MANF Ablation Causes Prolonged Activation of the UPR without Neurodegeneration in the Mouse Midbrain Dopamine System

Visual Abstract

lack of MANF from the brain led to the chronic activation of UPR by upregulation of the endoribonuclease ac-

Introduction
Mesencephalic astrocyte-derived neurotrophic factor (MANF) is a small, endoplasmic reticulum (ER) localized protein with survival-promoting effects and neuroprotective properties (Petrova et al., 2003;Voutilainen et al., 2009). Importantly, MANF has been associated with ER function as its expression is increased in response to several ER stress inducers and it is secreted by ER calcium depletion (Apostolou et al., 2008;Glembotski et al., 2012). ER stress is a phenomenon caused by high levels of misfolded proteins or protein aggregates. To alleviate ER stress, cells activate the unfolded protein response (UPR) involving distinct pathways initiated by three ER transmembrane stress receptors: inositol-requiring enzyme 1a (IRE1a), double-stranded RNA-activated protein kinase-like ER kinase (PERK), and activating transcription factor 6 (ATF6; Walter and Ron, 2011). These pathways enhance the degradation of misfolded proteins, the protein-folding capacity, and attenuate protein translation. If the UPR fails to restore homeostasis, sustained ER stress may trigger cell death. In fact, upregulation of UPR genes has been found in the postmortem brains of Parkinson's disease patients and chronic ER stress is speculated to be a contributor to neurodegenerative diseases (Hoozemans et al., 2007;Hetz and Saxena, 2017;Lindholm et al., 2017). However, we lack critical studies that show the causative role of chronic ER stress and neurodegeneration.
MANF and its paralogue, cerebral dopamine neurotrophic factor (CDNF; Lindholm et al., 2007), are structurally characterized by two domains: the amino-terminal domain is predicted to bind lipids and the carboxy-terminal domain contains an ER-retention signal and is important for the anti-apoptotic activity of MANF (Parkash et al., 2009;Hellman et al., 2011;Mätlik et al., 2015). The Manf gene promoter carries ER stress response elements I and II, which are recognized by UPR transcription factors AFT6 and spliced X-box binding protein 1 (sXBP1), thus regulating Manf mRNA expression (Oh-Hashi et al., 2013;Wang et al., 2018). Fruit flies express a single CDNF/MANF homologous gene, DmManf, which was also shown to genetically interact with Xbp1 (Lindström et al., 2016). MANF interacts with an ER chaperone glucose-regulated protein 78 (GRP78; Glembotski et al., 2012) and recent data show that it is the C-terminal domain of MANF that binds to the nucleotide-binding domain of GRP78, contributing to protein-folding homeostasis as a nucleotide exchange inhibitor (Yan et al., 2019).
Inactivation of the Manf gene in fruit flies causes defects in the formation of dopaminergic axonal bundles and decreased dopamine levels (Palgi et al., 2009). However, knock-down of neuronal DmManf does not affect the dopamine system in adult flies (Stratoulias and Heino, 2015). In zebrafish, knock-down of Manf results in a decreased number of dopamine neurons without phenotypic defects (Chen et al., 2012). In nematodes, a mutation in the Manf homologue was suggested to result in the gradual degeneration of dopamine neurons (Richman et al., 2018), but a recent study implicated that MANF-depleted nematodes do not after all display any dopamine phenotype (Hartman et al., 2019). We have shown that conventional MANF knock-out (KO) mice and pancreas-specific conditional KO mice develop insulin-dependent diabetes due to prolonged ER stress in pancreatic b cells, leading to loss of the b cell mass and premature death (Lindahl et al., 2014;Danilova et al., 2019a). In MANF KO mice, it was also demonstrated that MANF regulates cortical development and migration of neuronal progenitor cells (Tseng et al., 2017). Despite the suggested role of increased UPR activation behind impaired cortical development, the role of MANF in the regulation of UPR in neurons in vivo has remained unsolved. Whereas exogenous MANF has demonstrated protection of dopamine neurons in animal models of Parkinson's disease (Voutilainen et al., 2009;Liu et al., 2018), the function of endogenous MANF in the dopamine neuron maintenance in mice is not known.
To study the role of endogenous MANF in midbrain dopamine neurons and in cortical neurons, we analyzed the brains of conventional and conditional MANF KO mice with the deletion of Manf from the neural lineage of the cells in the CNS. In this study, we aimed to characterize the survival and function of the midbrain dopamine system in conditional MANF KO mice and to examine the effect of MANF absence on ER stress and its consequences for the survival of neurons.

Animals
Mice were housed in an individually ventilated cage system (Mouse IVC Green Line, Tecniplast) in a specific pathogen-free facility with a 12/12 h light/dark cycle (lights on 6 A.M. to 6 P.M.) with access to food pellets (Harlan Teklad Global Diet) and water ad libitum. Cage enrichment was provided by bedding (aspen chips 5 Â 5 Â 1 mm, 4HP, Tapvei), nesting material (aspen strips, PM90L, Tapvei) and an aspen brick (100 Â 20 Â 20 mm, Tapvei). Development of KO mice used in the study have been previously described (Lindahl et al., 2014). Shortly, Manf -/mice were generated from an embryonic stem cell clone MANF_D06 (EPD0162_3_D06; C57Bl/6N-Manf tm1a(KOMP)Wtsi ). The clone included a b -galactosidase reporter cassette, which had a strong splice acceptor site between exon 2 and 3 and splicing from exon 2 to reporter cassette caused a null mutation. Manf fl/fl mice were produced by crossing Manf 1/with transgenic CAG-Flp mice following removal the b -galactosidase reporter cassette. Crossing of Manf fl/fl mice with transgenic Cre mice caused removal of exon three and splicing from exon 2 to exon 4. This resulted in a frameshift and premature stop codon and therefore with truncated Manf mRNA. Manf 1/1 and Manf -/mice were maintained in an outbred ICR background and bred from heterozygous Manf 1/parents. Manf fl/fl mice and Manf fl/fl ::Nestin Cre/1 were in a mixed ICR/C57BL/ 6JRccHsd genetic background, and Nestin Cre/1 mice were only in the C57BL/6JRccHsd background (Envigo). Mice used for most of the behavioral studies were bred over eight generations in the C57BL/6JRccHsd background. Only male mice were used in the behavioral testing, whereas in the other experiments the use of either female or male mice has been informed in the figure legends. Data groups of embryonic and postnatal (P1 and P14) mice include both genders. For sample collection, brains were dissected from euthanized mice, snap frozen in liquid nitrogen and stored at À75°C. For immunohistochemistry, mice were heavily anaesthetized and perfused transcardially with PBS first and then with 4% paraformaldehyde (PFA) in PBS. The experiments with mice were approved by the National Animal Experiment Board in Finland, and the project license numbers were ESAVI/10 564/04.10.07/2014 and ESAVI/11 865/04.10.07/2017.

Blood glucose level measurements and glucose tolerance test
Blood glucose levels were measured in random-fed animals with a glucometer (Accu-Chek Aviva Glucometer, Roche Diagnostics). For the glucose tolerance test, mice were starved for 16 h in their home cages. After measurement of basal blood glucose levels, D-glucose (2 g/kg) was injected intraperitoneally, and blood glucose levels were measured at different time points during 2 h.

Counting of dopamine neurons
To count the relative number of dopamine neurons in the substantia nigra (SN), four nigral sections within an interval of 100-120 mm were selected for the staining. TH-stained sections of the SN were scanned with an automated scanner (3DHistech). An area of the SN pars compacta (SNpc) from each hemisphere was defined separately based on morphologic landmarks and an estimation of dopamine neuron number was performed with a MATLAB algorithm based on intensity from TH-immunoreactive sections (Penttinen et al., 2016). In the paper of Penttinen et al. (2016), the method was validated by comparing stereological counting with this new method and analyzing dopamine neurons in different PD models. Counts were performed bilaterally and results are presented as relative change in the cell number per stained section.

Optical density measurement
Images of TH-immunostained and DAT-immunostained striatal sections were scanned with a 3DHistech scanner. Striata were defined from pictures, and optical density was measured with an Image-Pro Analyzer 7.0. The optical density of a cortical region above the striatum was subtracted as a background signal. Six sections per animal were used for the analysis. The results of integrated optical density are shown as an average and percentage of the control mice.

Measurement of dopamine and serotonin levels
The neurotransmitter levels from dissected striata were measured by high-performance liquid chromatography (HPLC) as previously described (Airavaara et al., 2006). Tissue samples were homogenized in a solution containing 0.2 M HClO 4 and antioxidants following centrifugation at 14,000 rpm for 35 min. Supernatant was further filtered with Vivaspin columns by centrifugation at 9000 rpm for 35 min. From striatal tissues samples, concentrations of dopamine, 3,4-dihydroxyphenylacetic acid (DOPAC), homovanillic acid (HVA), serotonin, and 5hydroxyindoleacetic acid were analyzed with a HPLC including an electrochemical detector.

Real-time quantitative PCR (qPCR)
Total RNA was isolated by phenol-chloroform extraction with TRI Reagent (Molecular Research Center). Briefly, TRI reagent (Invitrogen, Thermo Fisher Scientific) was added to samples, followed by mechanical disruption by grinding with a pestle or by pulling through a needle by a syringe. Chloroform (1:4) was added to the samples, which were centrifuged at 13,000 rpm for 15 min. The aqueous phase was collected and isopropanol (1:1) added. For small samples, glycogen (Thermo Fisher Scientific) was added to visualize the pellet. After incubation, the samples were centrifuged at 13,000 rpm for 15 min, followed by washing of the pellet twice with 75% ethanol. The pellet was air-dried and dissolved in water. RNA concentration was measured by NanoDrop and equal amounts of RNA were used for synthesizing complementary DNA. RNA quality was assessed by measuring UV absorbances of the RNA samples. Reverse transcription reaction was catalyzed by RevertAid Premium Reverse Transcriptase (Thermo Fisher Scientific) or Maxima H Minus Reverse Transcriptase (Thermo Fisher Scientific) in the presence of 10 mM NTP mix (Fermentas UAB) and oligo-d(T) (Metabion International). qPCR was performed with Lightcycler 480 SYBR Green I Master mix (Roche Diagnostics) using Lightcycler 480 Real-Time PCR System (Roche). The mRNA levels of the target gene were normalized to levels of b -actin as a reference gene, and quantification was performed by a standard curve method. Primer sequences were designed to span exon-exon junctions. Primer sequences were: Atf4

Primary dopamine cell cultures
To culture midbrain dopamine neurons from single embryos, the method of micro-island cultures was used. Micro-islets were drilled into four-well plates and coated with poly-ornithine. Midbrain floors were dissected from E13.5 embryos, collected in Dulbecco's media and washed with HBSS. Cells were detached by trypsin treatment at 137°C for 20-30 min. To stop the reaction, fetal bovine serum (FBS) was added and cells were triturated. After the sedimentation of cells, the supernatant was collected and the trituration repeated. Cells were centrifuged for 5 min at 1000 rpm. The pellet was washed twice with the media and dissolved in the media. Approximately 2000-3000 cells were added to each micro-island. Some plates were fixed on the first day in vitro (DIV1). For other plates, half of the media was changed at DIV2 and fixed at DIV5. Cells were fixed with 4% PFA, washed and permeabilized with 0.2% Triton X-100 in PBS. Samples were blocked with 5% horse serum 1 0.2% Triton X-100 in PBS and incubated with anti-TH antibody (1:1000, MAB318) overnight. Specimens were incubated with secondary antibody (1:400, Alexa Fluor 488 Goat Anti-mouse, Life Technologies, A11029) for 1 h, and nuclei were stained with DAPI (1:5000) for 10 min. Images from micro-islands were taken with a Leica fluorescent camera DC300F and TH1 cells were quantified from images with ImageJ (Rueden et al., 2017). Each data point was an average of at least two micro-islands.

Primary cortical cell cultures
Cortex was collected from mouse fetuses at gestational day 16 (E16) or E17 separately and later genotyped.
Cortical pieces were washed twice with HBSS. Samples were incubated with 0.5% trypsin for 15 min at 137°C, followed by addition of HBSS 1 10% FBS and DNase I. Centrifuged pellets were washed twice with HBSS1FBS solution. Neurobasal medium (Invitrogen) supplemented with B27 (1Â, Invitrogen), 500 mM glutamine, and primocin was added on top of the pellet and pipetted until tissue pieces were separated. Cells were centrifuged and washed with media. Dissociated neurons were counted, and cells were plated on poly-ornithine coated 96-well or 12-well plates at the amount per well of 2.5 Â 10 4 and 4 Â 10 5 , respectively. Half of the medium was changed every 2-4 d. Cortical neurons cultured on 12-well plates were treated with 20 mM 5-fluoro-29-deoxyuridine (F0503, Sigma-Aldrich) and 20 mM uridine (U3750, Sigma-Aldrich) from DIV2 to DIV5 to reduce the number of glial cells. Treatments were performed during media change at DIV9. Neurons cultured on 96-well plates were fixed with 4% PFA. They were permeabilized with 0.2% Triton X-100 in PBS for 15 min and blocked with 5% horse serum in 0.2% Triton X-100 in PBS. A primary antibody (anti-NeuN, 1:500, MAB377, Millipore) was added to the blocking medium and incubated overnight. After incubation with a secondary antibody (1:400, Alexa Fluor 488 goat anti-mouse, Life Technologies, A11029) for 1 h, the nuclei were stained with DAPI. Cultures were imaged with a CellInsight CX5 instrument (Thermo Fisher Scientific) and the number of NeuN-positive cells was analyzed automatically by designed image analysis workflow in CellProfiler and CellProfiler Analyst programs (Carpenter et al., 2006;Jones et al., 2009). Neurons grown on 12-well plates were used for RNA extraction or Western blotting in the same way as described above.

Multiple static rods
The performance of a mouse on five rods with increasing diameter (rod 1: 27 mm, rod 2: 21 mm, rod 3: 15 mm, rod 4: 11 mm, rod 5: 8 mm) was measured. At the beginning of a test, a mouse was placed at the end of the round beam facing away from the supporting platform. The latencies to turn 180°, walk 60 cm, and possible fall off were measured within a maximum of 2 min.

Wire hanger test
A mouse was placed in the middle of the horizontal hanger grid. The mouse was observed for 1 min, and the latency to fall was measured.

Accelerating rotarod
Mice were trained once on an accelerating rotarod (from 4 to 40 rpm during 4 min). Later on, the same day, the latency to fall was measured with the 4-min cutoff time.

Open-field activity
Mice were placed in the corner of an open-field arena (30 Â 30 cm, Med Associates). The horizontal activity was monitored for 30 min. For testing amphetamine-induced locomotor activity, mice were first habituated in the open-field arenas for 30 min. Mice were then injected with 3 mg/ kg D-amphetamine intraperitoneally and placed back in the same chambers. The horizontal activity was measured for another 90 min.

Barnes maze
A circular elevated platform (diameter 100 cm) contained 20 equally distributed holes (diameter, 5 cm) around the perimeter. An escape box was placed under one of the holes. Mice were trained to find the escape box during training sessions of a maximum of 3 min per time. The trial was ended either when the animal entered the box or 3 min had elapsed; in the latter case the animal was guided to the box by hand and allowed to enter it. Training was conducted three times per day (with an inter-trial interval of at least 1 h) on three consecutive days. On day 4, for testing of memory in probe trial 1, the box was removed and mice had 1.5 min to navigate to the right place. For reverse training, the escape box was set at the opposite side of the maze. Reverse trainings of 3 min occurred three times per day on two consecutive days. The maze was cleaned with water between every animal. The trials were recorded with video tracking software Ethovision XT10 (Noldus). The distance traveled and the latency to find the escape box were measured during the training trial; in addition, the time spent in a zone surrounding the target hole was measured in the probe trial.

Statistical analysis
Statistical analysis was performed using either SPSS Statistics (version 24, IBM) or GraphPad Prism (version 7.04, GraphPad Software). Student's t test was used when comparing two groups when data were normally distributed. If data were not normally distributed or Levene's test for equality of variances showed unequal variances, the Mann-Whitney U test was used to compare two groups. If only part of the large dataset was not normally distributed, the nonparametric Mann-Whitney U test was still used to make consistent analysis. Data with three groups and one factor were analyzed with one-way ANOVA, followed by Tukey's post hoc test, and if data were not normally distributed, the nonparametric Kruskal-Wallis test followed by Mann-Whitney U test was used. For tests with many measurements from the same subject, we used two-way ANOVA for repeated measures. For the data, where thapsigargin was used as a treatment, two-way ANOVA was used to analyze genotype-treatment interaction, and it was followed by Tukey's post hoc test. The behavioral tests and analysis were performed blinded. All results are presented as mean 6 SEM and regarded as significant when p , 0.05. The statistical test used are listed in Table 1

Conditional deletion of MANF in the nervous system
Conditional MANF KO mice were produced by deleting an exon three of the Manf gene by crossing the floxed mice (Manf fl/fl ) with transgenic Nestin Cre/1 mice. Cre recombinase expressed under the Nestin promoter in Nestin Cre/1 transgenic mice is induced predominantly in the CNS (Tronche et al., 1999). Later, it was reported that in the Nestin Cre/1 line used, recombination is observed in neural stem cells and neural progenitor cells from E12.5 but with incomplete efficiency (Dubois et al., 2006). Since MANF is mainly expressed in neurons in the adult mouse brain (Lindholm et al., 2008;Tseng et al., 2017;Danilova et al., 2019a,b), this recombination is expected to remove most of the MANF in the brain. Consequently, qPCR analysis showed barely detectable levels of Manf mRNA analyzed from the brain lysates from different brain areas of Manf fl/fl ::Nestin Cre/1 mice compared with littermate Manf fl/fl mice ( Fig. 1A; Table 1). Brain regions analyzed were selected based on their normally high MANF expression (Lindholm et al., 2008). Importantly, Manf mRNA levels were similar in the cortex between control Manf fl/fl and Nestin Cre/1 mice, and reduced as expected in Manf fl/fl :: Nestin Cre/1 mice ( Fig. 1B; Table 1). This result indicates the comparable levels of Manf mRNA in Manf fl/fl and Nestin Cre/1 mice. Consistent with the Manf mRNA levels, immunoreactivity for MANF with a validated MANF-specific antibody was detected in the brains of Manf fl/fl mice, but the signal was almost fully lost in Manf fl/fl ::Nestin Cre/1 mouse brain counterstained with hematoxylin (Fig. 1C). The loss of MANF did not affect Cdnf mRNA expression levels in the Manf fl/fl ::Nestin Cre/1 brain areas ( Fig. 1D; Table 1) or the Bdnf mRNA levels measured from the cortical and hippocampal tissue ( Fig. 1E; Table 1).
In contrast to global Manf -/mice in the ICR strain that die prematurely (Lindahl et al., 2014) and Manf -/mice in the C57BL/6 strain that are perinatal lethal (Neves et al., 2016;Bell et al., 2019), Manf fl/fl ::Nestin Cre/1 mice were viable and visibly similar to littermates when followed until the age of 18 months. The body weights of Manf fl/fl ::Nestin Cre/1 mice were reduced at four and 16 months of age compared with the Manf fl/fl controls but when comparing Manf fl/fl :: Nestin Cre/1 mice with Nestin Cre/1 mice, there was no difference in body weight ( Fig. 1F; Table 1). Since conventional MANF KO mice develop diabetes by the age of seven weeks (Lindahl et al., 2014), we followed the blood glucose levels to rule out the possible development of hyperglycemia. In line with the previously published data (Lindahl et al., 2014), blood glucose levels of random-fed Manf fl/fl ::Nestin Cre/1 mice stayed similar to Manf fl/fl controls until the age of 16 months ( Fig. 1G; Table 1). The glucose tolerance test measuring insulin secretion and sensitivity was performed by injecting glucose followed by measuring the rate of glucose clearance from the blood stream. Both Manf fl/fl and Manf fl/fl ::Nestin Cre/1 mice showed similar glucose clearance over time ( Fig. 1H; Table 1).

MANF deficiency causes robust activation of IRE1a signaling pathway in the brain
Previously it has been shown that loss of MANF in pancreatic b cells leads to chronic UPR activation in vivo (Lindahl et al., 2014). Recently, it was also demonstrated that ablation of MANF in cartilage results in upregulation of ER chaperones in vivo (Bell et al., 2019). Therefore, we examined the expression of UPR-related genes in the brains of both conventional Manf -/and conditional Manf fl/fl :: Nestin Cre/1 KO mice. The studied UPR-related genes were selected from all three UPR branches in addition to the major UPR sensor GRP78 (alias BiP), which in normal conditions binds to these ER transmembrane UPR transducers but dissociates when misfolded proteins start to accumulate in the ER (Lee, 2005).
First, the qPCR analysis performed on cDNA from the brains of Manf 1/1 and Manf -/mice indicated that the mRNA levels of all the UPR-related genes from three UPR pathways, Atf4, Atf6a, C/EBP homologous protein (Chop), Grp78, sXbp1 (activated by IRE1a endoribonuclease activity), and total Xbp1 (tXbp1), were upregulated in the embryonic E13.5 Manf -/mice brains and postnatally in the brains from newborn and two-week-old Manf -/-mice ( Fig. 2A; Table 1). Next, the UPR induction was analyzed in the brain tissue of five-week-old Manf -/mice at the age of manifestation of high blood glucose levels and diabetes (Lindahl et al., 2014). The brain regions analyzed were the cortex, striatum, hippocampus, SN and cerebellum. The Atf6a mRNA levels were not increased in any brain region analyzed from five-week-old Manf -/mice ( Fig. 2B; Table 1). However, mRNAs encoding for Chop and Atf4, which are synthesized and activated despite the translational initiation block in the PERK pathway, were both upregulated in the cortex of Manf -/mice ( Fig. 2B; Table 1). Interestingly, mRNA levels of the spliced form of a transcription factor XBP1 (sXbp1) were significantly upregulated by 6.1-to 11.5-fold in the cortex, striatum, SN, and cerebellum, indicating the activation of the IRE1a signaling pathway (Fig. 2B; Table 1). In addition to sXbp1,  the mRNA levels of Grp78 were significantly increased in all examined brain areas of five-week-old Manf -/mice. As experimentally-induced hyperglycemia and high-fat diet induced diabetes have been shown to increase ER stress and UPR activation in rodent brains (Zhao et al., 2015;Sims-Robinson et al., 2016), we studied the effect of MANF removal devoid of the hyperglycaemic effect in the brain tissue of Manf fl/fl ::Nestin Cre/1 mice. We examined UPR-related gene expression in the same brain regions of two-month-old Manf fl/fl ::Nestin Cre/1 mice and their Manf fl/fl littermates. The increase in Xbp1 splicing in Manf fl/fl ::Nestin Cre/1 mice varied between brain regions, there was a 4.7-fold increase in the cortex, 3.7fold increase in the striatum, 7.1-fold increase in the Mann-Whitney U test, Student's t test, one-way and two-way ANOVA followed by Tukey's post hoc test, or Kruskal-Wallis followed by Mann-Whitney U test were used for statistical analysis. Data are presented as mean 6 SEM. hippocampus, 4.4-fold increase in the SN, and as much as an 11.2-fold increase in the cerebellum ( Fig. 2C; Table  1). In addition, mRNA levels of Grp78 were significantly increased in all brain regions studied. Thus, upregulation of UPR genes Grp78 and sXbp1 was shared between conventional Manf -/mice and conditional Manf fl/fl ::Nestin Cre/1 mice, confirming that this is a phenotype caused by MANF deficiency and not diabetes.
Next, we estimated protein levels of two ER stress markers, GRP78 and phosphorylated eukaryotic translation initiation factor 2a (p-eIF2a), in the brain homogenates of Manf fl/fl and Manf fl/fl ::Nestin Cre/1 mice. During ER stress, activation of PERK results in phosphorylation of eIF2a, leading to global translational repression (Walter and Ron, 2011). Results showed that the protein levels of GRP78 were significantly increased in the cortex, striatum and cerebellum (Fig. 2D,E; Table 1). Furthermore, the phosphorylation of eIF2a was significantly increased in the cerebellum, suggesting activation of the PERK pathway as well (Fig. 2D,F), but the expression of eIF2a was not increased (Fig. 2D,G). Despite the increase in the phosphorylation of eIF2a, the mRNA levels of Atf4 were not upregulated in conditional Manf fl/fl ::Nestin Cre/1 brain.
In addition to its endoribonuclease (RNase) activity, IRE1a has a kinase activity that in ER stress, is responsible for recruitment of TRAF2 and further activation of the pro-inflammatory pathway, the nuclear factor k subunit (NF-k B; Walter and Ron, 2011;Tam et al., 2012). Recently, it was shown that exogenously added MANF reduces activation NF-k B pathway and downregulates Bcell lymphoma 10 (BCL10) mRNA expression in cytokinetreated human pancreatic islets in vitro (Hakonen et al., 2018). BCL10 is a pro-apoptotic protein and an upstream regulator of the NF-k B pathway (Ruland et al., 2001). Silencing of MANF in a cytokine-treated human b cell line results in an increase of BCL10 mRNA expression (Hakonen et al., 2018). Therefore, we also analyzed the mRNA levels of Bcl10 in Manf fl/fl ::Nestin Cre/1 mice. However, we did not see changes in the expression levels of Bcl10 mRNA in the cortex and SN of Manf fl/fl ::Nestin Cre/1 mice compared with Manf fl/fl mice ( Fig. 2H; Table 1).
Activation of IRE1a RNase activity has been shown to induce expression of thioredoxin interacting protein (Txnip) by degradation of Txnip mRNA destabilizing miRNA-17 (Lerner et al., 2012). TXNIP expression has been associated with guiding b cells to terminal UPR following IRE1a hyperphosphorylation, oligomerization and increased inflammatory signaling (Morita et al., 2017). Therefore, we measured the mRNA levels of Txnip, but results showed no changes in the levels of Txnip mRNA expression in the cortex or SN of two-month-old Manf fl/fl :: Nestin Cre/1 mice ( Fig. 2H; Table 1).
All three UPR pathways are activated in MANFdeficient mouse brain during aging To examine whether UPR activation continues during aging, we analyzed UPR genes in the cortex and SN of 16-month-old Manf fl/fl ::Nestin Cre/1 mice. Similarly to the two-month-old mice, both Grp78 and sXbp1 mRNA levels were almost identically upregulated in the cortex of 16month-old mice ( Fig. 3A; Table 1). This time, also Nestin Cre/1 mice were included as another control, but the results showed similar mRNA levels between Manf fl/fl and Nestin Cre/1 mice. Interestingly, in the SN of 16-month-old Manf fl/fl ::Nestin Cre/1 mice, expression of Atf6a and Atf4 was also upregulated when compared with control mice, indicating activation of the ATF6 and PERK pathways in addition to the IRE1a pathway ( Fig. 3B; Table 1). In comparison with the situation in younger mice, when only the IRE1a pathway was activated, 16-month-old Manf fl/fl ::Nestin Cre/1 mice had all three UPR pathways activated in the SN.
We also analyzed mRNA levels of both Bcl10 and Txnip in the SN of 16-month-old Manf fl/fl , Nestin Cre/1 and Manf fl/ fl ::Nestin Cre/1 mice. These markers, however, had still not changed ( Fig. 3C; Table 1). Finally, GRP78 expression was confirmed to localize to TH-positive dopamine neurons in the SN of Manf fl/fl and Manf fl/fl ::Nestin Cre/1 mice by immunohistochemical costaining of GRP78 and TH (Fig. 3D).

MANF deficiency does not affect the number of midbrain dopamine neurons during ageing
Previous studies have implied that chronic UPR activation has an impact on neuronal survival in the brain as ER stress is associated with neurodegeneration (Hetz and Saxena, 2017;Lindholm et al., 2017). Thus, the chronic ER stress found in the brains of Manf fl/fl ::Nestin Cre/1 mice would indicate decreased survival, degeneration, and loss of neurons in the brains of these mice. Furthermore, MANF has been shown to be expressed in a portion of the TH-positive neurons in the SNpc (Lindholm et al., 2008) and therefore could affect their maintenance. To investigate the role of MANF as a physiological trophic factor for midbrain dopamine neurons, we analyzed the number of surviving nigrostriatal dopamine neurons of mice lacking endogenous MANF in the brain. Dopamine neurons in the SNpc were visualized by immunoreactivity to TH, a rate-limiting enzyme in dopamine synthesis. Relative quantification of TH-stained neurons in the SNpc showed no loss of continued Figure 2. Upregulation of UPR genes in the brains of both conventional and conditional MANF KO mice. A, qPCR analysis of UPR genes in the brains of E13.5 embryos and P1 and P14 pups of Manf -/and Manf 1/1 mice (n = 5-9). B, qPCR analysis of UPR markers in specific brain regions of five-week-old Manf -/and Manf 1/1 female mice (n = 3-6). C, qPCR analysis detecting changes of UPR markers in different brain parts of two-month-old Manf fl/fl ::Nestin Cre/1 female mice and Manf fl/fl female littermates (n = 4-9). D, Representative Western blottings of brain homogenates comparing protein levels of GRP78, p-eIF2a, eIF2a, and GAPDH between Manf fl/fl ::Nestin Cre/1 and Manf fl/fl male mice. Optical density-based quantitation of protein levels in immunoblots, where GRP78 is normalized to GAPDH (E), p-eIF2a to eIF2a (F), and eIF2a to GAPDH (G; n = 4-6). H, mRNA levels of Bcl10 and Txnip in the cortex and SN of two-month-old Manf fl/fl ::Nestin Cre/1 and Manf fl/fl female mice (n = 4-8). Results are scaled to the average value of the control samples. Mann-Whitney U test or Student's t test was used for statistical analysis. Data are expressed as mean 6 SEM. dopamine neurons in one-year-old Manf fl/fl ::Nestin Cre/1 mice ( Fig. 4A; Table 1). The axonal branches and striatal terminals of nigral neurons were detected by immunostaining of striatal sections with antibodies detecting TH and dopamine transporter (DAT; Fig. 4B). Optical densities of TH-positive and DAT-positive staining in the dorsal striatum were not altered, indicating no detectable changes in the striatal innervation of nigral dopamine neurons in one-year-old Manf fl/fl :: Nestin Cre/1 mice ( Fig. 4C; Table 1). Moreover, we stained coronal brain sections including cortex and striatum with a neuronal NeuN antibody, but no obvious changes in the structure or number of positive cells were observed between one-year-old Manf fl/fl ::Nestin Cre/1 and Manf fl/fl mice (Fig. 4D).
In addition to morphology, striatal monoamine levels were measured to examine possible changes in the activity of dopamine neurons. The levels of dopamine and its metabolites, DOPAC and HVA, were measured from the striatal samples of 16-month-old mice by HPLC, i.e., at the time when all three UPR pathways are activated in MANFdeficient mice. The results showed no differences in the production of dopamine or its metabolites between Manf fl/fl and Manf fl/fl ::Nestin Cre/1 mice ( Fig. 4E; Table 1). Unexpectedly, Nestin Cre/1 mice had significantly elevated dopamine levels when compared with Manf fl/fl ::Nestin Cre/1 mice, but there were no differences in the levels of DOPAC or HVA. Moreover, the levels of serotonin (5-hydroxytryptamine) and its metabolite 5-hydroxyindoleacetic acid were measured, and as a result, no differences were observed between the groups ( Fig. 4E; Table 1).
To assess possible changes in TH and DAT levels, we studied their expression at mRNA and protein levels. The mRNA levels of Th and Dat in the striatum and levels of Th  Table 1-1). Additionally, Western blot analysis indicated no changes in the protein levels of TH and DAT in the striatum or TH protein levels in the SN in Manf fl/fl ::Nestin Cre/1 mice compared with control Manf fl/fl mice (Extended Data Fig. 4-1B,C; Extended Data Table 1-1). Thus, MANF deficiency does not affect the expression of TH or DAT in the midbrain of young adult mice.

Mice lacking MANF in the brain exhibit normal motor behavior
To investigate the effect of MANF absence from the nervous system on functional level, we characterized behavior of Manf fl/fl ::Nestin Cre/1 mice using a battery of different behavioral tests. The nigrostriatal pathway controls the initiation of movements and the cerebellum is important for the coordination of movements. Therefore, we analyzed motor behavior of MANF-deficient mice. Although the Nestin Cre/1 line has been reported to have unaltered motor behavior, learning and memory, we still used this mouse line as a control, with Manf fl/fl littermate controls for Manf fl/fl :: Nestin Cre/1 mice (Giusti et al., 2014). Mice were tested at the age of four months and most tests were repeated in the same mice at the age of one year.
The functionality of the dopamine system in Manf fl/fl :: Nestin Cre/1 mice was analyzed by their responses to a dopaminergic stimulant. Amphetamine releases dopamine from vesicles in dopamine neurons and elevates striatal extracellular dopamine levels causing hyperactivity in mice (Yates et al., 2007). The response to D-amphetamine measured by locomotor activity was similar between Manf fl/fl :: Nestin Cre/1 and control mice at the age of four months ( Fig.  5G; Table 1) and 14 months ( Fig. 5H; Table 1), although the response to amphetamine was in general more modest in older mice. These results suggest that both dopamine uptake and release in the MANF-deficient striatum are normal.
Long-term memory formation relies on the synthesis of new mRNA and proteins, but in the case of ER stress, and particularly PERK pathway activation, protein translation is attenuated (Ohno, 2018). Indeed, hippocampal ER stress has been demonstrated to cause memory dysfunction (Zhang et al., 2014). Since we observed increased UPR activation in Manf fl/fl ::Nestin Cre/1 mice, we decided to investigate their spatial long-term memory by using the Barnes maze test, which consists of training trials and probe test trials measuring learning and long-term memory, respectively. In the second part of the test, the mice are trained to find the new location of the escape box to test behavioral flexibility. The results showed that fourmonth-old mice had a similar learning pattern independent of the genotype ( Fig. 5I; Table 1) and we could not see differences between mice at one year of age either, indicating no alteration in learning (data not shown). In the memory test, the time spent in the target hole zone represents spatial long-term memory. The first and second probe trials for four-month-old mice showed that Manf fl/fl , Manf fl/fl ::Nestin Cre/1 , and Nestin Cre/1 mice spent a similar amount of time in the target hole zone ( Fig. 5J; Table 1) and similar behavior was observed for one-year-old mice (data not shown). Furthermore, there were no differences between genotype groups in any other of the parameters tested, which were latency to find the target hole zone, number of visits in the target hole zone, and errors made finding the target hole zone (data not shown). Thus, we found no defects in learning and long-term memory of Manf fl/fl ::Nestin Cre/1 mice.

Cultured cortical neurons from embryonic MANF KO mice show increased ER stress and vulnerability to chemically-induced ER stress
To study the survival of MANF KO neurons in vitro, we isolated dopamine neurons from the midbrain floor of E13.5 Manf 1/1 and Manf -/embryos and cultured them for 5 d. The number of surviving TH-positive cells was measured by comparing the initial number of cells on DIV1 with the number of cells on day 5. The results showed no changes in the survival of non-treated THpositive neurons from MANF KO mice compared with wild-type controls in vitro ( Fig. 6A; Table 1).
Because of the scarce material obtained from a midbrain floor of one embryonic mouse, we could not plan controlled chemical treatments and biochemical analysis for MANF KO dopamine neurons. Therefore, we continued to study UPR activation in vitro using primary cortical neurons isolated from Manf 1/1 , Manf 1/-, and Manf -/embryos at day E16 or E17 as they can be easily obtained with large quantities from embryonic mice. First, we evaluated the ER stress levels of cortical neurons cultured for two weeks. The cultures consisted mostly of a heterogenic population of cortical neurons as they were treated with mitotic inhibitors to reduce non-neuronal cell proliferation. Increased UPR activation was found in Manf -/cortical neurons when compared with neurons from both Manf 1/1 and Manf 1/embryos. Similarly to increased expression of UPR genes in the brain of Manf -/embryos in vivo ( Fig.  2A), expression of Grp78, sXbp1, and tXbp1, but not Chop and Atf6a, was increased in Manf -/neurons cultured for 14 d in vitro (Fig. 6B; Table 1). In addition, the mRNA levels of Atf4 were increased, suggesting activation of the PERK pathway. The Bcl10 mRNA levels did not differ between genotypes ( Fig. 6B; Table 1).
The vulnerability of Manf -/cortical neurons was assessed by applying a chemical ER stressor, thapsigargin, to the cultures. Thapsigargin blocks the sarcoendoplasmic reticulum calcium transport ATPase (SERCA) pump and depletes Ca 21 stores from the ER, leading to UPR activation and cell death in various cell types, including neurons (Li and Hu, 2015). Thapsigargin at a concentration of 10 nM for 48 h led to a 66.3 6 10.4% reduction in Manf -/neuronal survival compared with dimethyl sulfoxide (DMSO)-treated neurons, whereas the decrease in survival of control Manf 1/1 neurons was 34.9 6 7.9% and Manf 1/neurons 37.1 6 13.4% (genotype-treatment interaction, p = 0.039; Fig. 6C; Table 1). At 5 nM thapsigargin, no changes in the number of surviving neurons were detected (Fig. 6C). As a result, neurons without endogenous MANF were more vulnerable to thapsigargin-induced ER stress in vitro.
To examine further the effect of thapsigargin on neuronal cultures, the expression of several UPR markers was analyzed after a 48-h-long exposure to 10 nM thapsigargin. Upregulation of many UPR genes, Atf6a, Atf4, Chop, Grp78, sXbp1, and tXbp1, was detected after thapsigargin treatment in all genotypes (Fig. 6D-I; Table 1). Furthermore, expression of Bcl10 was higher in thapsigargin-treated neurons compared with DMSO-treated ones ( Fig. 6J; Table 1). The treatment with 10 nM thapsigargin resulted in a significant genotype-treatment interaction only regarding mRNA levels of sXbp1 (p = 0.012; Table 1). Thus, the results showed higher upregulation of sXbp1 in Manf -/neurons compared with Manf 1/1 and Manf 1/neurons (Fig. 6H). We also measured the mRNA levels of Txnip in the samples, and the levels were upregulated in response to thapsigargin but did not differ between the genotypes ( Fig. 6K; Table 1).
Finally, as sustained ER stress is known to induce NF-k B activation by IRE1a activation (Kaneko et al., 2003), we also evaluated the phosphorylation of NF-k B by Western blotting (Fig. 6L). Because of a small sample size, we combined Manf 1/1 and Manf 1/samples together. Quantification of intensity in Western blotting bands showed only a trend towards increased phosphorylation of NF-k B in Manf -/neurons after 10 nM Figure 6. ER stress in embryonic cortical cultures of Manf -/mice and their vulnerability to thapsigargin (TG). A, Survival of embryonic midbrain dopamine neurons in vitro isolated from Manf 1/1 and Manf -/mice (n = 7-8). B, Expression of UPR related and Bcl10 genes in Manf 1/1 , Manf 1/-, and Manf -/cortical neurons cultured for 14 d and studied by qPCR (n = 4-8). C, Effect of the TG treatment on neuronal survival. Cortical neurons were treated with TG at DIV9, fixed after 48 h, and stained with the NeuN antibody (n = 3-5). The loss of cells was quantified by comparing the neuron number with vehicle-treated cultures. Neurons were similarly treated with TG at DIV9, collected after 48 h and analyzed by qPCR for the expression of Aft6a (D), Atf4 (E), Chop (F), Grp78 (G), sXbp1 (H), tXbp1 (I), Bcl10 (J), and Txnip (K) expression (n = 3-5). L, Western blottings presenting phosphorylated NF-k B and NFk B expression in Manf 1/1 , Manf 1/-, and Manf -/cortical cultures after 48-h-long exposure to TG. M, Quantification of Western blottings, where p-NF-k B is normalized to NF-k B (n = 4-5). Results are scaled to the average value of the control samples. Each data point represents one animal and its value is an average of at least two wells. Data are expressed as mean 6 SEM of at least three independent experiments. Experimental groups were compared by the Student's t test, one-way ANOVA, or two-way ANOVA followed by Tukey's multiple comparison test. thapsigargin treatment compared with that in control neurons ( Fig. 6M; Table 1).

Discussion
In this study, we elucidate the previously unknown function of endogenous MANF in the mouse brain. Previous studies have suggested a function for MANF in the maintenance of neurons based on neuroprotective and neurorestorative effects of MANF in different rodent degenerative disease models (Airavaara et al., 2009;Voutilainen et al., 2009;Yang et al., 2014;Neves et al., 2016). However, our results indicate for the first time that endogenous MANF is not needed for the maintenance of midbrain dopamine neurons in mice. It has been implicated, that accumulation of misfolded or aggregated proteins followed by severe chronic UPR causes neuronal death in neurodegenerative diseases. This study, however, questions this present view and shows that sustained but moderate UPR in the rodent brain is not detrimental.
The phenotype of conventional Manf -/is dominated by the development of insulin-dependent diabetes, a severe growth defect and premature death (Lindahl et al., 2014). On the contrary, Manf fl/fl ::Nestin Cre/1 mice are viable and have a normal lifespan without diabetes or gross defects, but adult Manf fl/fl ::Nestin Cre/1 mice have a decreased body weight when compared with littermate Manf fl/fl controls. In a recent study, Manf knock-down by siRNA from the hypothalamus resulted in hypophagia and reduced body weight , suggesting that Manf fl/fl :: Nestin Cre/1 mice could be smaller due to hypophagia. However, Nestin Cre/1 mice themselves have been reported with a growth defect due to reduced levels of mouse growth hormone (Galichet et al., 2010;Declercq et al., 2015). Therefore, the reduced weight in Manf fl/fl :: Nestin Cre/1 mice is affected by the Nestin Cre/1 transgene and the effect of hypothalamic MANF deficiency on food intake requires more investigations.
Upon ER stress, the IRE1a receptor is autophosphorylated and its RNase activity leads to degradation of mRNAs and unconventional splicing of Xbp1 mRNA (Walter and Ron, 2011). Our study reveals increased activation of the IRE1a RNase pathway in the brains of both conventional Manf -/mice and conditional Manf fl/fl :: Nestin Cre/1 mice. In young adult Manf -/mice, UPR activation seems more severe with increased expression of Chop mRNA, which may be partially due to the manifestation of hyperglycemia as cerebral ER stress has been reported in diabetic mice (Zhao et al., 2015). Spliced XBP1 induces the expression of UPR target genes, such as Grp78 (Hetz and Saxena, 2017). Indeed, Grp78 expression is also increased in tissue samples where splicing of Xbp1 is elevated. The increased phosphorylation of eIF2a in the brains of Manf fl/fl ::Nestin Cre/1 mice does not result in increased expression of its downstream components Atf4 nor Chop indicating only minor PERK pathway activation at young age. However, in 16-month-old Manf fl/fl :: Nestin Cre/1 mice, expression of Atf4 and Atf6a are increased suggesting activation of both PERK and ATF6 pathways in the brain at older age. Interestingly, the IRE1a RNase pathway is chronically stimulated through the lifespan of Manf fl/fl ::Nestin Cre/1 mice. Therefore, supporting previous findings in other tissues (Wang et al., 2018;Hartman et al., 2019), we conclude that also in neurons MANF regulates UPR mainly through the IRE1a pathway.
In rats, exogenously added and striatally overexpressed MANF protects dopamine neurons in the 6-hydroxydopamine model of Parkinson's disease (Voutilainen et al., 2009;Hao et al., 2017). Here, we analyzed whether MANF deficiency and chronic UPR in the mouse brain leads to nigrostriatal neurodegeneration. It has been shown that in Parkinson's disease dopamine neurons start to degenerate by losing their synapses and axons. This is observed both in patients and in Parkinson's rodent disease models with clear Parkinson's disease phenotype (Domanskyi et al., 2015). In this study, we measured the optical density of striatal dopamine fibers by two different markers, TH and DAT, and did not observe any differences in density between the genotypes. We also used even more sensitive measurements for dopamine neuronal degeneration by quantifying striatal dopamine concentrations and its metabolite levels with HPLC, and did not find any differences between the genotypes. The concentrations of the striatal samples were in range of 7.6-25.4 mg/g of tissue, and similar that we have published before (Albert et al., 2019). Based on these results, maintenance of dopamine neurons is not affected by the embryonic onset of MANF deficiency in the mouse brain, which supports another study showing MANF not being a crucial survival factor for dopamine neurons in Caenorhabditis elegans (Hartman et al., 2019).
Synaptic alteration and loss occur at early stages of neurodegenerative diseases, such as Alzheimer's, Parkinson¨s and Lewy body diseases (Hunn et al., 2015;Overk and Masliah, 2014). It has been demonstrated in mice that synapses can degenerate, although striatal dopamine fibers and dopamine levels stay unchanged (Galli et al., 2014). Therefore, in the future, it will be important to quantify the number of synapses (DeKosky and Scheff, 1990) and examine individual synapses of dopamine neurons in MANFdeficient mice using highly sensitive technologies, such as mass synaptometry (Gajera et al., 2019). Furthermore, changes in dopamine transmission at the synapses can also occur without dopamine neuron loss as shown in CDNF KO mice (Lindahl et al., 2020). Despite the unchanged response to amphetamine, detailed investigation of the dopamine neurotransmission in MANF-deficient mice would be justified in the future.
It is possible that embryonic deletion of MANF is compensated by factors that remain unknown. We confirmed that at least in the adult brain, loss of MANF is not compensated by its protein family member CDNF, but future studies on CDNF-MANF double KO mice are relevant. With possible functional compensation in mind, we also quantified the mRNA levels of Bdnf, Th, and Dat and protein levels of TH and DAT but found no changes in their levels between the genotypes. In future studies it will be important to eliminate embryonic compensatory effects in the MANF KO mice by deleting MANF specifically from the brain in adult mice. It has been shown that in ischemic brains, MANF is expressed in activated microglia (Shen et does not seem to be needed for the proper function of midbrain dopamine neurons in the aging brain.