Conditional Loss of Ikkα in Sp7/osterix+ Cells Has No Effect on Bone, but Leads to Cell Autonomous, Age-related Loss of Peripheral Fat


 NF-κB has been reported to both promote and inhibit bone formation. To further explore its role in osteolineage cells, we conditionally deleted IKKα, an upstream kinase required for non-canonical NF-κB activation, using Sp7/Osterix (Osx)-Cre. Surprisingly, we found no effect on either cancellous or cortical bone, even following mechanical loading. However, we noted that IKKα conditional knockout (cKO) mice began to lose body weight after 6 months of age with severe reductions in fat mass in geriatric animals. Low levels of recombination at the IKKα locus were detected in fat pads isolated from 15 month old cKO mice. To determine if these effects were mediated by unexpected deletion of IKKα in peripheral adipocytes, we looked for Osx-Cre-mediated recombination in fat using reporter mice, which showed increasing degrees of reporter activation in adipocytes with age up to 18 months. Since Osx-Cre-driven recombination in peripheral adipocytes increases over time, we conclude that loss of fat in aged cKO mice is most likely caused by progressive deficits of IKKα in adipocytes. To further explore the effect of IKKα loss on fat metabolism, we challenged mice with a high fat diet at 2 months of age, finding that cKO mice gained less weight and showed improved glucose metabolism, compared to littermate controls. Thus, Osx-Cre mediated recombination beyond bone, including within adipocytes, should be considered as a possible explanation for extraskeletal phenotypes, especially in aging and metabolic studies.


Introduction
Although NF-κB is primarily considered key to acute in ammatory responses, this is not universally true, particularly for the alternative or non-canonical pathway. Unlike the canonical pathway, which is activated in minutes and generally inactivated within hours, alternative NF-κB is induced over many hours and typically persists for days. NIK functions as a central signaling component in this pathway, orchestrating signals from multiple stimuli and activating downstream kinase IKKα. This triggers phosphorylation of p100 and its partial processing, subsequently leading to persistent activation of the p52/RelB transcriptional complex 1. Besides participating in in ammatory responses due to its activation in immune cells, alternative NF-κB is involved developmentally in lymph node organogenesis, via the stroma 2-4, and plays a cell-extrinsic role in myelopoiesis 5. Expression of NIK in intestinal epithelial cells controls specialized antigen-presenting cells in the gut 6. Outside of its effects on the immune system, alternative NF-κB has been shown to control pathologic angiogenesis via direct actions in endothelial cells 7. Upregulation of NIK in skeletal muscle occurs in patients with metabolic syndrome and decreases with weight loss after gastric bypass 8. Alternative NF-κB also plays a role in metabolism via direct actions in pancreatic beta cells and hepatocytes 9. Thus, there is ample evidence that the alternative NF-κB pathway is important in a variety of cell types and physiologic contexts.
Bone is a dynamic organ, maintained by the coordinated actions of osteoblasts, which produce bone matrix, osteoclasts, which degrade bone, and osteocytes, which act as mechanosensors directing osteoblast and osteoclast activities. Osteocytes differentiate from osteoblasts, and together these cell types comprise the osteolineage. Few studies, all employing global knockout models, have directly addressed the role of alternative NF-κB in osteoblasts 10-13. These displayed complex skeletal phenotypes, largely pointing to positive effects on bone mass with pathway inhibition. However, we recently explored the role of alternative NF-κB in bone using a constitutively active NIK allele (NT3) lacking a negative regulatory domain, expressed in the osteolineage, nding increased bone mass 14. In contrast, the effect of alternative NF-κB signaling in the osteoclast is well-established. We and others have previously shown that NIK, IKKα, and RelB support osteoclastogenesis, particularly during pathological osteolysis 13 , 15-20. Due to these direct effects of alternative NF-κB on osteoclasts and physiologic coupling between osteoclasts and osteoblasts, the cell autonomous role of this pathway in osteoblasts remains unclear.
To better understand the role of alternative NF-κB in bone formation, we set out to conditionally inhibit it in osteoblasts. As in the previous study with activated NIK [ref - Davis JBMR], we chose to target early osteoblasts for modulation of alternative NF-κB throughout the lifespan of osteoblasts and osteocytes.
The transcription factor Osterix, also known as Sp7, is upregulated as mesenchymal stromal cells become committed to the osteoblast lineage, and in adult mice, its expression in the skeleton is largely con ned to osteoblasts and most osteocytes 21 , 22. Therefore, the Osterix promoter has been widely used to drive Cre expression in many studies of bone 23 , 24. At the initiation of this study, mice with a conditional allele for deletion of NIK were not yet available, so we employed IKKα / mice, ablating the second kinase in the alternative NF-κB pathway.

Mice
Mice were communally housed in a pathogen-free barrier facility, with controlled temperature and 12-hour light/dark cycles. They had ad libitum access to fresh water and standard rodent chow (Purina 5058, St. Louis, MO, USA) unless otherwise indicated. Protocols were approved by Institutional Animal Studies Committee at Washington University School of Medicine (ASC protocols 20170025 and 19-1059) and all methods were performed in accordance with the relevant guidelines and regulations, including with ARRIVE guidelines.

Micro-computed tomography
The right tibia of mice was scanned by microCT in vivo (VivaCT 40, Scanco, Brüttisellen, Switzerland) at 10.5 mm resolution (70 kVp, 114 mA, 8W, 100ms integration time). Cancellous bone parameters were measured at a 1 mm region distal to the end of the tibial growth plate. Cortical measurements were made at the tibial mid-shaft (1 mm region de ned 5 mm proximal to the distal tibio bular junction). Bone indices are reported in accordance with established standards 41.
Dual-energy X-ray absorptiometry (DXA) Whole body DXA scans were performed on 18 month male CON and cKO mice using a Faxitron UltraFocus 100 machine (Faxitron, Buffalo Grove, IL, USA).

Mechanical loading and dynamic histomorphometry
Unilateral axial tibial compression (Electropulse 1000; Instron, Norwood, MA, USA) and dynamic histomorphometry (Bioquant Osteo software v18.2.6; Bioquant Image Analysis Corp., Nashville, TN, USA) was performed on the right tibiae of 16-week-old, male mice as previously described 14. The left tibia served as the contralateral non-loaded control. Strain gauging was performed to determine the force necessary for a 2000 microstrain deformation (9.6N, both genotypes). In samples where the mineral apposition rate was zero, an imputed value of 0.1 was used to allow for statistical comparisons. All measurements were acquired in a blinded fashion and reported in accordance with published standards 41.

Genomic DNA recombination
Long bones were ushed of marrow and crushed in TRIzol (15596026; Invitrogen, USA) using a Navy RINO lysis kit and Bullet Blender Tissue Homogenizer (Next Advance, Troy, NY, USA). Peripheral fat depots were processed similarly to crushed bone samples. Cell cultures were rinsed 2x with PBS, then lysed directly in TRIzol. Genomic DNA was extracted using a back extraction buffer (4M guanidine thiocyanate, 50mM sodium citrate, 1M Tris) and alcohol precipitation. 50ng of input gDNA per sample using GoTaq

High Fat Diet
Mice were fed a 60% kcal/fat diet (#D12492, Research Diets Inc, New Brunswick, NJ, USA), ad libitum, beginning at 8-9 weeks of age. Body weight was measured once a week. At sacri ce, peripheral fat pads were weighed after removal of any gross contaminating tissue.

EchoMRI
Lean and fat mass was assessed by EchoMRI body composition analysis (EchoMRI LLC, Houston TX, USA) in non-fasted mice. Each mouse was scanned twice and the average value was used for analysis.

Glucose and insulin tolerance tests
Mice were housed on aspen bedding and administered intraperitoneally a 1mg/g dose of dextrose or 0.75U/kg of insulin (HumulinR U-100, Eli Lily, USA) after a 6h or overnight fast (food only), as indicated. Tail vein blood was sampled at intervals over a 2 hour period and blood glucose was measured using a Bayer Contour meter (#9556C, Bayer HealthCare, Mishawaka, IN, USA) and accompanying test strips.

Immunostaining and Fluorescence Imaging
Inguinal fat was embedded in optimal cutting temperature compound (OCT; 23-730-571, Thermo Fisher Scienti c, USA) and ash frozen at -80°C. Embedded tissues were post xed in 10% neutral buffered formalin and cut at 50 µm on a cryostat (Leica, Buffalo Grove, IL, USA). For immunostaining, cut sections on glass slides were blocked in 10% goat serum in Tris-NaCl-Tween (TNT) buffer before incubation for 24 hours with primary antibodies; anti-perilipin 1 (N-terminus) guinea pig polyclonal (# LS-C665927-

Statistical Analysis
All statistics were computed using GraphPad Prism software (Version 9.2.0, GraphPad Software, Inc., La Jolla, CA, USA). Values of p<0.05 were considered signi cant and data are presented as mean ± SD. For pairwise comparisons, either a student's, unpaired, two-tailed t-test or student's paired, two-tailed t-test was used. A Welch's correction was applied to the t-test if the F-test to compare variances was signi cantly different. For multiple group comparisons, a 2-way ANOVA followed by Tukey's multiple comparison test was performed. For repeated measure group comparisons, a 2-way repeated measures ANOVA followed by Sidak's multiple comparison test was performed. A mixed-effect model was used if sample sizes differed. Speci c statistical tests and sample sizes are indicated in the respective gure legends.

Conditional deletion of IKKα in the osteoblast lineage does not alter bone mass
We mated IKKα / and Osx-cre mice to generate Osx-Cre;IKKα / (cKO) and Cre-negative (CON) littermates. Dams were kept on doxycycline throughout pregnancy and until pups were weaned to prevent early Cre expression that can impact skeletal growth. Activation of Cre and excision of the oxed IKKα allele were assessed in bone marrow-derived mesenchymal stromal cell cultures (BMSCs) under osteogenic conditions and in ushed, crushed bones from cKO and CON littermates. Cre expression was undetectable in unstimulated BMSCs and rose during osteogenesis ( Fig S1A). Interestingly, however, recombination of the IKKα allele was robust, even prior to addition of osteogenic media (Fig S1B). Importantly, Cre expression and recombination did not occur in CON bones but were readily detected in cKO samples (Fig S1C,D). We next assessed the effect of IKKα de ciency on osteogenesis in vitro and found a modest increase in mineralization as well as expression of osteoblast markers (Fig S2).
We used in vivo microCT to screen for bone effects in male mice at multiple ages and found no differences in cortical or cancellous parameters in the tibia at any time (Fig. 1a,b, S3-6). Aged males were also screened by DXA, but again no differences in bone mass were seen (Fig. 1c). In females, ex vivo microCT at 16 weeks of age also failed to detect bone changes after IKKα deletion (Fig S7). To determine if an anabolic condition would elicit a bone-speci c role for IKKα, we applied unilateral tibial compression for 2 weeks in 16 week old male mice. However, again, we saw no differences by genotype (Fig. 2). We therefore concluded that IKKα plays little or no role in osteogenesis in basal or mechanical loading conditions.
Aging IKKα cKO mice lose fat, associated with improved glucose metabolism As we aged mice to examine their bone phenotype, we noted that the cKO mice appeared smaller. In fact, analysis of body weights showed not only lower weights for the cKO cohorts overall at both 12 and 18 months, but also a 13% decrease in weight in the same animals between 6 and 12 months of age, compared to a 9% increase over the same period in CON (Fig. 3a,b). EchoMRI was then used to quantitate fat and lean mass. cKO mice displayed distinctly lower fat mass at 12 and 18 months (47% and 62% respectively), with a more modest, but still statistically signi cant, decrease in lean mass (Fig. 3c,d). Post mortem, both gonadal and inguinal fat pads were smaller in cKO than CON mice (Fig. 3e,f).
Because reductions in fat are often accompanied by changes in glucose metabolism, we challenged mice with glucose tolerance testing (GTT) after a 6h fast. At 7-9 mo of age, the response to glucose challenge was nearly identical (Fig. 4a). By middle-age (13-15 mo), cKO mice showed a mild but signi cant decrease in peak blood glucose levels (Fig. 4b). Old (18-20 mo) cKO animals displayed a trend towards lower peak glucose (Fig. 4c), and this difference was further accentuated after an overnight fast (Fig. 4d), indicating improved glucose tolerance in old cKO mice.
Osx-Cre mediates increasing recombination in adipocytes with age Based on these changes in fat mass and glucose metabolism, we next sought to determine if the IKKα oxed locus was recombined in cKO fat. We performed PCR to detect the deleted allele in inguinal, gonadal, renal, and brown fat pads and observed the recombination product at multiple sites in cKO animals, but not CON (Fig. 5a). Fat tissue contains many cell types besides adipocytes, including endothelium and other vascular components, as well as hematopoietic cells such as macrophages. To speci cally examine Osx-Cre driven recombination, we examined fat pads from Osx-Cre;TdTomato (TdT) reporter mice raised in the same manner as the IKKα cKO mice, using confocal microscopy. We found limited reporter expression in inguinal fat at 6 and 12 weeks of age, with substantially increased signal at 6 and especially 18 months (Fig. 5b). The pattern of TdT coincided with immunostaining for perilipin, demonstrating that recombination occurs in adipocytes. In contrast, we found no co-staining of TdT with CD45 (Fig. 5c), indicating that recombination in hematopoietic cells in fat is unlikely to be responsible for the observed low-fat phenotype. Based on this data, we conclude that Osx-Cre drives recombination in peripheral adipocytes, in an age-dependent manner.

IKKα cKO mice show a blunted response to high fat diet
To further investigate the fat phenotype observed in aged cKO mice, we deployed a well-established high fat diet (HFD) model in younger animals. Mice were fed this diet (60%kcal/fat) beginning at 8 weeks of age, prior to any difference in body weight between CON and cKO mice ( Fig. 6a and S8a). Male cKO mice showed blunted weight gain which did not reach statistical signi cance following 8 weeks of HFD, although total fat mass was decreased by ECHO MRI (S8). Intriguingly, attenuation of weight gain in female cKO mice was signi cant (60% increase in CON vs 40% increase in cKO) (Figure 6b), associated with a marked reduction in fat mass (Figure 6c). Total lean mass was unchanged in both sexes, compared to CON (Figure 6d and S8d).
After 8-11 weeks on HFD, male cKO mice showed a trend towards better glucose tolerance, but this did not reach statistical signi cance in our cohorts (S8d). Glucose tolerance in females after 8-11 weeks similarly trended better in cKO mice (Fig. 7a,b). Insulin tolerance was not signi cantly different in either sex at this point, although female cKO showed a slight trend towards better tolerance (S8f and Fig. 7c,d). We then decided to continue HFD in a subset of females for 20-23 weeks. After this extended period, body weight continued to be lower in cKO, and both glucose and insulin tolerance were signi cantly improved compared to CON (Fig. 7e-i). Because signi cant changes in weight and fat mass occurred long before alterations in glucose metabolism, it is likely that the metabolic effects are secondary to, rather than the primary drivers of, weight gain.

Discussion
In this study, we set out to examine the role of alternative NF-κB signaling in the osteolineage by targeting a key upstream kinase, IKKα, using Osx-cre. Previously, we found that forced activation of this pathway using a constitutively active allele of NIK with the same Osx-cre driver enhanced both basal and stimulated bone formation 14. Here, male cKO mice showed no differences in bone mass up to 18 months of age, and mechanical loading by tibial compression failed to generate any differences in bone formation. Female mice also had normal bone mass at 4 months of age. Thus, under basal and nonin ammatory loading conditions, IKKα does not seem to have a clear role, either positive or negative, in bone formation. One limitation of our bone analysis is that we did not follow females over time, despite our nding that global loss of alternative NF-κB components NIK and RelB has greater effect in females.
However, in those models, the differences are clear by 10 weeks of age. It is possible that IKKα may impact bone formation in the context of strong in ammatory stimuli such as in ammatory arthritis models, which were not examined here.
Although Osx-Cre e ciently drives recombination in osteoblasts, not all phenotypes identi ed can be attributed to effects in the osteolineage. The Osx-Cre allele has established recombination activity in many extraskeletal tissues including synovium 25, intestinal epithelium 26 , 27,, and kidney 28, as well as in some hematopoietic stem and progenitor cells 29. In tumor bearing mice, Osterix, as well as the Osx-Cre allele, is also expressed in a subset of cancer-associated broblasts with a dual broblast/osteogenic signature 29. Recently, we found subcutaneous sarcomas, but not bone tumors, in mice expressing a transgene driven by Osterix-cre 30. Therefore, given the striking loss of fat in aging cKO mice in the absence of any changes in bone, we considered the possibility that recombination outside of bone was responsible for the phenotype. PCR of genomic DNA from peripheral fat depots in middle aged mice showed some recombination of the IKKα allele, albeit less robust than in bone. Examination of inguinal fat pads from reporter mice showed robust TdT expression only in adipocytes from aged mice. Despite identi cation of TdT in several subsets of CD45+ cells in a previous study using the same line of Osx-Cre;TdT mice 29, we did not identify any such cells in the inguinal fat sections at any age. We therefore concluded that Osx-Cre can drive recombination in adipocytes, and this cell autonomous effect is the most likely cause of the cKO fat phenotype. Future experiments using an adipose-speci c cre driver are required to con rm this and would be useful to determine the mechanism by which IKKα impacts adipocyte metabolism.
In addition to the unexpected nding of a fat phenotype driven by Osx-Cre, we also did not anticipate the strong effect of age on activation of this Cre. Previous studies utilizing this Cre driver primarily utilize mice under 6 months of age, and the few studies with mice at or beyond 1 year of age did not describe, or speci cally look for, Cre expression outside of bone 31-33. More comprehensive analysis of conditional mouse models with sensitive reporters like TdTomato is likely to uncover other so-called off-target effects that arise with aging in many Cre lines.
Most studies of the role of NF-κB in adipocytes have focused on IKKβ, an apex kinase in the canonical pathway. Loss of IKKβ in mature adipocytes using Adiponectin-Cre or Adipoq-Cre did not affect body weight in normal or HFD conditions 34 , 35. However, although fat mass was decreased in these models, the mice were insulin resistant on HFD. In contrast, with conditional loss of IKKβ in adipocyte progenitors, lower fat mass on HFD was accompanied by improved glucose homeostasis 36. Constitutive activation of IKKβ in adipocytes prevented diet-induced obesity and improved glucose tolerance 37. Interestingly, IKKβ in adipocytes may act via direct phosphorylation of targets such as β-catenin and BAD 34 , 38, rather than via activation of canonical NF-κB. Thus, the role of IKKβ in the adipocyte lineage is complex. Furthermore, the effects of aging have not been studied in any of these models.
Data on the role of IKKα or other components of alternative NF-κB in metabolism is even more limited, with effects described in pancreatic islets, hepatocytes, and skeletal muscle 9. However, there has been an absence of conditional knockout studies for the pathway in peripheral adipocytes. Our cKO model, although not initially designed to target fat or metabolism, nevertheless highlights the importance of IKKα in peripheral fat, both with aging and diet-induced obesity. Whether IKKα in adipocytes acts through alternative NF-κB signaling, via other downstream kinase targets, or in a kinase-independent fashion 39, remains to be established.
Since this study was undertaken to examine the role of IKKα in bone, studies of fat and metabolism were not initially planned. By the time the fat phenotype in aging males was discovered, it was not practical to undertake a similar study in females. Therefore, we decided to investigate whether differences in fat could be accelerated using HFD in both sexes. Interestingly, although both male and female cKO mice had less fat than CON after 8 weeks on the obesogenic diet, the difference in overall body weight was more pronounced in females. Thus, like osteoclasts 13, adipocytes may have differential sensitivity to the alternative NF-κB pathway by sex.
In sum, using an Osx-Cre driven conditional knockout approach, we found no clear role for IKKα in the osteolineage in either basal or mechanically stimulated conditions, but rather an intriguing role for IKKα in fat accumulation with age or diet-induced obesity. Although further experiments speci cally targeting adipocytes are needed, the results shown here suggest that inhibition of IKKα, or potentially other alternative NF-κB pathway components, may reduce fat deposition with age and improve glucose metabolism. With increasing recognition of bone as an endocrine organ, it is tempting to conclude that phenotypes arising from conditional alleles driven by Osx-Cre are due to the osteolineage. However  cKO = red, all male mice. Right tibiae were loaded (Load, triangles) and left tibiae served as non-loaded (NL, circles) controls. Results are presented as mean SD. n=12 per genotype. 2-way ANOVA followed by Tukey's multiple comparison test; response to load within genotypes. There were no signi cant differences between genotypes, or interactions between genotype and loading.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.