Osteoblastic Swedish mutant APP expedites brain deficits by inducing endoplasmic reticulum stress-driven senescence

Patients with Alzheimer’s disease (AD) often have osteoporosis or osteopenia. However, their direct link and relationship remain largely unclear. Previous studies have detected osteoporotic deficits in young adult Tg2576 and TgAPPsweOCN mice, which express APPswe (Swedish mutant) ubiquitously and selectively in osteoblast (OB)-lineage cells. This raises the question, whether osteoblastic APPswe contributes to AD development. Here, we provide evidence that TgAPPsweOCN mice also exhibit AD-relevant brain pathologies and behavior phenotypes. Some brain pathologies include age-dependent and regional-selective increases in glial activation and pro-inflammatory cytokines, which are accompanied by behavioral phenotypes such as anxiety, depression, and altered learning and memory. Further cellular studies suggest that APPswe, but not APPwt or APPlon (London mutant), in OB-lineage cells induces endoplasmic reticulum-stress driven senescence, driving systemic and cortex inflammation as well as behavioral changes in 6-month-old TgAPPsweOCN mice. These results therefore reveal an unrecognized function of osteoblastic APPswe to brain axis in AD development.

A lzheimer's disease (AD) is the most common form of dementia. It is pathologically characterized by cortical and cerebrovascular β-amyloid (Aβ) plaques, phosphor-tau containing neurofibrillary tangles, reactive glial cell (astrocyte and microglial cell)-associated chronic brain inflammation, and neuron-loss 1,2 . Interestingly, in addition to brain pathologies, patients with AD, both early and late onset, often have osteopenia or osteoporosis 3-10 , a condition characterized by the loss of bonemass or bone mineral density (BMD) with micro-architectural deterioration of bone tissue, and a higher rate of hip fracture. However, little is known regarding the underlying mechanisms of AD association with bone loss.
A growing list of genetic risk genes has been identified in patients with early onset and late onset AD. Intriguingly, many of the AD risk genes, such as TREM2 (triggering receptors expressed on myeloid cells-2) and PYK2, are highly expressed in immune cells and bone cells, and encode proteins that regulate not only neuron synaptic functions, but also immune responses and bone homeostasis [11][12][13][14][15] . APOE, another AD risk gene, is also identified as a risk factor for osteoporosis [16][17][18] . Among the various risk genes for AD development, we chose Swedish mutant APP (APP swe ) to address the question regarding AD association with bone loss for the following reasons. The Swedish mutations in the APP gene are initially identified in patients with early-onset (EO) AD, which promote the generation of Aβ by favoring its proteolytic cleavage performed by βand γ-secretases [19][20][21] . Much research has focused on the impacts of Aβ on the brain, even though APP or APP swe is known to be expressed not only in the brain, but also in periphery tissues, including osteoblast (OB)lineage cells 22,23 . Although APP swe is only detected in a small fraction of AD patients, it is commonly used to generate AD animal models, such as Tg2576 and 5XFAD 24,25 . APP swe in these animal models (in particularly Tg2576) is expressed ubiquitously, in both the brain and periphery tissues, including OB cells 22,23 . While investigating the phenotypes of these APP swe -based animal models have provided valuable insights into Aβ brain pathology and impairments in mouse cognitive functions, the function of APP swe in peripheral tissues, such as OBs, remains poorly understood. Previous examinations of bone structures in Tg2576 mice have identified early-onset osteoporotic deficits, months before any brain-pathologic defect that was detected 22,23 . Knocking out App (in APP −/− mice), or selective expressing APPswe in osteocalcin (OCN) promoter driven Cre (OCN-Cre) + OB-lineage cells (in TgAPP swe OCN mice) recapitulates the osteoporotic defects in Tg2576 mice 23,26 . These observations raise an interesting question, could problems in the bone cells conversely contribute to AD pathology in the brain?
Here, we provide evidence that TgAPP swe OCN mice express APP swe largely in the OB-lineage cells, with little to weak expression in the dorsal dentate gyrus (dDG) of the hippocampus. These mice develop age-dependent [starting at 6month-old (MO)] and brain-region selective pathologies, and exhibit anxiety-and depression-like behaviors, as well as altered cognitive functions. While these mice at 6-MO showed brainpathy (including glial activations and elevated pro-inflammatory cytokines) largely in the cortex, these mice at 12-MO showed brain-pathy mainly detected in the hippocampus. Further mechanistic studies demonstrate that APP swe, but not APP wt or APP lon (London mutant), in OB-lineage cells increases endoplasmic reticulum (ER)-stress, senescence, and SASPs (senescence associated secretory phenotypes). Inhibition of ER-stress abolishes APP swe -induced senescence, and suppression of senescence diminishes brain and behavioral phenotypes in 6-MO TgAPP swe OCN mice. Taken together, these observations suggest that APP swe in OB-lineage cells contributes to the brain-region selective inflammation and glial activation and induces anxiety-and depression-like behaviors in age-dependent manner, which are largely due to elevated OB-senescence, SASPs, and systemic inflammation. These results thus uncover a link between APP swe in the OB-lineage cells and AD development.

Selective APP swe expression in OB-lineage cells in
TgAPP swe OCN mice. To investigate osteoblastic APP swe 's function in AD development, we took advantage of TgAPP swe OCN mice, in which human APP swe expression in LSL-hAPP swe mice depends on the removal of LSL by the OCN-Cre (Fig. 1a) 23 . Although OCN-Cre mice express Cre primarily in mature/adult OB-lineage cells 27,28 , our recent study showed Cre activity in neurons of dDG hippocampus, olfactory bulb, and cerebellum 29 . Thus, it is important to verify APP swe 's expression in bone cells and brain tissues of TgAPP swe OCN mice. Notice that the hAPP swe protein was detected in the OB-lineage BMSCs (bone marrow stromal cells), but not in the hippocampus or cortex of the TgAPP swe OCN mice (6-MO) (Fig. 1b, c). We then asked if this is due to hAP-P swe 's cleavage (to produce Aβ 40 or Aβ 42) in the brain tissues. ELISA measuring human Aβ 40 and Aβ 42 levels showed little-tono Aβ increase in the hippocampus, cortex, or serum samples (Fig. 1d, e); but slight increases of both Aβ 40 and Aβ 42 in the OBlineage cells, of TgAPP swe OCN mice (6-MO), as well as in the brain tissues and serum samples of 6-MO Tg2576 mice (Fig. 1d, e). These results eliminate the possibility of βand γ-cleavages of hAPP swe in the brain of 6-MO TgAPP swe OCN mice, suggesting little hAPP swe expression in the mutant brain at this age. We further tested this view by RT-PCR analysis of hAPP swe 's transcripts in the mutant mice. Using specific primers for human APP, a weak hAPP swe expression (~1.5 fold over control) was detected in the TgAPP swe OCN brain regions (e.g., hippocampus, olfactory bulb, and cerebellum) where OCN-Cre is expressed 29 , but not in the OCN-Cre negative cortex (Fig. 1f). Notice that the hAPP swe 's transcripts were much more abundant in the BMSCs (~70 fold over control) than in the brain (Fig. 1f), implying a much weaker Cre activity in neurons than in OB-lineage cells of the OCN-Cre mice. This viewpoint is consistent with the RT-PCR findings that Cre is expressed largely in the OB-lineage cells (~128 fold over control), weakly (~18 fold over control) in the hippocampus, and undetectable in the cortex of OCN-Cre mice (Fig. 1g). Taken together, these results suggest that the hAPP swe is highly expressed in OCN-Cre + OB-lineage cells, but little to weakly expressed in the OCN-Cre + dDG, olfactory bulb, and cerebellum neurons, of TgAPP swe OCN mice.
Age-dependent and brain region-selective elevations in reactive astrocytes, microglial cells, and inflammatory cytokines, and an impairment in DG neurogenesis in TgAPP swe OCN mice. We then addressed whether TgAPP swe OCN mice exhibit any brain pathology that is similar to those of APP swe -based AD animal models (e.g., Tg2576) 24,25,[32][33][34] , by performing the following studies.
First, we measured both Aβ 40 and Aβ 42 levels in the bone cells and brain tissues of TgAPP swe OCN mice at ages of not only 6-MO, but also 12-MO. Although little Aβ 40 or Aβ 42 levels were detected in 6-MO TgAPP swe OCN cortex and hippocampus (Fig. 1d, e), Aβ 42 , but not Aβ 40 , was slightly elevated in 12-MO TgAPP swe OCN hippocampus, but not cortex nor serum samples ( Supplementary  Fig. 1a, b). Additionally, little to no Aβ plaque was detected in 12-MO TgAPP swe OCN bone and brain sections, in contrast from brain sections from 5XFAD mice (4.5 MO) ( Supplementary Fig. 1c-e). These findings support the view for a weak hAPP swe /Aβ 42 expression in 12-MO TgAPP swe OCN hippocampal DG neurons.
Second, we examined neuronal distribution patterns and densities in the cortex and hippocampus of TgAPP swe OCN mice (at age of~7-MO) by conducting co-immunostaining analysis using antibodies against NeuN (a marker for all neurons) and Ctip2 (a marker for Layer V-VI neurons in the cortex and neurons in CA1-2 and DG). Little change in the NeuN + and Ctip2 + neuron distribution patterns and densities was detected in TgAPP swe OCN brains ( Supplementary Fig. 2).
Third, we assessed the morphologies and densities of glial cells, including Olig2 + oligodendrocytes, S100β + ependymal cells, GFAP + astrocytes, and IBA1 + microglial cells, in the brain sections of control (LSL-APP swe ) and TgAPP swe OCN mice. The Olig2 + oligodendrocytes and S100β + ependymal cells appeared to be unchanged in the TgAPP swe OCN cortex or brain (Supplementary Fig. 3). Intriguingly, both GFAP + astrocytes and IBA1 + microglial cells were increased in 6-MO TgAPP swe OCN cortex, particularly in layers I-III, but not in hippocampus ( Fig. 2a-d), suggesting a brain region-selective activation of these glial cells. This view was further verified through a Western blot analysis, which showed increased GFAP and IBA1 protein levels in 6-MO TgAPP swe OCN cortex, but not in hippocampus (Fig. 2e, f). Because glial cell activation is often associated with increased inflammation 30,31 , we examined expressions of inflammation associated cytokines (e.g., Il1b, Il6, Il10, and Tnfa), growth factors (e.g., Tgfb1 and Csf2), and proteinase (e.g., Mmp3) in both the cortex and hippocampus of control and TgAPP swe OCN mice (at 6-MO) using RT-PCR analysis. The transcripts of Il1b, Il10, Tnfa, and Mmp3 were all increased in TgAPP swe OCN cortex, but not in hippocampus (Fig. 2g, h), supporting the view of cortex as a vulnerable brain region in 6-MO TgAPP swe OCN mice.
Fourth, we found that the glial activation and inflammatory phenotypes in TgAPP swe OCN mice were not only brain-region selective, but also age-dependent. Whereas the cortex displayed the glial activation/inflammation in 6-MO TgAPP swe OCN mice, these phenotypes were not detected in 3-MO TgAPP swe OCN ( Supplementary Fig. 4), but evidently more obvious in 12-MO TgAPP swe OCN hippocampus than the cortex ( Supplementary  Fig. 5), suggesting age-dependent changes in the brain-region selectivity of the glial activation/inflammation phenotypes.
Finally, we examined adult neurogenesis in hippocampal DG (dentate gyrus), which is also impaired in AD animal models 32 . EdU was injected into the mice~12 h before sacrifice to label proliferative neural stem cells (NSCs). Hippocampal sections were coimmunostained EdU with antibodies against DCX (doublecortin) (a marker for newborn neurons derived from NSCs). While TgAPP swe OCN mice at 3-MO showed no difference in EdU + and DCX + cell densities compared to the controls, TgAPP swe OCN mice at 6-MO displayed significant reductions in EdU + and DCX + cell densities at both dorsal and ventral DG ( Supplementary Fig. 6), demonstrating an age-dependent impairment in the hippocampal DG neurogenesis of TgAPP swe OCN mice, similar to that described in AD animal models 32 .
In aggregate, TgAPP swe OCN mice (starting at 6-MO) exhibit partial AD relevant brain pathologies, which include increased  42 (e) levels in serum, BMSCs (50 μg in total protein), and brain homogenates including hippocampus and cortex (300 μg total protein) from 6-MO control, TgAPP swe OCN , and Tg2576 mice. f RT-PCR analysis of hAPP gene expression in BMSCs, olfactory bulb, cerebellum, hippocampus, and cortex of 6-MO control and TgAPP swe OCN mice. g RT-PCR analysis of Cre expression in BMSCs, hippocampus, and cortex of 6-MO control (LSL-APP swe ) and TgAPP swe OCN mice. All data were presented as mean ± SD. *p < 0.05, **p < 0.01, ***p < 0.001 (n = 3 mice). Mann-Whitney U test was used in c and g, and one-way ANOVA followed by Tukey post hoc test was used in d-f. reactive astrocytes, microglial cells, and inflammatory cytokines in the cortex (at 6-MO)/hippocampus (at 12-MO), impaired DG neurogenesis, and elevated Aβ 42 in 12-MO hippocampus.
Age-dependent anxiety-and depression-like behaviors in TgAPP swe OCN mice. Glial activation, brain inflammation, and decreased DG neurogenesis are often associated with depression-and/or anxiety-like behaviors [33][34][35][36][37][38] . We thus subjected TgAPP swe OCN and control mice to an open field test (OFT) for evaluation of TgAPP swe OCN mice's anxiety and locomotor activity. TgAPP swe OCN mice at 6-and 12-MO, but not 3-MO, showed reduced center duration time but comparable total distance traveled to the controls (Fig. 3a, b and Supplementary  Fig. 7a, b), suggesting a reduced exploratory, but not locomotor, activity, in the mutant mice, and implicating anxiety and/or depression. We further examined their behaviors using elevated plus maze test (EPMT) and light/dark transition test (LDT), both tests widely used to assess anxiety-related behavior in mouse models 39,40 . Indeed, TgAPP swe OCN mice, again at 6-and 12-MO, but not 3-MO, showed decreased open arm duration time and entries by EPMT (Fig. 3c, d and Supplementary  Fig. 7c, d), and reduced time in light box room in the LDT ( Fig. 3e and Supplementary Fig. 7e), supporting the view for anxiety-like behaviors. We then assessed their depression-like behaviors using tail suspension test (TST), force swimming test (FST), and sucrose preference test (SPT). TgAPP swe OCN mice (6and 12-MO, but not 3-MO) appeared to be depressed, exhibiting increased immobility times in both TST ( Fig. 3f and Supplementary Fig. 7f) and FST ( Fig. 3g and Supplementary Fig. 7g) and reduced sucrose preference ( Fig. 3h and Supplementary   Fig. 7h). Together, these results suggest that TgAPP swe OCN mice experience age-dependent (starting at 6-MO) anxiety-and depression-like behaviors.
Since hAPP swe is weakly expressed in dDG neurons in TgAPP swe OCN hippocampus (Fig. 1f), we wondered whether such a weak dDG expression of APP swe could induce similar behavior phenotypes to that in TgAPP swe OCN mice. The AAV-CamkII-Cre (Cre under the control of CamkII promotor for excitatory neuron expression) and AAV-CamkII-GFP (as control) were specifically injected into the dDGs of both sides of the hippocampus in LSL-APP swe mice (at age of 4-MO); and mice at 6-MO were subjected to the behavior tests ( Supplementary Fig. 8a). While dDG neurons in LSL-APP swe mice were successfully infected with the viruses (indicated by the GFP, hAPP swe expression, and Aβ 42 increase) (Supplementary Fig. 8b-g), little to no differences in behavior tests using EMPT, LDT, TST, FST, and SPT were detected between Cre and GFP virus injected mice (Supplementary Fig. 8h-l), unlike the TgAPP swe OCN mice. These results thus implicate that the anxiety-or depression-like behaviors in 6-MO TgAPP swe OCN mice are in large due to APP swe 's expression in OBlineage cells, but not dDG neurons.
Age-dependent alterations in spatial learning and memory in TgAPP swe OCN mice. Although anxiety-and depression-like behaviors are present in AD animal models (e.g., Tg2576 and 5XFAD) [41][42][43] and AD patients [44][45][46] , a key AD relevant functional deficit is the age-dependent cognition decline 47,48 . Therefore, we subjected TgAPP swe OCN and control (LSL-APP swe ) mice to the Morris water maze (MWM) test (to access mouse spatial learning and memory function) 49 , and the novel object recognition (NOR) Fig. 2 Elevated reactive astrocytes, microglial cells, and inflammatory cytokines in 6-MO TgAPP swe OCN cortex, but not hippocampus. a Representative images of co-immunostaining with IBA1 (green), GFAP (magenta), and DAPI (blue) of hippocampal sections from 6-MO control (LSL-APP swe ) and TgAPP swe OCN mice. Scale bars: 200 µm (upper) and 20 µm (lower). b Quantification of data in a. c Representative images of co-immunostaining with IBA1 (green), GFAP (magenta), and DAPI (blue) of cortex sections from 6-MO control (LSL-APP swe ) and TgAPP swe OCN mice. Scale bars: 100 µm (upper) and 20 µm (lower). d Quantification of data in c. e Representative Western blots using antibodies against hAPP, GFAP, and IBA1 in homogenates of cortex and hippocampus of control and TgAPP swe OCN mice. GAPDH was used as a loading control. f Quantification of the data in e. g-h Real-time PCR (RT-PCR) analysis of indicated gene expressions in 6-MO control (LSL-APP swe ) and TgAPP swe OCN cortex (g) and hippocampus (h). All quantification data were presented as mean ± SD (n = 3-4). *p < 0.05, **p < 0.01, ***p < 0.001. Student's t test was used in b, d, and f-h.
test (to evaluate mouse recognition memory) 50,51 . Interestingly, age-dependent changes in both MWM and NOR tests were detected in TgAPP swe OCN mice. No obvious difference in MWM or NOR task performance was observed between TgAPP swe OCN and control mice at 3-MO (Fig. 4a-c). Un-expectedly, at 6-MO, TgAPP swe OCN mice exhibited faster learning and better long-term memory in MWM (Fig. 4d, e), but no obvious difference in NOR task performance (Fig. 4f), compared to the age-matched controls, suggesting an improvement in spatial learning and memory in 6-MO TgAPP swe OCN mice. Interestingly, at~12-MO, In all these behavior tests, 6-MO and 12-MO control (LSL-APP swe ) and TgAPP swe OCN mice (males) were examined. All quantification data were shown as mean ± SD (n = 10 mice). *p < 0.05, **p < 0.01, ***p < 0.001, Student's t test.

Increased senescence and SASPs in APP swe + OB-lineage cells.
To investigate if and how APP swe in OCN-Cre + OB-lineage cells gives rise to the brain and behavior phenotypes, we purified OCN-Cre + BMSCs (marked by tdTomato + , believed to be OB progenitors 28 ) from both 6-MO control (OCN-Cre; Ai9) and TgAPP swe OCN ; Ai9 mice using fluorescence-activated cell sorting (FACS), and then subjected them to RNA-seq analysis (Fig. 5a). 917 up-and 1825 down-regulated genes were identified in APP swe + OB progenitors (Fig. 5b). Among these genes, 154 upand 269 down-regulated genes encode secreted proteins (Fig. 5b). Interestingly, GO analysis showed that most up-regulated genes are involved in inflammatory response, cytokine production, and cytokine/chemokine-mediated signaling pathways; and most down-regulated genes are implicated in cell cycle, cell proliferation, and bone mineralization (Fig. 5c). Further heat map analysis illustrated the up-and down-regulated genes for bone mass regulators, cytokines and chemokines, AD risk genes, and growth factors critical for neurogenesis (Fig. 5d). Some of these up/down regulated genes were verified by RT-PCR analyses (Fig. 5e).
Finally, we wondered whether other tissues in TgAPP swe OCN mice develop senescence-like phenotypes. The mRNAs from various tissues [including cortex, hippocampus, TA (Tibialis anterior) muscles, kidneys, and livers] of 6-MO control and TgAPP swe OCN  ) and NOR (f) tests. g-i 12-MO control (LSL-APP swe ) and TgAPP swe OCN male mice were subject to MWM (g, h) and NOR (i) tests. In MWM tests, the latencies to reach the hidden platform during the training period were showed in a, d, and g; and the representative tracing images and quantification of time spent in target quadrant, platform crossing time, and swim speed were shown in b, e, and h. In NOR tests, the time spent with novel object per total time with both objects as the novel object preference was quantified, shown in c, f, and i. All values were presented as mean ± SD (n = 10 mice). *p < 0.05, one-way ANOVA followed by Tukey post hoc test was used in a, d, and g, and Student's t test was used in b, c, e, f, h, and i. mice were subjected to RT-PCR analyses with P16 Ink4a and P53 transcripts--both markers of senescence. Interestingly, both P16 Ink4a and P53 were increased in the cortex and TA muscles, but not kidney or liver, of 6-MO TgAPP swe OCN mice ( Supplementary  Fig. 10a-e). These results suggest brain-region and tissue selective senescence-like phenotypes in TgAPP swe OCN mice.
Diminished behavior phenotypes and brain pathology in TgAPP swe OCN mice treated with senescence inhibitor.
Moreover, the GFAP + reactive astrocytes, IBA1 + cells, and SASP-like factors (e.g., Il1b, Tnfa, but not Il10 or Mmp3) in TgAPP swe OCN cortex were attenuated ( Supplementary  Fig. 12a-c), and the impaired hippocampal DG neurogenesis in TgAPP swe OCN mice was restored ( Supplementary Fig. 12d, e) by D + Q treatments. In aggregates, these results suggest that APP swe -induced senescence and SASPs are likely to prompt cortical brain inflammation and glial activation, which may underlie the behavioral phenotypes in 6-MO TgAPP swe OCN mice. Systemic inflammation in TgAPP swe OCN mice likely due to APP swe -induced OB-senescence and SASPs. To further understand how APP swe -induced OB-senescence and SASPs contribute to the brain pathology and behavior changes in TgAPP swe OCN mice, we speculate that APP swe induced OB-senescence and SASPs contribute to systemic inflammation, which promotes brain inflammation and behavior changes. To test this speculation, we addressed the following questions.
Second, is APP swe in OCN-Cre + cells a key contributor to the systemic inflammation? Although APP swe is largely expressed in OB-lineage cells of TgAPP swe OCN mice (Fig. 1), we cannot rule out the potential contribution of APP swe 's weak expression in the hippocampal dDG to systemic inflammation. To this end, we examined the serum inflammatory cytokines and chemokines in mice (LSL-APP swe ) injected with AAV-CaMKII-Cre or AAV-GFP into their dDGs; and the Cre-injected mice exhibited similar levels of APP swe /Aβ42 in the hippocampus compared to 12-MO TgAPP swe Ocn mice ( Supplementary Fig. 8g). Using a small-scale antibody array containing antibodies against multiple SASP-like pro-inflammatory cytokines and chemokines ( Supplementary  Fig. 13a), little to no change was detected between the serum samples from the Cre and GFP injected mice ( Supplementary  Fig. 13a, b). These results thus eliminate the possibility of dDG APP swe /Aβ 42 contribution to the systematic inflammation, supporting APP swe in OCN-Cre + OB-lineage cells as a major contributor of systemic inflammation. We also measured serum inflammatory factors in Tg2576 mice, a well-studied AD animal model that expresses APP swe ubiquitously 24 , using multiplexed antibody-based arrays, and compared the changes (upregulated secreted proteins in Tg2576 over control mice) with TgAPP swe OCN mice. Among 49 upregulated secreted proteins in Tg2576 mice, 31 (~63%) were increased in TgAPP swe OCN mice (Fig. 8f), providing additional support for the view.
Third, is the systemic inflammation results from the APP swe induced-senescence and SASPs? Measuring serum SASP-like cytokines and chemokines in TgAPP swe OCN mice treated with and without D + Q, as illustrated in Fig. 7a, demonstrate that many cytokines (IL1β, 2, 23, 27) and chemokines (CCL2, 11 and CXCL1, 2) were increased in serum samples of TgAPP swe OCN mice treated with Veh, but decreased in the mice with D + Q treatments ( Supplementary Fig. 13c, d). Together, these results suggest that the systemic inflammation in TgAPP swe OCN mice is likely in large due to the APP swe -induced OB-senescence and SASPs.
Induction of ER stress-driven OB-senescence by expression of APP swe , but not APP wt or APP lon . To understand how APP swe in OB-lineage cells induces senescence, we re-analyzed the RNA-seq data (APP swe + vs control OB progenitors) and found that, in addition to the increases in mRNAs of senescence genes, the transcripts of ER stress genes (e.g., Grp78, Atf6, and Hsp90) were elevated in APP swe + OB-progenitor cells (Fig. 9a, b). The increase in ER stress proteins (e.g, GRP78 and ATF6) were further verified by Western blot (Fig. 9c, d). To investigate the relationship between APP swe -induced ER stress and senescence, we treated APP swe + OB-progenitors with 4-PBA (4-Phenylbutyric acid), an inhibitor of ER stress 58 . 4-PBA treatments abolished the increases of the senescence marker proteins P16 Ink4a , P53, and SA-β-gal (Fig. 9e-h), suggesting that APP swe likely increases OB-senescence by inducing ER stress.
Notice that ER stress can be induced by the overexpression of membranous proteins 59 . It thus is necessary to determine if the increased ER stress in APP swe + cells results from its over expression. To this end, MC3T3 cells (an OB cell line) expressing APP wt -YFP (wild type), APP swe -YFP, and APP lon -YFP were examined. MC3T3 cells expressing APP swe -YFP, but not APP wt -YFP or APP lon -YFP, showed an obvious increase in GRP78 (an ER stress sensor) ( Supplementary Fig. 14a, b), indicating a more dramatic effect on ER stress by APP swe -YFP and demonstrating its specificity.
Additionally, a more prominent co-localization of GRP78 with APP swe -YFP than those with APP wt -YFP or APP lon -YFP was observed ( Supplementary Fig. 14a, c). Moreover, APP swe -YFP had an increased co-localization with EEA1, an early endosome marker, but decreased co-location with GM130, a marker for Trans-Golgi, compared with those of APP wt -YFP or APP lon -YFP (Supplementary Fig. 14d-g). These results demonstrate APP swe 's distinctive cellular features in its increase of GRP78 and its subcellular localizations. Finally, the senescence marker, SA-β-gal, was selectively increased in MC3T3 cells expressing APP swe -YFP, but not APP wt nor APP lon ( Supplementary Fig. 14h-i), providing additional support for the specificality of the detrimental effects by APP swe , but not by the overexpression of APP wt or APP lon .

Discussion
Patients with AD often have osteopenia or osteoporosis [3][4][5][6][7][8][9][10] . The lower bone mineral density is often reported in the earliest clinical stages of AD patients (both men and women) and associated with their brain atrophy and memory decline 8 . However, it remains unclear if the AD patients carrying the Swedish mutations have osteoporosis-like deficit. Here, using TgAPP swe OCN mouse model that selectively expresses APP swe largely in the OBlineage cells, we found that APP swe in OB-lineage cells induces senescence and SASPs, which appear to be a key contributor of systemic inflammation, and thus promote anxiety-and depression-like behaviors in TgAPP swe OCN mice. Our studies also suggest that the senescence may be insufficient to induce the cognitive decline detected in 12-MO TgAPP swe OCN mice, which may be associated with a weak expression of APP swe /Aβ 42 in the dDG neurons of the hippocampus. These observations, summarized in Fig. 10a, lead to a working hypothesis depicted in Fig. 9 APP swe induction of OB-senescence via ER stress. a Heat map of differentially expressed ER stress or anti-stress related genes identified by RNAseq in control (OCN-Cre; Ai9) and TgAPP swe OCN ; Ai9 Td + OB-progenitors (detail analysis was described in Methods). b RT-PCR analysis of ER stress-related genes Grp78, Atf6, Hsp90b1, Eif2ak3, Ern1, Hsp90aa1, and Hspa2 and anti-stress related gene Sirt3 gene expression in purified Td + BMSCs from 6-MO control (OCN-Cre; Ai9) and TgAPP swe OCN ; Ai9 mice, *p < 0.05, **p < 0.01, ***p < 0.001, mean ± SD, n = 3, Mann-Whitney U test. c Western blot analysis of indicated protein expression in BMSCs from mice with indicated genotypes (at 6-MO). GAPDH was used as a loading control. d Quantification of data in c, *p < 0.05, **p < 0.01. mean ± SD, n = 4, Student's t test. e Western blot analysis of indicated protein expression in BMSCs from 6-MO control and TgAPP swe OCN with or without 0.25 mM 4-PBA (4-Phenylbutyric acid) treatment. f Quantification analyses of the data in e, *p < 0.05, n = 3. g SA-β-gal staining of 6-MO control and TgAPP swe OCN BMSCs with vehicle (Veh)(PBS) and 4-PBA treatment, respectively, scale bar, 20 µm. h Quantification of SA-βgal + cell densities in g (mean ± SD; n = 5, **p < 0.01, ***p < 0.001). Two-way analysis of variance test was used in f and h. A prerequisite to a better understanding of the mechanisms of TgAPP swe OCN mice' brain/behavior phenotypes is to reveal where exactly the APP swe is expressed. TgAPP swe OCN mice are generated by crossing OCN-Cre with the LSL-hAPP swe mice, and thus the APP swe expression is controlled not only by the CAG promoter in LSL-hAPP swe mice (for its mRNA expression), and but also by the OCN-Cre dependent removal of LSL (for hAPP swe protein expression) 23 . Although OCN-Cre mice express Cre largely in OB-lineage cells 60 , our recent study demonstrates the Cre activity in neurons at the dDG, olfactory bulb, and cerebellum of the brain in OCN-Cre mice 29 . Our further studies in this paper lead us to conclude that hAPP swe or OCN-Cre is largely expressed in the OB-lineage cells, but weakly expressed in the dDG neurons, in 12-MO TgAPP swe OCN mice (Fig. 1b-g). We thus believe that the cortical brain and behavior phenotypes in 6-MO TgAPP swe OCN mice are likely induced by the APP swe in OB-lineage cells. However, it is possible that the weak APP swe /Aβ 42 expression in aged (12-MO) dDG hippocampal neurons contributes to the inflammation phenotypes in the mutant hippocampus and the cognitive decline (Fig. 10b).
How does APP swe in OB-lineage cells induce brain pathology? Several lines of evidence support the hypothesis that APP sweinduced OB-senescence and SASPs may underlie its effects on the brain, particularly the cortex, via systemic inflammation (Fig. 10b). Many SASP-like proteins were induced in cultured APP swe + OB progenitors and increased in serum samples of TgAPP swe OCN mice (Figs. 5 and 8). Cultured APP swe + OB progenitors and MC3T3 cells showed increased senescence cells (Fig. 6 and Supplementary Fig. 14h-i) 52,61 . While the OB-senescence phenotypes were temporally associated with APP sweinduced bone-deficits 23 , they occurred earlier than brain deficits, in TgAPP swe OCN mice (Fig. 6). The inhibition of senescence in TgAPP swe OCN mice diminished nearly all the brain and behavior phenotypes ( Fig. 7 and Supplementary Fig. 12). In line with this hypothesis are the multiple literature reports that demonstrate cellular senescence as tightly linked to skeleton and brain aging and various degenerative diseases, including AD [62][63][64][65][66][67] , and the use of senolytic drugs to attenuate the disease process has been shown in several animal models of AD 68,69 .
In terms of the systemic inflammation, while it can be induced by deficits in multiple organs, our results suggest that APP swe -induced senescence and SASPs in OB-lineage cells appear to be a key contributor to this event. Many (31 over 49,~63%) upregulated SASP-like factors detected in serum samples of TgAPP swe OCN mice were also detected in Tg2576 mice (Fig. 8f). Although APP swe is weakly expressed in the dDG neurons of TgAPP swe OCN mice (Fig. 1), examining the serum inflammatory cytokines and chemokines in mice (LSL-APP swe ) injected with AAV-CaMKII-Cre or AAV-GFP into their dDGs showed an increase in APP swe /Aβ42 in the hippocampus of Cre injected mice ( Supplementary Fig. 8g), but a comparable level of serum cytokines and chemokines between Cre and GFP injected mice ( Supplementary Fig. 13a, b). Treatments with senescence inhibitors (D + Q) abolished nearly all the increased inflammatory cytokines in the serum samples of TgAPP swe Ocn mice ( Supplementary Fig. 13c, d). These results thus eliminate the possible contribution of the APP swe /Aβ 42 at the dDG to systematic inflammation, and support the view.
How does APP swe in OB-lineage cells induce senescence and SASPs? We believe that APP swe -induced ER stress may underlie this process for the following reasons. First, expressing APP wt , APP swe , or APP lon in osteoblastic cell line, MC3T3 cells, results in an increased of β-gal + SnCs specifically in APP swe + , but not APP wt + or APP lon + , cells ( Supplementary Fig. 14h-i), although APP or Aβ levels were increased in all three types of cells. These results not only suggest the specificity of APP swe in the induction of the senescence, but also implicate Aβ's insufficiency or independency to this event. Second, APP swe , compared to APP wt or APP lon , exhibited distinctive features in its subcellular localizations and its induction of ER-stress, in addition to senescence ( Supplementary Fig. 14a-g), revealing an association between the selective induction of the ER stress and senescence by APP swe , but not APP wt or APP lon , in line with the view that APP swe is processed by β-secretase or BACE1 in Golgi-derived vesicles, and APP wt is cleaved in the endosomes 70 . Third, both RNA-seq and Western blot analyses showed that APP swe + OB progenitors have increased expressions of not only senescence associated genes, but also ER stress genes (e.g., Grp78, Atf6, and Hsp90) (Fig. 9a-d); and treatment of APP swe + OB progenitors with an ER stress inhibitor 4-PBA abolished the increase of senescence marker proteins P16 Ink4a , P53, and β-gal + SnCs (Fig. 9e-h), supporting the view for ER stress as an driver of senescence. Notice that Hashimoto et al. report an absence of ER stress responses in App NL-G-F (App knock-in mice harboring Swedish mutation) brain 59 . We thus speculate that this event may be cell type/tissue specific, and OB-lineage cells may be more sensitive to APP swe than neurons in its induction of ER stress.
Are senescence and SASPs induced by osteoblastic APP swe involved in the behavior changes observed in TgAPP swe OCN mice? Our results suggest that they are likely contributors to anxiety and depression, but insufficient to cause cognitive decline. In addition to the temporal association between the increased SASPs and the behavior changes, inhibition of senescence and SASPs by D + Q diminishes nearly all the behavior changes in TgAPP swe OCN mice at 6-MO (Fig. 7). Among the SASPs induced by APP swe , IL-1β is noteworthy, because IL-1β is found to mediate bi-functions in regulating spatial learning and memory [71][72][73] . Expressing IL-1β in the brain (in particular, the cortex) exhibits enhanced spatial learning and memory in young adult, but not aged, mice 74 , a similar behavioral phenotype examined in the TgAPP swe OCN mice (Figs. 2g and 4 and Supplementary Fig. 5g). This IL-1β's function is also in agreement with numerous reports, that IL-1β is upregulated by long term potentiation (LTP) (an event critical for learning and memory) [75][76][77] . The overexpression of IL-1ra, an endogenous IL-1R antagonist or IL-1R KO (knock-out), blocks spatial memory 78,79 as well as LTP 73,80 . In the light of these reports, we speculate that the osteoblastic APP swe , via increasing IL-1β, a key SASP, may improve hippocampal/cortex-dependent spatial learning and memory function in an age-dependent manner. We are also aware of controversial reports, which claim that IL-1β plays a detrimental role in regulating learning and memory 81,82 . While IL-1β plays a role in modulating learning and memory, its precise function appears to strongly depend on the site of IL-1β injection/increase, timing, and dosage 73,79 . Notice that Il1b was increased in the hippocampus but not the cortex of 12-MO TgAPP swe OCN mice ( Supplementary Fig. 5g-h); and such IL-1β increase was accompanied by elevated Aβ 42 and glial activation in the hippocampus, and cognitive decline behaviors ( Supplementary Figs. 1b and 5 and Fig. 4g-i). We thus speculate that the hippocampal inflammation phenotype may be induced by the weak APP swe /Aβ 42 expression in dDG hippocampal neurons, which may also impair cognitive function in 12-MO TgAPP swe OCN mice (Fig. 10b). It would be of interest to further test this view in future experiments. Finally, it is highly possible that complex mechanisms underlie APP swe regulation of brain and behavior phenotypes in TgAPP swe OCN mice. In addition to IL-1β and TNFα, other SASPs and growth factors may also contribute to the brain pathology. In addition to senescence and SASPs, the weak expression of APP swe /Aβ 42 in the OCN-Cre + dDG neurons may be exacerbated by systemic inflammation and be responsible for the hippocampal pathology and cognitive decline in aged (e.g., 12-MO) TgAPP swe OCN mice. It is also noteworthy that while chronic inflammation is believed to be one of the environmental risk factors for AD development 30,83 , our studies suggest that the chronic systemic inflammation associated with AD patients (either EOAD or LOAD) may be induced by a combination of AD genetic risk gene(s), a primary hit, and environmental risk factors (e.g., aging, infection), a secondary hit, in line with the two-hit hypothesis 84 . Further investigations that address how chronic inflammation is induced, how it promotes the brain pathology and behavior changes, and what is the function/ contribution of APP swe /Aβ 42 in dDG neurons to the AD development may gain more insights into the two-hit hypothesis and AD pathogenesis.
Immunofluorescence staining and image analysis. Immunostaining was performed as described previously 29 . In brief, mice were anesthetized with isoflurane and were transcardially perfused with PBS (50 mL) followed by 4% (w/v) paraformaldehyde (PFA) in phosphate buffer (PBS) (pH 7.4) (50 ml) to remove intravascular plasma proteins. The dissected brains were post-fixed in 4% PFA at 4°C overnight. Coronal sections (40 μm) were washed 3 times with PBS (10 min each) and treated with blocking reagent (10% Donkey Serum + 0.5% Triton 100×) for 1 h, then incubated overnight at 4°C with the primary antibody. Sections were washed 3 times and incubated with corresponding conjugated secondary antibody for 1 h. DAPI was used for nucleus counter staining. Stained sections were imaged by confocal microscope at room temperature. Fluorescent quantification was performed using ZEN software according to the manufacturer's instructions (Carl Zeiss).
Western blotting. Western blotting was performed as described previously 85 . Brain tissues and cultured BMSCs were homogenized in modified RIPA buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1 mM EDTA,) containing 0.5% sodium deoxycholate, 0.1% SDS, 1 mM PMSF, 1 mM Na3VO4, 1 mM NaF, 1 mM DTT, and protease inhibitor cocktail (Millipore, 539134). Lysates were centrifuged at 10,000 x g for 10 min at 4°C to remove debris and to obtain homogenates. Samples were resolved by SDS-PAGE and transferred to a nitrocellulose membrane (1620112, Bio-Rad Laboratories). After incubation with 5% milk in TBST (10 mM Tris, 150 mM NaCl, and 0.5% Tween 20, pH 8.0) for 1 h, membranes were immunoblotted with indicated antibodies overnight at 4°C. Membranes were washed with TBST three times and incubated with a 1:2000 dilution of horseradish peroxidase-conjugated anti-mouse or anti-rabbit antibodies for 1 h. Blots were washed with TBST three times and immunoreactive bands were visualized using the LI-COR Odyssey infrared imaging system. Intensity of immunoreactive bands were quantitated by using ImageJ (NIH).
EdU injection and labeling. Control (LSL-APPswe) and TgAPP swe OCN mice were given four intraperitoneal injections of EdU (50 mg/kg/time, 1 time/4 h) within 12 h. 12 hours after their last injection, mice were euthanized and transcardially perfused first with 50 ml of cold PBS and then with 50 ml of 4% PFA. The dissected brains were post-fixed in 4% PFA at 4°C overnight. Coronal sections (40 μm) were obtained for staining. Cultured BMSCs were incubated with 10 µM EdU for 2 hours, and then cells were fixed with 4%PFA for 10 min. EdU staining was performed using a Clik-iT EdU imaging kit with Alexa-Fluor 488 (Invitrogen) following the manufacturer's instructions.
Behavioral tests. Mice (male) at ages of 3-, 6-or 12-MO (month old) were subjected to behavioral studies. Behavioral tests were done blind to genotypes or treatments. For all behavioral experiments, mice were transferred to the testing room 4 h before any test to acclimate to the environment. All behavioral instruments were cleaned with 70% ethanol prior to each trial.
Open field test (OFT), Elevated plus maze test (EPMT), and Light/dark transition test (LDT) were performed as described previously 29 . In brief, for OFT, each mouse was placed in a chamber (L × W × H = 50 × 50 × 20 cm) and its movement was monitored for 10 min using an overhead camera. Light intensity was about 150 lux. The video was analyzed by a tracking software (Etho Vision, Noldus). The total distance and center (25 × 25 cm) duration time were quantified. For elevated plus maze test (EPMT), the EPM was placed 50 cm above the ground. Each mouse was initially placed in the center square facing one of the open arms (L × W = 60 × 5 cm). Light intensity was about 100 lux. Mice movement was recorded for 5 min using an overhead camera and tracking software (Etho Vision, Noldus). The time spent in the open arms and the number of open arm entries were quantified. For light/dark transition test (LDT), mouse was firstly placed in the dark compartment, overhead camera was turned on, and the door between lit and dark chambers was opened. Light intensity was about 200 lux in the lit chamber. 10 min of movement was recorded using a tracking software ((Etho Vision, Noldus). The time spent in the lit chamber and the number of transitions were quantified.
The tail suspension test (TST), forced swimming test (FST), and sucrose preference test (SPT) were performed as described previously 86 . For the TST and FST, the last 4-min of a 6-min test were analyzed, and the immobility time was measured directly. The sucrose preference test was carried out using a two-bottle choice procedure. Single housed mice were habituated to drink 2% (wt/vol) sucrose solution (dissolved in water) for 3 days, then mice were given access to the two preweighed bottles, one containing water and the other containing 2% sucrose solution. Bottle positions were changed every day and water and sucrose solution consumption was assessed daily for 4 days. The consuming ratio of sucrose over total solution consumed was used for measuring the sucrose preference.
The Morris water maze (MWM) was performed as previously described 87 . Specifically, a 120 cm pool and 10 cm platform were used for water maze and nontoxic bright white gel (Soft Gel Paste Food Color, AmeriColor) was added to the water to make the surface opaque and to hide the escape platform (1 cm below the surface). Mice were trained for 5 days, four trials per day with 20 min interval between trials and 60 s per trial to locate the hidden platform. Eight spatial cues were placed on the pool wall, visible for mice to find the hidden platform. On the 6 th day, the platform was removed, and mice were placed into the pool at a new starting position. The time spent in each platform quadrant and the number of platform-crossing within 60 s were analyzed. The swim speed and the amount of time spent in each quadrant were quantified using the video tracking system (Noldus). The investigators were blind to genotype during data acquisition and analysis.
The Novel Object Recognition Task (NOR) was based on a previous published procedure 88 . It consists of a habituation phase followed by a testing phase. During the habituation phase, each mouse was allowed to freely explore the empty arena over two days. On the third day, the testing phase begun. Habituation consisted of one ten-minute session administered one per day. The testing phase consisted of a (1) familiarization trial followed by a (2) test trial. During the familiarization trial, a single mouse was placed in the arena containing two identical objects and released against the center of the opposite wall with its back to the objects. This was done to prevent coercion to explore the objects. Object interaction is defined as entrance into the object-containing zone resulting in direct or nearly direct object contact with the nose or whiskers. The test trials were administered after delays of 1-hour post-familiarization. The test trial was administered in the aforementioned way except that one sample object from the familiarization trial and one novel object were presented. During the test trials, time spent with novel object per total time with both objects as the novel object preference was quantified.
AAV virus injection. AAV9-CamkII-GFP (105541-AAV9) and AAV9-CamkII-Cre (105551-AAV9) were purchased from Addgene. Virus injection was performed as described previously 29 . In brief, male LSL-APPswe mice (4-month-old) were anesthetized with Ketamine/Xylazine (HENRY SCHEIN #056344), and the head was fixed in a stereotaxic device (David Kopf Instruments). After the antiseptic treatment, the skull was exposed and cleaned using 1% H 2 O 2 . Holes were drilled into the skull and viruses were bilaterally injected into DG at the coordinates relative to bregma: caudal: −2.06 mm; lateral:±1.3 mm; ventral: −1.75 mm. After injection, the needle was left in place for 5 min to allow for diffusion of injected viruses before being slowly withdrawn. For the following 5 days, mice were daily injected with Meloxicam to reduce pain. Injection locations were validated in each mouse after experiments.
In vitro primary OB-progenitor (BMSCs) cultures. OB-progenitor (BMSCs) culture was carried out following a standard protocol as described previously 28,85 . In brief, the whole bone marrow cells flushed out from long bones of mice with DMEM were filtered through a 70-mm filter mesh, washed, re-suspended, and then plated in 100-mm dishes with growth medium (DMEM plus 10% FBS), which were incubated at 37°C with 5% CO 2 . 3 days later, the non-adherent cells were removed. The attached bone marrow cells were cultured with the growth medium for 7 days. These cells were passaged and cultured for another 3-6 days with the same growth medium. These cells, so called BMSCs, were used for Western blot, RT-PCR, and SA-β-gal staining.
Flow cytometry analysis. Flow cytometry analysis was done as previously described 28 . BMSCs were flushed from femurs and tibias of 6-MO OCN-Cre; Ai9 and TgAPPswe OCN ; Ai9 mice, the attached bone marrow cells were cultured with the growth medium for 7 days. These cells were passaged and cultured for another 3 days with the same growth medium. Then cell media were removed from culture dishes and cells were rinsed with PBS. Trypsin solution was added to incubate at 37°C for 2 min. The detached adherent cells were centrifuged, and the pellet cells were washed with 1 ml cold PBS, and finally resuspended in 0.5 ml PBS with 1% FBS for flow cytometry analysis. Flow cytometric analysis was performed by use of a flow cytometer in CWRU core facility. Acquisition and analysis were performed by using FACSDiva 8.0.1 software (BD).
instruction. Serum IL-6 was measured with mouse IL-6 ELISA kit (550950, BD Biosciences), following the manufacturers' instruction. Serum, Brain and BMSCs homogenization was obtained for human Aβ 40/42 Elisa assay. Brain tissues were homogenized as previously described 87 . Human Aβ40 and Aβ42 level in serum, brain (300 µg in total protein) and BMSCs (50 µg in total protein) homogenates were measured by the Aβ 40 human ELISA kit (Invitrogen, catalog #KHB3481) and the Aβ 42 human ELISA kit (Millipore, catalog #EZHS42), respectively. Their concentrations were determined by comparing readings against their standard curves.
L-Series label-multiplex antibody arrays. Mice blood samples were collected and allowed to clot for 30 min at room temperature and centrifuged for 15 min at 3000 rpm. Serum was frozen and aliquot at −80°C until use. The antibody arrays were performed using an L-Series Glass Slide antibody arrays kit (AAM-SERV-LG, Raybiotech, USA) according to the manufacturer's instructions. In brief, the serum was dialyzed before the biotin-labeling step. The primary amine of the proteins in the sample was biotinylated, followed by dialysis to remove free biotin. The newly biotinylated sample was added onto the glass slide and incubated at room temperature. After incubation with Fluorescent Dye-Strepavidin, the signals were visualized by fluorescence.
Mouse cytokine array. Serum samples were collected as described above. Cytokines were measured with Mouse Cytokine Array Panel A (ARY006, R&D Systems). In Brief, the serum was mixed with a cocktail of biotinylated detection antibodies. The sample/antibody mixture was then incubated with the Mouse Cytokine Array membrane. Any cytokine/detection antibody complex present was bound by its cognate immobilized capture antibody on the membrane. Following a wash to remove unbound material, streptavidin-horseradish peroxidase and chemiluminescent detection reagents were added sequentially. Light was produced at each spot in proportion to the amount of cytokine bound.