et al., 1984; Baumstark et al., 2003; Carter et al., 2009; de Winter et al., 2009; Robinson et al., 1983; Turner et al., 1989; Weber et al., 1978). Levels of the gonadotropins LH and FSH were raised in some cases but lowered in others (Table S1), leading to the hypothesis that BFLS may involve both central and peripheral defects of the gonadotropin axis (Robinson et al., 1983).

It is important to note that these endocrine tests have been performed in various laboratories over a span of more than 40 years. Consequently, it is unclear whether the discrepancy in results represents true variation between individuals (perhaps due to different genetic variants in the PHF6 gene, polymorphisms in the genetic background and/or environmental factors), variable testing procedures or other factors, such as the age at which patients were tested.

Although the endocrine test results are variable among individuals with BFLS, the symptoms of BFLS, including frequent short stature (with a normal birth weight), small genitalia and gynecomastia (Gécz et al., 2006), along with altered pituitary hormone levels identified in a number of BFLS patients led us to hypothesize that loss of PHF6 may disrupt hypothalamic-pituitary axis function. Therefore, we sought to characterize the role of PHF6 in regulation of this axis. Here, we report the phenotype of a novel mouse model of BFLS and identify PHF6 as a key regulator of the GHRH/GH/IGF1 axis, which controls growth, at least in part, through regulation within the nervous system.

To study the role of PHF6 in the endocrine system, we used a murine Phf6 allele with loxP sites flanking exon 4 and 5 (McRae et al., 2019), which results in undetectable Phf6 mRNA upon cre-mediated recombination (Fig. S1A). As previously reported, males with germline deletion of Phf6 on a C57BL/6 background died shortly after birth, while heterozygous females survive to adulthood (McRae et al., 2019). Female mice with germline mutation in one copy of Phf6 had a significantly reduced body weight compared with control mice (P<0.001; Fig. 1A). A significant difference was detected by 3 weeks of age using Bonferroni's post-hoc test for multiple comparisons (7.47±0.33 g versus 9.00±0.25 g for Phf6^+/− and Phf6^+/+ mice, respectively, P=0.01; Fig. 1A). Male Phf6^−/Y mice isolated just before birth at E18.5 had an equal body weight compared with controls (1.13±0.04 g versus 1.13±0.05 g, P=0.96; Fig. S1B), indicating the effect of PHF6 on growth and survival occurs postnatally.

The survival of human BFLS patients with mutations expected to result in complete loss of PHF6 suggest that hemizygous PHF6 loss-of-function mutations should be compatible with postnatal life. We hypothesized that the genetic background may be an important determinant of survival. In order to model an outbred genetic background that may more closely represent the effect of loss-of-function mutations in humans, we bred the Phf6⁻ allele onto a mixed FVB/BALB/c background. In contrast to the C57BL/6 background, Phf6^−/Y males survived postnatally on the mixed FVB/BALB/c background, although they were slightly underrepresented by weaning with 17.7% versus the expected 25% (P=0.0498; Table S2), for reasons yet to be determined. Once weekly weighing showed that adult Phf6^−/Y mice have a reduction in body weight compared with wild-type mice (P<0.001), with a significant difference detected by 2 weeks of age with Bonferroni's post-hoc test for multiple comparisons (5.13±0.45 g, versus 8.27±0.39 g, P<0.001; Fig. 1B). The largest difference in body weight was at 3 weeks of age where Phf6-deleted mice were ∼50% smaller than controls (5.40±0.66 versus 10.72±0.47, P<0.001; Fig. 1B). Body length (79.75±2.02 mm versus 96.00±1.08 mm) and femur length (9.75±0.32 mm versus 11.75±0.14 mm) was significantly reduced in Phf6^−/Y mice compared with controls (P=0.0004 and 0.001, respectively; Fig. 1C-E), indicating the reduced body weight is a result of reduced overall growth rather than reduced adipose tissue. Indeed, relative to body weight, Phf6^−/Y mice had an increase in the inguinal subcutaneous and dorsal lumbar white fat compared with controls (0.019±0.002 g versus 0.011±0.001 g, respectively, P=0.02), but no change in the interscapular brown adipose tissue (Fig. 1F). In absolute terms, the lean body volume, but not the adipose tissue volume, was significantly reduced in the Phf6^−/Y mice compared with controls (Fig. 1G). Males of the Phf6^−/Y genotype were fertile. Intercrosses between Phf6^−/Y and Phf6^−/+ demonstrated that FVB/BALB/c females with homozygous mutation in Phf6 could survive and, as was demonstrated for Phf6^−/+ heterozygous females, they were smaller than control females (10.41±0.80 g, 8.32±0.33 g and 7.00±0.01 g for Phf6^+/+, Phf6^−/+ and Phf6^−/− genotypes, respectively, at 3 weeks of age; P=0.01 for Phf6^−/− versus Phf6^+/+ and P=0.02 for Phf6^−/+ versus Phf6^+/+; Fig. S1C).

To test whether the loss of PHF6 altered circulating levels of pituitary hormones, we assayed the concentrations of ACTH, GH, FSH, LH, PRL and TSH using a multiplex assay in male mice on the FVB/BALB/c background. Because of the pulsatile nature of pituitary hormone secretion (Steyn et al., 2011, 2013), we assayed blood isolated every 10 min over 6 h (Fig. 2A, Fig. S2).

Given the smaller size of the Phf6^−/Y animals, we hypothesized that there would be lower levels of GH. As expected, Phf6^−/Y mice displayed a reduction in the mean concentration and total secreted GH (area under the curve) to 55% of control mice (Fig. 2A, Table 1). Deconvolution analysis using the AutoDecon algorithm (Johnson et al., 2008, 2009) indicated that the basal secretion rate of GH and the number of peaks were unaffected by PHF6, but there was a decrease in the maximum peak mass to 14% of controls and a decrease in the average peak mass in Phf6^−/Y animals to 15% of Phf6^+/Y littermates (Table 1).

To test whether the reduction in secreted GH levels corresponded to a reduction in pituitary GH levels, we performed western blots for GH using pituitary lysates and observed a reduction in GH protein, relative to the actin loading control, in Phf6^−/Y pituitaries versus Phf6^+/Y controls (Fig. 2B,C). These results indicate that pituitary GH and secreted GH levels are decreased in mice lacking PHF6.

In contrast to the effect of PHF6 on GH secretion, there was no effect of PHF6 loss on the concentration or secretion parameters of the other pituitary-derived hormones assessed: ACTH, FSH, LH, PRL and TSH (Fig. S2, Tables S3 to S7). Despite no change in the circulating levels of ACTH, FSH, LH, PRL and TSH between Phf6^−/Y and control mice, we identified changes in the mRNA levels in hypothalamic pituitary axis genes of some of these pathways through assessing a panel of candidate genes by RT-qPCR. In male FVB/BALB/c Phf6^−/Y mice, we found reduced mRNA levels of the prolactin gene Prl in the pituitary gland (Fig. S3A). Levels in the pituitary gland of the pro-opiomelanocortin gene Pomc, which encodes the ACTH precursor, were significantly reduced in Phf6^−/Y versus control adult mice, whereas mRNA levels of corticotrophin-releasing hormone (Crh), which encodes the Pomc stimulating factor, were unchanged in the brain (Fig. S3B,C). Pituitary transcript levels of the gene encoding the common alpha subunit (Cga) for TSH, FSH and LH, the thyrotropin-releasing hormone β-subunit (Tshb), and the gonadotropin β-subunits (Fshb and Lhb), were not affected by Phf6 status (Fig. S3D-G). Trh mRNA levels were unaffected in the adult brain (Fig. S3H); however, gonadotropin releasing hormone (Gnrh1) transcript levels were slightly reduced in the brain of adult Phf6^−/Y versus control mice (Fig. S3I). These results indicate some evidence of a minor effect of PHF6 on the central hypothalamic-pituitary-gonadal axis in the form of an effect on brain Gnrh1 mRNA levels, in addition to reduced pituitary gland transcript levels of Prl and Pomc caused by loss of PHF6. However, these changes at the mRNA level did not manifest as corresponding changes to the circulating levels of gonadotropins, PRL or ACTH proteins in FVB/BALB/c mice.

Assessment of transcript levels of endocrine genes in the GH pathway in mice on the FVB/BALB/c background demonstrated a significant reduction in mRNA levels of the growth hormone-releasing hormone gene (Ghrh) in the brain of adult mice lacking PHF6 (Fig. 3A). On the other hand, loss of PHF6 had no effect on the expression of somatostatin (Sst), a negative regulator of GH release and secretion (Fig. 3B). In concordance with the reduced GH pituitary protein levels, transcripts of the growth hormone gene, Gh, in the pituitary gland were also significantly reduced both at postnatal day 14 (P14) and in adults (Fig. 3C). Transcript levels of the downstream effector gene of GH, Igf1, were reduced in the liver at P14 and in older mice in the absence of PHF6 (Phf6^−/Y) compared with controls (Fig. 3D). The absence of PHF6 (Phf6^−/Y) also resulted in a reduction in plasma IGF1 protein levels (Fig. 3E). These results indicate that downregulation of the GHRH/GH/IGF1 axis is likely to be responsible for the observed growth defect resulting from loss of PHF6.

The effect of PHF6 on transcription of GHRH/GH/IGF1 axis genes in the brain, pituitary gland and liver suggests that PHF6 could be playing a role in multiple tissues and at multiple levels of the GHRH/GH/IGF1 axis. To assess whether a neuroendocrine role of PHF6 was contributing to the growth defect, we employed the Nestin-cre transgene, which is expressed in the nervous system, including the hypothalamus, but not in the pituitary gland or the liver (Tronche et al., 1999). By crossing Phf6 (Phf6^lox) mice with mice expressing the Nestin-cre transgene (Nes-cre^Tg/+), we deleted Phf6 specifically in the nervous system on the C57BL/6 background (Fig. S4A,B). Males with nervous system deletion of Phf6 (Phf6^lox/Y;Nes-cre^Tg/+) survived postnatally on the C57BL/6 background, indicating the perinatal death observed in C57BL/6 Phf6^−/Y mice was not due to nervous system-specific effects. As the Nestin-cre transgene alone causes a mild postnatal growth defect (Declercq et al., 2015; Galichet et al., 2010), evident by 4 weeks of age (Fig. S4C), we compared Phf6^lox/Y;Nes-cre^Tg/+ mice with Nes-cre^Tg/+ controls to measure the effect of Phf6-deletion specifically. Phf6^lox/Y;Nes-cre^Tg/+ mice displayed a reduction in body weight compared with Phf6^+/Y;Nes-cre^Tg/+, with a statistically significant difference in multiple comparisons evident by postnatal week 3 (9.06±0.52 versus 5.13±0.30, respectively), corresponding to a 43% reduction in body weight (P<0.001; Fig. 4A). This effect was proportional, affecting overall body size (Fig. S4D). The effects of nervous system-specific deletion of Phf6 were accompanied by a reduction in Igf1 mRNA in the Phf6^lox/Y;Nes-cre^Tg/+ liver compared with controls (Fig. 4B). The effects of nervous system-specific deletion of Phf6 on overall body growth and peripheral organs demonstrate that control of body growth by PHF6 is at least in part mediated by neuroendocrine signaling.

A reduction in anterior pituitary hormone protein in the pituitary gland in the absence of PHF6 (see GH in Fig. 2B,C) could result from a reduction in the number of hormone producing cells as a result of a defect in pituitary development or from a reduction in the amount of hormone produced per cell. To distinguish these two possibilities, we assessed the histology of the pituitary and GH, PRL, ACTH, TSH and FSH levels per cell by immunofluorescence in sections of Phf6^+/Y and Phf6^−/Y pituitary glands (Fig. 5, Fig. S5). The pituitary gland displayed a normal histological structure (Fig. 5A), but was smaller in Phf6^−/Y mice compared to controls (not shown). Likewise, the volume of the anterior lobe specifically was smaller in Phf6^−/Y mice than in control mice (Fig. 5B). However, the decrease in volume was proportional to the decrease in total body volume (Fig. 5C). The absence of PHF6 caused a reduction in GH immunofluorescence signal per cell and per animal (Fig. 5D-F), but not in the total cell number or the GH⁺ area per view field. Similarly, the PRL signal was reduced per cell Phf6^−/Y mice compared with controls, but not the PRL⁺ area (Fig. S5A). Only ACTH, in addition to being reduced per cell, appeared to be present in a smaller area of the tissue in Phf6^−/Y mice compared with controls (Fig. S5B). No significant differences were observed in TSH or FSH staining (Fig. S5C,D). These results suggest that PHF6 is required either directly or indirectly for the production of normal levels of GH, PRL and ACTH per cell, and possibly also for the development or survival of normal numbers of ACTH-producing cells.

Suppressor of cytokine signaling (SOCS) proteins provide important negative-feedback mechanisms for many signaling pathways. SOCS proteins act by binding to and blocking activation of phosphorylation sites and by serving as E3 ubiquitin ligase adapter proteins, committing signal cascade proteins to proteasomal degradation (Linossi et al., 2013; Rico-Bautista et al., 2006). SOCS2 is an important negative regulator of growth hormone receptor (GHR) signaling and, as such, Socs2^−/− mice display gigantism and elevated Igf1 levels in several tissues (Favre et al., 1999; Greenhalgh et al., 2005; Metcalf et al., 2000).

To investigate whether upregulation of the GHRH/GH/IGF1 pathway could rescue the growth defects of Phf6^−/Y mice, we generated Phf6;Socs2 double mutant mice and determined the effects of loss of SOCS2 on the Phf6^−/Y growth defect. We bred the Socs2^−/− mice for three generations onto the mixed FVB/BALB/c background. Strikingly, deletion of Socs2 partially rescued the growth defect resulting from loss of PHF6, with a significant increase detected by multiple comparisons between Phf6^−/Y; Socs2^+/+ and Phf6^−/Y;Socs2^−/− mice from 5 weeks of age onwards (11.71±1.65 g versus 16.98±1.30 g at 5 weeks, P=0.04; Fig. 6A, Table S8). After 8 weeks, there was no statistically significant difference between the double mutant Phf6^−/Y;Socs2^−/− and wild-type Phf6^+/Y;Socs2^+/+ mice, demonstrating a distinct rescue (27.78±1.13 g versus 24.42±0.73 g, respectively at 8 weeks, P=0.29; Fig. 6A, Table S8). Consistent with an effect on skeletal growth, deletion of Socs2 increased the femur length of Phf6^−/Y mice (Fig. 6B,C). These results indicate that the growth defects caused by loss of PHF6 can be reversed by elevated GH signaling.

An analysis of 19 individuals with BFLS showed that 47% of BFLS males have a height that is more than two standard deviations below the population average (Turner et al., 2004). Body size is controlled by both environmental and genetic factors (Grasgruber et al., 2014; Grunauer and Jorge, 2018; Lango Allen et al., 2010; Perkins et al., 2016). Genetic variants that impact body size typically have either a described role in cell proliferation (Bicknell et al., 2011a,b; Griffith et al., 2008; Guernsey et al., 2011; Klingseisen and Jackson, 2011; Rauch et al., 2008) or endocrine function (Abuzzahab et al., 2003; Pfäffle and Klammt, 2011; Phillips et al., 2001; Pugliese-Pires et al., 2011; Wajnrajch et al., 1996; Wit et al., 2016). In this article, we have shown that PHF6 is an important regulator of postnatal growth, and that deletion of germline Phf6 in mice models the short stature of individuals with BFLS. The proportional growth defect resulting from Phf6 deletion specifically within the nervous system indicates that PHF6 controls growth at least in part through neuroendocrine regulation.

We have demonstrated that the GHRH/GH/IGF1 axis is a key pathway downregulated by loss of PHF6. We identified reduced transcription of Ghrh in the brain, Gh in the pituitary gland and Igf1 in the liver. As GHRH regulates the transcription and secretion of Gh (Barinaga et al., 1983), which in turn regulates transcription of Igf1 (Lupu et al., 2001), it is possible that the effects on Gh and Igf1 mRNA could be an indirect downstream result of the reduction in Ghrh. Indeed, if the primary effect was GH or IGF1 reduction, increased Ghrh expression would be expected, owing to alleviation of the negative-feedback loop (Phelps et al., 1993). Furthermore, the effect of Phf6 deletion appears to be similar to the Ghrh (Alba and Salvatori, 2004), Ghrhr (Eicher and Beamer, 1976) and Ghr (Lupu et al., 2001; Meyer et al., 2004; Zhou et al., 1997) mutant mice, which display postnatal growth defects. In contrast, Igf1 mutant mice are already ∼60% smaller than controls at birth (Baker et al., 1993; Liu et al., 1993; Powell-Braxton et al., 1993). These observations, combined with the finding that nervous system-specific deletion of Phf6 is sufficient to cause a growth defect, are consistent with a primary effect of PHF6 on regulating Ghrh. Notwithstanding, it is possible that PHF6 may also play an additional role in regulating Gh expression in the pituitary gland and Igf1 in the liver, as Phf6 is also expressed in these tissues (Voss et al., 2007). Further work is required to establish the exact molecular mechanisms by which PHF6 regulates endocrine pathways. Interestingly, PHF6 has been co-immunoprecipitated with PHIP (BRWD2), a reader of methylated H3K4 (Morgan et al., 2017), which has been shown to potentiate the effects of IGF1 signaling (Podcheko et al., 2007). PHIP mutations have been described in humans with developmental delay and intellectual disability (Jansen et al., 2018; Webster et al., 2016), while mice lacking the full-length PHIP protein display a postnatal growth defect (Li et al., 2010). Therefore, investigation into whether PHF6 interacts with PHIP to control the GH/IGF1 axis may offer insights towards our understanding of the molecular details of how PHF6 controls the GH/IGF1 pathway.

The rescue of the growth defect in Phf6^−/Y mice by additional deletion of Socs2 demonstrates that loss of negative regulation of GH signaling is able to override the effects caused by loss of PHF6, supporting the hypothesis that PHF6 regulates the GHRH/GH/IGF1 axis. We conclude that the level of GH in the absence of PHF6 must be sufficient to enable the lack of negative regulation by SOCS2 to rescue the growth defect (in contrast to the Ghrhr knockout ‘little’ mouse, in which the almost complete absence of GH renders mice unresponsive to deletion of Socs2) (Greenhalgh et al., 2005). These results provide a proof of principle that the growth defect caused by Phf6 mutation can potentially be rescued with GH treatment. However, it will be important for clinicians to consider whether the potential benefits of GH treatment for humans with BFLS will outweigh the risks (Allen, 2006; Radcliffe et al., 2004). A 9-year-old female BFLS patient recently treated with daily growth hormone injections developed edema after 3 weeks of treatment (Zhang et al., 2019). Furthermore, treatment with growth enhancers must be carefully considered in the context of PHF6 mutations being associated with leukemia (Van Vlierberghe et al., 2010). Two individuals with BFLS have developed hematopoietic malignancy (Carter et al., 2009; Chao et al., 2010). PHF6 regulates hematopoietic stem and progenitor cells (McRae et al., 2019; Wendorff et al., 2019), and notably, mice with germline or blood-specific mutation of Phf6 have an accelerated onset of hematopoietic malignancy compared with controls (McRae et al., 2019), establishing PHF6 as a bona fide tumor suppressor. This indicates that other BFLS patients may be susceptible to leukemia. Although long-term studies have found no overall increased risk of leukemia with GH treatment in individuals with no known risk factors (Allen et al., 1997; Stochholm and Kiess, 2018), GH-treated patients with predisposing factors may be at higher risk (Nishi et al., 1999; Watanabe et al., 1993, 1988); thus, endocrine therapies for BFLS patients should be approached with consideration of these additional factors.

Despite having not observed a significant effect on PRL or ACTH secretion, we found that Prl and Pomc transcripts were downregulated in the pituitary gland and that PRL and ACTH levels per cell were reduced. Similarly, although we did not observe an effect on FSH or LH secretion, nor on mRNA levels, the levels of Gnrh1 mRNA in the hypothalamus were reduced. The clinical signs of BFLS that these pathways may be related to (such as genital hypoplasia and gynecomastia) do not appear to manifest in mice. Any effect of PHF6 on these pathways at the transcriptional level appears to be compensated for, resulting in normal secretion levels in mice. Whether these pathways are important in the pathogenesis of human BFLS will be an important avenue for further research.

At least three infants with BFLS have been treated for failure to thrive and a further two developed hypoglycaemia in the neonatal stage (Turner et al., 2004). Two deaths have been reported in childhood BFLS; one male at 2 years with severely retarded growth and another male at 4 years who presented with a respiratory illness (Birrell et al., 2003; Turner et al., 2004). No direct correlation between the specific PHF6 mutation and the severity of presentation has been defined for BFLS in humans (Turner et al., 2004). The survival of PHF6^C99F hypomorphic mutant mice on the C57BL/6 background (Cheng et al., 2018), in contrast to the perinatal death of C57BL/6 mice with loss of function Phf6 mutation reported here and in another loss of function Phf6 model (Cheng et al., 2018), suggests that there is indeed an effect of mutation type on phenotypic presentation. The survival of FVB/BALB/c Phf6^−/Y males shown here in contrast to the perinatal death of C57BL/6 Phf6^−/Y males implicates the genetic background as an important additional mediator of the severity of phenotype presentation. Therefore, the specific PHF6 mutation inherited in addition to the overall genetic makeup of the individual is likely to influence the presentation of clinical signs in BFLS. Environmental factors are also likely to mediate the impact of PHF6 mutations, as it has recently been shown that loss of PHF6 specifically in AgRP neurons has no effect on body weight in mice on an ad libitum diet; however, mice lacking PHF6 in AgRP neurons on an alternate fasting/feeding diet have reduced feeding and reduced body weight (Gan et al., 2020).

Overall, we have shown that PHF6 positively controls postnatal growth, at least in part through neuroendocrine signaling. We have demonstrated that PHF6 is an important regulator of the GHRH/GH/IGF1 axis; specifically, of levels of pituitary GH and secreted GH, and of Ghrh, Gh and Igf1 transcripts. This study provides a better understanding of the etiology of BFLS and further insight into the genetic control of the endocrine system.

The Phf6^lox targeted deletion allele has been previously described (McRae et al., 2019) and was used in combination with cre-deleter mice (Schwenk et al., 1995) for germline deletion. The Nestin-cre transgene (Tronche et al., 1999) was crossed to Phf6 mice with the Phf6^lox targeted deletion allele mice maintained on a C57BL/6 background. As hemizygous Phf6 deletion is perinatal lethal on a C57BL/6 background (McRae et al., 2019), mice were crossed to F1 hybrid FVB/BALB/c mice, as carried out previously (Thomas et al., 2006), and kept on a mixed genetic background with 50% FVB and 50% BALB/c contribution by crossing to F1 hybrid FVB/BALB/c mice in every generation. C57BL/6 mice with the Socs2 null allele (Metcalf et al., 2000) were crossed for at least three generations to F1 FVB/BALB/c mice prior to experimental use. Genotyping was performed using primers as described in Table S9. All experimental procedures complied with the animal ethics regulations of the NHMRC of Australia and with approval of the WEHI Animal Ethics Committee.

No animals were excluded. Where possible, automated assessment was used to avoid biases, e.g. multiplex hormone assays, automated image analysis (micro-CT, immunofluorescence), densitometry of western blots and RT-qPCR.

The sampling procedure, adapted from Steyn and colleagues (Steyn et al., 2011) used a modified version of the tail-clip blood collection procedure (Abatan et al., 2008). A scalpel blade was used to create a small (1-2 mm) cut at the tail tip, followed by the collection of 4 µl of whole blood using a P20 pipette from the tail tip every 10 min over 6 h. Blood (4 µl) was added to 22 µl of 3.38 mM EDTA in PBS, immediately spun for 5 min at 2000 g to isolate the supernatant, which was snap frozen on dry ice before being stored at −80°C until analysis. The experiment was repeated a total of three times, with each experiment involving one Phf6^−/Y animal and a Phf6^+/Y littermate control.

Mice were imaged using a Bruker Skyscan 1276 Micro-CT system. Imaging parameters were set to balance contrast with acquisition time and resolution, and images were acquired with 41 µm³ voxels. The x-ray projection images were reconstructed into 3D volumes using Bruker's NRecon software, before being analyzed in FIJI using a custom-written macro. Briefly, the images were run through a series of intensity and morphology filters, before being thresholded to remove signal from bone, skin and other organs. Total volumes of the animal, as well as the adipose tissue were recorded.

The multiplex pituitary hormone assay (MPTMAG-49K, Millipore) with beads to ACTH, FSH, GH, prolactin, TSH and LH was performed as per the manufacturer's instructions, in technical duplicates for each time point. The plates were read on the Luminex 200 3.1 xPONENT System. Standard curves and concentration values were obtained using the Milliplex Analyst 5.1 Software. Technical duplicates were averaged prior to deconvolution analysis and only excluded in the rare cases where no value for that analyte was returned from the multiplex assay. Precision values for the AutoDecon program were calculated as described (Johnson et al., 2008) using the minimal detectable concentration and the coefficient of variation values provided by the multiplex kit manufacturer. The AutoDecon algorithm (Johnson et al., 2008, 2009) was run independently for each animal and hormone with the peak triage P-value set to 0.15 and software-estimated values for the secretion standard deviation and the half-life. Serum IGF1 levels were measured using a Quantikine ELISA kit for IGF1 (R&D Systems, MG100) according to the manufacturer's instructions.

Pituitary gland lysates were derived from micro dissected pituitary glands using KALB lysis buffer [1% Triton X, 1 mM EDTA, 150 mM NaCl, 50 mM Tris HCl (pH 7.4), 0.02% NaN₃] and proteinase inhibitors (Roche). Denatured lysates were separated by electrophoresis on a 4-12% NuPAGE Bis-Tris gel, and transferred onto a PVDF membrane (Osmonics). Membranes were blocked with 5% skim-milk/PBST, probed with anti-GH IgY antibody (XW-7687, Pro-Sci) followed by anti-IgY-HRP secondary (sc-2428, Santa Cruz) or anti-actin-HRP (sc-1616, Santa Cruz). Blots were developed on film using chemiluminescence and scanned for densitometry.

Pituitary glands were dissected and positioned horizontally in cryomolds (Tissue-Tek, Y557) pre-filled with OCT mounting medium (Tissue-Tek, IA018) then immediately snap-frozen on dry ice before storage at −80°C. Specimens were sectioned in 5 µm sections using a Microm HM525 Cryostat and placed on Superfrost positively charged slides, air-dried briefly, then placed on dry ice and stored at −80°C.

For immunofluorescence staining, frozen sections were fixed in 4% (w/v) PFA (Sigma, 158127-500G) in phosphate-buffered saline (PBS) for 30 min, followed by permeabilization in 0.2% (v/v) Triton-X-100 (Sigma, T8787), then blocking in PBSG [0.25% (w/v) gelatine (Sigma, G-1890) in PBS] for 15 min each. Sections were incubated with primary antibody diluted in PBSG for 1 h in a wet chamber at room temperature. Antibodies used were raised against GH (XW-7687, Pro-Sci; 1:100), ACTH (orb10034, Biorbyt; 1:25), prolactin (orb4237, Biorbyt; 1:20), FSH (MA5-14711, Thermo Fisher; 1:50) and TSH (LS-C50523, Lifespan Biosciences; 1:100).

Thereafter, sections were washed three times in PBS for 5 min each, then incubated with secondary antibody for 1 h. Sections were rinsed in PBS and counterstained with 0.1 µg/ml of DAPI (Sigma, 10236276001) for 5 min at room temperature, rinsed in PBS then mounted with Dako fluorescence mounting medium (Dako, S3032). Slides were stored in the dark at 4°C and imaged 1-3 days after staining. Images were acquired on a compound fluorescence microscope (Axioplan 2, Zeiss) using a digital camera (AxioCam HR, Zeiss). Mean fluorescence intensity per cell was determined in areas of 112,230 µm² to 142,800 µm² per animal (corresponding to 3 to 4 40× view fields) using Image J (version 2.0.0-rc-39/1.50b). Grayscale images were thresholded, and hormone⁺ and DAPI⁺ particles of at least 3.2 µm² were measured.

Liver, brain, kidney and testis RNA were isolated by guanidine hydrochloride extraction and differential precipitation as described previously (Thomas and Dziadek, 1993). Pituitary gland RNA was isolated using an RNeasy mini kit (Qiagen) as per the manufacturer's instructions. RNA quality was assayed using the Agilent 2100 Bioanalyzer or the Agilent 4200 Tapestation, and samples with an RNA integrity number of less than 7 were not used for analysis. cDNA synthesis was performed using Superscript III (18080-085, Invitrogen) according to the manufacturer's instructions.

RNA (10 μg) was denatured in RNA loading buffer [200 μl 10× MOPS buffer (pH 7.0), 350 μl formaldehyde, 1000 μl formamide] at 55°C for 15 min. Samples were loaded with 2 μl of gel loading dye [50% (v/v) glycerol, 1 mM EDTA (pH 8.0), 0.25% bromophenol blue and 0.2% xyan cyanol in MQ-H₂O] and separated by gel electrophoresis (1% ultra-pure agarose, 18% formaldehyde, 10% MOPS buffer, recirculating) at 40 V overnight. Following separation, the gel was washed, then stained in 0.5 μg/ml ethidium bromide. The gel was de-stained in distilled H₂O then photographed using an ultraviolet trans-illuminator. RNA was transferred to a Hybond-N nylon membrane (Amersham Pharmacia Biotech) overnight, dried in an oven at 80°C for 1 h and cross-linked with 50 mJ of UV-light using the GS gene linker (BioRad, Hercules). A Phf6 cDNA template corresponding to a region within exon 11, bases 2377 to 3156 of the published Phf6 sequence (GenBank Accession Number NM_027642) was incubated with in vitro DNA polymerase reagents (Megaprime DNA Labelling System, Amersham Biosciences) and ³²P-dCTP (EasyTides, PerkinElmer). Synthesized radiolabeled probes were added to 10 ml of preheated hybridization buffer and incubated with the nylon membrane at 65°C overnight. After washing steps the membrane was exposed to X-ray film for 7 days at −70°C.

Genomic DNA or cDNA was added to 10 μl, SensiMix SYBR (Bioline Reagents), 10 pmol each of a forward and reverse primer (Tables S10 and S11) designed using Primer3 (Koressaar and Remm, 2007; Untergasser et al., 2012), and 6 μl MilliQ-H₂O in a LightCycler 480 384-well Plate (Roche). Data were acquired on the LightCycler 480 (Roche) under the following conditions: one cycle of 95°C for 10 min; and 40 cycles of 95°C for 20 s, 60°C for 20 s, 72°C for 30 s. Standard curves based on arbitrary units corresponding to the Cp values obtained from serially diluted standard samples were used to calculate the initial concentration of the unknown samples. All samples were assayed in technical triplicates that were averaged prior to normalization with housekeeping gene(s): B2m for genomic DNA, and Gapdh and/or Pgk1 for RT-qPCR. A triplicate measurement was excluded in the rare case where there was a greater than 20% difference compared with the other two replicates for a particular sample (and the average of the other two replicates only was used for analysis).

Skeletal preparations were performed as previously described (Voss et al., 2009), using a protocol adapted from Kaufman (1992). Specifically, adult carcasses were skinned, fixed in ethanol for 7 days, followed by acetone for 7 days. Subsequently, specimens were stained for 10 days in a solution containing 1 volume of 0.4% Alcian Blue in 70% ethanol, 1 volume of glacial acetic acid, 14 volumes of 95% ethanol, 4 volumes of H₂O and 0.2 volumes of 0.5% Alizarin Red S in H₂O. Samples were rinsed, then kept in a solution of 20% glycerol and 1% potassium hydroxide for 3 days at 37°C, then for 4 days at room temperature. Skeletons were transferred to 50% glycerol prior to imaging and femur length quantification.

Statistical analysis, including unpaired (paired in Table 1) t-tests, Chi-squared tests and ANOVA, were performed using Prism 8.

We thank F. Dabrowski, G. Panoschi, L. Johnson, L. Wilkins and WEHI Bioservices for assistance with animal experiments; M. Corbett and M. Dixon for creating the Phf6 targeted construct; and J. Chue for preliminary work on the Phf6 mutant mice. We thank R. E. May, H. Pehlivanoglu and L. Potenza for excellent technical support. We are grateful to E. Bandala-Sanchez and T. A. Nguyen for assistance with the multiplex assay, and to G. Schütz for providing the Nestin-cre transgenic mice.

The peer review history is available online at https://dev.biologists.org/lookup/doi/10.1242/dev.187021.reviewer-comments.pdf