Microfluidic chip production

The 3-compartment chip master was constructed as described in Kanaan et al. (37). For chip production, polydimethylsiloxane (Sylgard 184, PDMS; Dow Corning, Midland, MI, USA) was mixed with a curing agent (9:1 ratio) and degassed under a vacuum. The resulting preparation was poured onto a polyester resin replicate and reticulated at 70°C for at least 2 h. The elastomeric polymer print was detached, and 2 reservoirs were punched for each macrochannel. The polymer print and a glass coverslip, cleaned with isopropanol and dried, were treated for 3 min in an air plasma generator (98% power, 0.6 mbar; Diener Electronic, Ebhausen, Germany) and bonded together. The chips were placed under UV for 20 min and then coated with a solution of poly d-lysine (10 μg/ml, P7280; Millipore-Sigma, St. Louis, MO, USA) overnight and washed with Dulbecco-PBS (D-PBS) (14190169; Thermo Fisher Scientific, Waltham, MA, USA) before cell seeding. Alternatively, 3-compartment chips were purchased from Microbrain Biotech (Paris, France).

Primary neuronal culture

All animals were ethically maintained and used in compliance with the European Policy on Ethics. Cortices were microdissected from E14 embryos of Swiss mice (Janvier, Le Genest Saint Isle, France) in D-PBS supplemented with 0.1% (w/v) glucose (Thermo Fisher Scientific). C57BL/6N mice were used in Nmrk2-knockout (KO) experiments. Dissected structures were digested with papain (15 U/ml in DMEM; 76220; Millipore-Sigma) for 5 min at 37°C. After papain inactivation with 10% (v/v) fetal bovine serum (GE Healthcare, Little Chalfont, United Kingdom), structures were mechanically dissociated with a pipette in DMEM Glutamax-1 (31966; Thermo Fisher Scientific) containing DNAse-I (D5025, Millipore-Sigma). After 2 7 min centrifugations at 700 g, cells were resuspended in a medium containing DMEM Glutamax-1, penicillin (100 U/ml)/streptomycin (100 μg/ml) (15140; Thermo Fisher Scientific), 5% (v/v) fetal bovine serum, N2 supplement (17502048; Thermo Fisher Scientific), and B-27 supplement (17504-044; Thermo Fisher Scientific) to a final density of 50 million cells/ml. Cortical cells were then seeded in the somatic compartment. Cell culture medium was added equally to the 4 reservoirs. Microfluidic chips were placed in Petri dishes containing 10% EDTA (Millipore-Sigma) and incubated at 37°C in a 5% CO₂ atmosphere. The culture medium was renewed every 6 d. Upon differentiation, cortical axons entered the microchannels and reached the second chambers after 4–5 d. Cortical axons continued growing thereafter.

Nmrk2-KO mouse

Nmrk2-KO mice were obtained by the injection of embryonic stem cells (Nmrk2^{tm1(KOMP)Vlcg} allele) from the Knockout Mouse Project (KOMP) into C57BL/6N blastocysts at Centre National de la Recherche Scientifique–Transgenesis, Archiving and Animal Models (CNRS–TAAM) Transgenic Animal Facility (SEAT; Villejuif, France). All exons and introns of the Nmrk2 gene are replaced by a lacZ reporter gene to create a null allele in embryonic stem cells (University of California, Davis, Davis, CA, USA; https://www.komp.org/geneinfo.php?geneid=65082). Chimeric male mice were used to obtain heterozygous Nmrk2-mutant mice on a C57BL/6N background. Homozygous Nmrk2-KO mice were viable and fertile. We validated the deletion by sequencing the genomic DNA from the KO mice.

Pharmacological treatments

Cells were pretreated and/or treated between 11 and 13 d in vitro. To ensure fluidic isolation, a differential hydrostatic pressure between compartments was maintained.

The following compounds were used: NMDA, (100 μM; M3262; Millipore-Sigma), FK866 (10 μM; F8557; Millipore-Sigma), NAD⁺ (N3014; Millipore-Sigma), NMN (N3501; Millipore-Sigma), NR (ChromaDex, Irvine, CA, USA), NAM (N0636; Millipore-Sigma), dipyridamole (DP, 50 μM; D9766; Millipore-Sigma), cytidine monophosphate (CMP; 25 mM; C1006; Millipore-Sigma), and (+)MK-801-maleate (10/50 μM; 0924; Tocris Bioscience, Bristol, United Kingdom).

Immunofluorescence detection and image acquisition

At various times, cultures were fixed with 4% (w/v) paraformaldehyde (15714-S; Euromedex, Souffelweyersheim, France) +4% (w/v) sucrose (S0389; Millipore-Sigma) for 20 min at room temperature. Cells were then washed once with PBS for 10 min and permeabilized for 30 min with D-PBS containing 0.2% (v/v) TritonX-100 (Millipore-Sigma) and 1% (w/v) BSA (Millipore-Sigma). Immunostaining was performed as described in Magnifico et al. (6) using the following conjugated antibodies diluted in D-PBS: anti-βIII tubulin-Alexa Fluor 488 (1:500, AB15708A4; Millipore-Sigma) and anti-microtubule-associated protein 2 (MAP2)-Alexa 555 (1:500, MAB3418A5; Millipore-Sigma). Cell nuclei were stained by using Hoechst 33342 (2 μg/ml). The chips were then rinsed once with PBS and filled with PBS+0.1% sodium azide.

Images were acquired with an Axio-observer Z1 microscope (Zeiss, Wetzlar, Germany) fitted with a cooled CCD camera (CoolsnapHQ2; Roper Scientific, Trenton, NJ, USA). The microscope was controlled with Metamorph (Berlin, Germany) software, and images were analyzed with ImageJ software (National Institutes of Health, Bethesda, MD, USA).

Quantification of axonal fragmentation and neuronal death

Axonal fragmentation was assessed with a macro developed in ImageJ software that contained the following steps: 1) subtract background, 2) plug in tubeness (σ = 1), 3) plug in Otsu thresholding, and 4) analyze particles by measuring total particle area (A_TOTAL) and identifying particles with circularity higher than 0.2 (A_0.2). The fragmentation index is obtained as the ratio A_0.2/A_TOTAL. This index is almost linearly correlated with fragmented axon percentages: indices under 0.2, ~0.5, and up to 0.8 correspond to <10, 50, and more than 90% of fragmented axons, respectively. Neuronal death was quantified by calculating the ratio of condensing nuclei to total nuclei.

PCR and real-time quantitative PCR

Primary cortical neurons were lysed at DIV 11, and tissues were lysed from 8-wk-old mice. Three wells were pooled to obtain 1 sample. Total RNA was isolated with an RNeasy kit (74104; Qiagen, Valencia, CA, USA) according to the manufacturer’s protocol. From the total RNA, 1 μg was RNase free-DNase treated and retrotranscribed with a SuperScript II reverse transcriptase kit (18064–022; Thermo Fisher Scientific) according to the manufacturer’s instruction. The remaining RNA was hydrolyzed using RNase H from E. coli (AM2293; Thermo Fisher Scientific).

PCR amplification was performed in a thermocycler by mixing 2 μl of cDNA with 0.5 μM of forward and reverse nucleotide sequences (Table 1) in a PCR buffer containing dNTP mix (200 μM each), MgCl₂ (1.5 mM), and 0.5 U of Taq DNA polymerase (18038; Thermo Fisher Scientific). Initially, the template was denatured by heating to 95°C for 5 min, followed by 30 amplification cycles. Each cycle consisted of 3 steps: the first for 45 s at 94°C, the second for 30 s at annealing temperature (depending of the amplified mRNA), and the third for 1 min at 72°C for polymerization. The final step was an extension at 72°C for 10 min. PCR products were electrophoresed on 1.5% (w/v) agarose gel and stained with ethidium bromide (0.5 μg/ml).

Quantitative PCR was performed with a LightCycler480 (Roche Applied Science, Penzberg, Germany), by mixing 1:50 diluted cDNA with the target-specific primers and Light Cycler 480 SYBRGreen I Master mix (4707516001; Roche, Basel, Switzerland) according to the manufacturer’s instructions. Amplification conditions were initial denaturation for 5 min at 95°C followed by 40 cycles of 10 s at 95°C, 15 s at annealing temperature and 10 s at 72°C. Results were analyzed with LightCycler 480 SW 1.5 software. Individual PCR products were analyzed by melting-point analysis. The expression level of a gene was normalized relative to that of mouse β-actin or hypoxanthineguanine phosphoribosyltransferase (HPRT).

NAD⁺ assay

NAD⁺ was measured using EnzyChrom NAD⁺/NADH Assay Kit (E2ND-100; Euromedex) according to the manufacturer’s protocol. Targeted quantitative NAD⁺ metabolomics were performed as described in Trammell and Brenner (38).

In vivo NMDA-induced excitotoxicity experiments

Experiments were performed on male C57/BL6 mice (23 ± 3 g, produced and provided by GIP Cycéron animal facilities) in accordance with European Union Directives (210/63/UE) and French Ethics laws (R214; 87-137; Ministère de l’Agriculture) on animal experimentation and were approved by the Université Caen Normandie Animal Welfare Committee. Mice were deeply anesthetized with isoflurane 5% and maintained under anesthesia with 2% isoflurane in a 70/30% gas mixture (N₂O/O₂) during surgery. The rectal temperature was maintained at 37 ± 0.5°C throughout the surgical procedure using a feedback-regulated heating system.

A cortical unilateral injection [coordinates: 0.5 mm posterior, 2.0 mm lateral, −0.5 mm ventral to the bregma (39)] of NMDA (5 nmol; in 0.33 μl), coinjected or not with NAD⁺ or NR (50 nmol) was performed after placing the animals in a stereotaxic frame. Solutions were injected by the use of a micropipette made with hematologic micropipettes (calibrated at 15 mm/μl, assistant ref 555/5; Hecht, Sondheim-Rhoen, Germany). The micropipette was removed 3 min later. The volume of the lesion was analyzed 48 h later.

MRI

Experiments were performed on a Pharmascan 7T (Bruker, Bremen, Germany). T2-weighted images were acquired with a multislice multiecho sequence: TE/TR 51.3/2500 ms 48 h after injection. Lesion volumes were quantified on MRI with ImageJ software.

Statistical analysis

Differences were assessed by a Mann-Whitney U test to compare 2 conditions or a Kruskal-Wallis test to compare >2 conditions.

NMDA treatment on neuronal cell bodies induces neuronal death and axonal degeneration

To dissect the protective effects of NAD⁺ and NR on AxD, we cultured cortical neurons in microfluidic devices containing 3 chambers connected to each other (37). When seeded in the left chamber, cortical neurons projected their axons from the proximal to the distal chamber through asymmetric filtering microchannels, whereas the somata and dendrites were confined in the left chamber. As a consequence of such compartmentalization, the right chamber of the microfluidic device contained only axons. When depressurized, the central chamber acts as siphon and ensures optimal fluidic isolation of the distal axonal and somatodendritic chambers (Fig. 1A, B).

Analysis of dendrites (MAP2, red), nuclear chromatin (Hoechst 33342, blue), and axons (β3-tubulin) indicated that the seeded cortical neurons were healthy after 2 wk in culture (Fig. 1C). Somatodendritic application of 100 μM NMDA in the proximal chamber induced dendrite degeneration marked by a loss of MAP2 immunostaining and nuclear condensation associated with a severe AxD visualized by punctiform β3-tubulin immunostaining in the distal chamber (Fig. 1C–G). By contrast, local application of NMDA to axons in the distal chamber had no effect on neurons, consistent with the fact that NMDA receptors (NMDARs) are exclusively localized to the somatodendritic compartment (22) (Fig. 1E–G). NMDAR involvement in NMDA-induced neurodegeneration was confirmed with the noncompetitive NMDAR antagonist MK-801. Nuclear condensation and AxD after 8 h NMDA treatment were fully blocked by coaddition of 10 or 50 μM MK-801 (Fig. 1H, I).

NR has a stronger neuroprotective effect than NAD⁺ during excitotoxicity, both in vitro and in vivo

To evaluate and compare the potential neuroprotective properties of NAD⁺ and NR, cortical neurons were treated with 100 μM NMDA in the presence or absence of 5 mM NAD⁺ or 1 mM NR. Nuclear condensation and AxD were then quantified over time. As shown in Fig. 2A, after 24 h of treatment, no inhibition of dendrite degeneration (MAP2) nor inhibition of nuclear condensation was observed after NAD⁺ or NR treatment, whereas both molecules delayed AxD (Fig. 2A–C). NAD⁺ protection was weak (67% of protection compared to NMDA effect after 8 h of treatment, 22% after 24 h), whereas NR fully protected axons after 8 h and still provided 72% protection 24 h after NMDA treatment (Fig. 2C). NAD⁺ and NR efficiencies were also compared in a dose–range assay in cortical neurons cotreated with 100 μM NMDA for 24 h (Fig. 2D). NAD⁺ started to delay AxD from 5 mM. In contrast, NR was efficient at a 10-fold lower dose (0.5 mM; 33 vs. 43% of protection) with a maximum protective activity at 1 mM. We asked whether the absence of protection of NAD⁺ and NR on NMDA-induced nuclear condensation was due to the high-dose of NMDA (100 μM) leading to irretrievable degeneration. Accordingly, we challenged 5 mM NAD⁺ or 1 mM NR with a lower 10 μM NMDA concentration during 24 h (Supplemental Fig. S1). A lower nuclear condensation was observed with 10 μM NMDA (30%) compared with that obtained with 100 μM NMDA (53%) but neither NAD⁺ nor NR had any protective effect on this phenomenon.

Considering the different neuroprotective activities of NAD⁺ and NR in our in vitro model of excitotoxicity, we asked if NAD⁺ and NR could be neuroprotective in an in vivo setting of excitotoxicity. For this purpose, we coinjected NMDA (5 nmol) and NAD⁺ or NR (50 nmol each) into mouse cortical structures as published (40). Strikingly, the volume of the NMDA-induced lesion as measured by MRI was significantly reduced after NR coinjection, whereas NAD⁺ did not reveal a significant protection (Fig. 2E, F). Thus, NR has a more potent neuroprotective effect than NAD⁺ in NMDA-induced neurodegeneration, both in vitro and in vivo.

NR is a more powerful brain NAD⁺-boosting compound than NAD⁺

We performed targeted quantitative NAD⁺ metabolomics to evaluate the effect of NR and NAD⁺ on NMDA-treated mouse brains (38). As shown in Fig. 2G, NR significantly elevated brain NAD⁺ from <400 pmol/mg to nearly 500 pmol/mg, whereas the increase in brain NAD⁺ from NAD⁺ injection was not significant. Consistent with an elevated NAD⁺ metabolome with NR as a neuroprotective compound, NR tended to elevate brain NAM and adenosine diphosphate receptor (ADPR), whereas no changes were evident in brains treated with NAD⁺ (Fig. 2H, I).

NR has a local protective effect mainly restricted to the axonal compartment

Previous results have shown that NAD⁺ delays Wallerian degeneration after axotomy on PNS-derived neuronal cells. Furthermore, both NAD⁺ and WldS protein prevent AxD in a soma-independent fashion (8, 41). Because NAD⁺ and NR protect AxD in our model, we examined if NAD⁺ and NR act locally at the axon or also depend on the somatic compartment. NAD⁺ or NR were applied either on both compartments or the somatodendritic or the axonal compartment simultaneously with somatodendritic addition of 100 μM NMDA for 24 h. Axonal application of NR partially protected axons from NMDA-induced degeneration (50% protection), whereas somatodendritic application did not prevent it (Fig. 3A–F). Local NAD⁺ treatment revealed no significant protective effects (Fig. 3G), but for both molecules NR and NAD⁺, the protection was increased or detected when applied on both compartments together. These findings indicate that local axonal treatment with NR is required for axonal protection and that somatic NAD⁺ or NR treatment is insufficient to fully prevent AxD. The data indicate that somatic application of these molecules helps to increase axonal protection induced by axonal application of the same compounds.

NR is converted to NAD⁺ in cortical neurons

Not all models of WldS-dependent neuroprotection invoke a requirement for NAD⁺ biosynthesis (42). Accordingly, we tested the hypothesis that the protective effect of NR depends upon its intracellular conversion to NAD⁺ (Fig. 4A). First, we asked whether NR enters cortical neurons for further conversion to NAD⁺. NR is a nucleoside reported to cross plasma membranes through NT (25, 27). Among NT isoforms, expression analysis in our neuronal cell culture model indicated that all equilibrative nucleoside transporter (ENT1-4) transcripts were expressed (Fig. 4B). However, none of the concentrative nucleoside transporter (CNT1-3) mRNAs could be detected. DP, a potent pharmacological inhibitor of ENT1 and -2 activities, blocks NR cell entry in human cell lines (25, 43). Cotreatment of cortical neurons with 100 μM NMDA, 1mMNR, and 50 μM DP almost completely prevented NR-induced axonal protection (axonal protection decreased by 68%; from 88 to 20%), establishing that the neuroprotective effect of NR requires DP-sensitive transport through the plasma membrane (Fig. 4C).

Intracellular NR conversion to NAD⁺ is initiated by Nmrk1/2, which generates the NAD⁺-proximal precursor, NMN (Fig.4A). We analyzed Nmrk1 and -2 mRNA expression by quantitative RT-PCR from both specific mouse brain areas and from cortical neurons in culture. We discovered that Nmrk1 mRNA is constitutively expressed in all tissues, whereas Nmrk2 mRNA was absent in all brain areas (cortex, hippocampus, and cerebellum) (Fig. 4D, E). However, Nmrk2 transcript was highly expressed in heart and striated skeletal muscles, as previously described (30, 44). Nmrk1 but not Nmrk2 mRNA was also expressed in cultured cortical neurons (Fig. 4E). Nmrk2 mRNA was reported to be induced in PNS neurons after sciatic nerve axotomy (7). To investigate a potential induction of Nmrk2 after an excitotoxic insult in cortical neurons, Nmrk2 mRNA was quantified after 100 μM NMDA treatment for 24 h. We found a low induction of Nmrk2 transcript, but a 4-fold increase of Nmrk1 mRNA. NR kinase activity leads to formation of NMN, which is adenylylated to NAD⁺ via Nmnat enzymes. We analyzed Nmnat1-3 mRNA expression in cortical neurons treated or not with NMDA 100 μM for 3 or 24 h. We showed that all the transcripts are expressed, with the highest level for the cytosolic isoform Nmnat2, followed by the mitochondrial enzyme Nmnat3 (Supplemental Fig. S2A). Each of the mRNAs was induced by 24 h NMDA treatment such that all components of the NR kinase pathway to NAD⁺ biosynthesis were upregulated during neuronal insult. NR conversion to NAD⁺ can be initiated by nucleoside phosphorylase, which converts NR to NAM, which is further converted to NAD⁺ (45). However, cotreatment of cortical neurons with NMDA 100 μM and NAM 5 mM for 24 h demonstrated that NAM is unable to prevent NMDA-induced AxD (Supplemental Fig. S2B). We next evaluated NAD⁺ levels in neurons during an excitotoxic challenge. Treatment of cortical neurons with 100 μM NMDA for 24 h induced a 36% decrease in intracellular NAD⁺. This effect was reverted by cotreatment with 1 mM NR (Fig. 4F). Together, these results establish that conversion of NR to intracellular NAD⁺ via the transcriptionally induced NR kinase pathway is critical for maintaining axoplasmic NAD⁺ in opposition to excitotoxicity-induced AxD.

Extracellular NAD⁺ involves a potentially unique mechanism to prevent excitotoxicity-induced AxD

Extracellular NAD⁺ has a lower axon-protective effect than NR in our model. As reported previously in cell lines and PNS-derived neuronal cells, NAD⁺ can be converted extracellularly to NMN by ENPPase and then to NR by 5′-ENTase (Fig. 5A). We asked whether extracellular conversion of NAD⁺ to NR is necessary for NAD⁺-mediated axonal protection or if this effect is independent of NR action. For this purpose, we probed 5′-ENTase activity in cortical neurons, using a high concentration of CMP, a competitive inhibitor of this enzyme (29). First, we treated cortical neurons with 100 μM NMDA, with or without 5 mM NMN for 24 h. Extracellular NMN significantly reduced NMDA-induced AxD and 25 mM CMP reversed the protective effect of NMN (Fig. 5B). Consistent with data in other tissues (28, 30), these data indicate that NMN must be converted to NR to mediate axonal protection. The beneficial effect of NMN was also prevented when cortical neurons were cotreated with NMDA and the ENT1/2 inhibitor DP, confirming that NMN is converted to the nucleoside NR before plasma membrane transport to prevent NMDA-induced AxD. Consistent with a dependence on conversion to NR, NR was more protective at a lower concentration than NMN. Surprisingly, 5 mM NAD⁺ was slightly but significantly protective, yet neither CMP nor DP blocked the axonal protection conferred by this high concentration of NAD⁺ (Fig. 5C).

NAD⁺ and NR protect neurons with the same efficiency after FK866-induced NAD⁺ depletion

Because excitotoxicity and SARM1 activation induce NAD⁺ depletion in a manner that can be protected by NR and NAD⁺, we tested whether pharmacological inhibition of NAD⁺ homeostasis causes AxD. FK866, a specific inhibitor of Nampt (46), which salvages NAM for resynthesis of NAD⁺ (Fig. 6A), was used to induce NAD⁺ depletion. Addition of 10 μM FK866 to the somatodendritic compartment of cortical neurons induced a rapid decrease of NAD⁺ levels after 24 h of treatment (Fig. 6B). At 72 h, this drop in NAD⁺ induced somatodendritic degeneration, nuclear condensation, and AxD (Fig. 6C, D). Cotreatment of cortical neurons with 10 μM FK866 and 50 μM NAD⁺ or NR fully prevented these effects (Fig. 6D–F). Notably, the protective concentrations of NAD⁺ and NR were lower than those used in the NMDA-induced neurodegeneration model (50 μM for both in the FK866 model and 5 mM NAD⁺ or 1 mM NR in the NMDA model) (Figs. 2A–D and 6D–F). NAD⁺ and NR remained fully protective when applied 24 h after treatment, at a time when NAD⁺ depletion was already observed, but dropped to 20% protection when applied 48 h later (Fig. 6E, F). These findings highlight the existence of a temporal window in which neuronal integrity can be rescued after a neurotoxic insult. Taken together, these data show that NAD⁺ and NR have the same neuroprotective effect against FK866-induced NAD⁺ depletion.

Extracellular NAD⁺ conversion into NR is needed to prevent FK866-induced neurodegeneration

We next evaluated the metabolic pathway from extracellular NAD⁺/NMN/NR to intracellular NAD⁺ in FK866-induced neurodegeneration. First, cortical neurons were cotreated with 50 μM FK866, 50 μM NR, and50 μM DP for 72 h, and axonal fragmentation was investigated. DP only partially limited the NR-protective effect against FK866 (axonal protection reduced by 40%), whereas the protective effect of NR was totally inhibited under NMDA stress (Figs. 4C and 7A). These data suggest that FK866 treatment induces expression of NT that is relatively resistant to DP. We analyzed ENT1-4 mRNA in cortical neurons treated or not with 10 μM FK866 for 72 h. ENT2 and -4 mRNA were induced 2.2- and 1.7-fold, respectively, by FK866 (Supplemental Fig. S3). NAD⁺ and NMN also prevented FK866-induced neurodegeneration in a manner that was partially prevented by DP (Fig. 7B, C). As NAD⁺ may depend on NT in protection against FK866 neurodegeneration, we tested whether extracellular NAD⁺ and NMN need to be converted to NR before transport through plasma membrane. Indeed, CMP fully blocked the axon-protective effect of NAD⁺ and NMN, indicating that both molecules require 5′-ENTase activity to be converted to NR and mediate their axon-protective effect (Fig. 7B, C). We also analyzed Nmrk1 and -2 mRNA expression in cortical neurons treated or not with FK866 for 24–72 h. As expected, Nmrk1 mRNA was highly expressed compared with the Nmrk2 transcript, which was not detectable (Supplemental Fig. S4B, C). As previously described with NMDA, FK866 treatment for 48 or 72 h induced Nmrk2 (3-fold at 72 h). As Nmrk2 induction was a common phenomenon observed in our experiments, we used cortical neurons from wild-type or Nmrk2-KO mice. These neurons were treated with 10 μM FK866 for 72 h before condensed nuclei and axonal fragmentation were analyzed. Even in the absence of the Nmrk2 gene, NAD⁺ and NR provided full protection against FK866-induced neurodegeneration, indicating that Nmrk2 is not required for these effects (Supplemental Fig. S4D, E). These results demonstrate that NAD⁺, NMN, and NR use a similar path way, likely independent of Nmrk2, to prevent FK866-induced neurodegeneration.