Immuno-EM characterization of vasopressin neurons in macaques by use of formaldehyde-fixed tissues stored for several years

Translational research often requires the testing of experimental therapies in primates, but research in non-human primates is now stringently controlled by law around the world. Tissues fixed in formaldehyde without glutaraldehyde have been thought to be inappropriate for use in electron microscopic analysis, particularly those of the brain. Here we report the immunoelectron microscopic characterization of arginine vasopressin (AVP)-producing neurons in macaque hypothalamo-pituitary axis tissues fixed with 4% formaldehyde and stored at –25°C for several years. The size difference of dense-cored vesicles between magnocellular and parvocellular AVP neurons was detectable in their cell bodies and perivascular nerve endings located, respectively, in the posterior pituitary and median eminence. Furthermore, glutamate and the vesicular glutamate transporter 2 were colocalized with AVP in perivascular nerve endings of both the posterior pituitary and the external layer of the median eminence, suggesting that both magnocellular and parvocellular AVP neurons are glutamatergic in primates. Both ultrastructure and immunoreactivity can therefore be sufficiently preserved in macaque brain tissues stored long-term for light microscopy. Taken together, these results suggest that this methodology could be applied to the human post-mortem brain and be very useful in translational research.


Introduction
Studies using macaque monkeys as model non-human primates are essential because of their high applicability to human clinical research which often requires the testing of experimental therapies in primates. However, primate research is now stringently controlled 5 by law around the world (Abbott, 2014). Rodents, in which research is less restricted, are often not analogous enough to humans in their physiology to allow an extrapolation from rodent to human. Medical science is urged to make minimal use of animals such as primates, and basic studies on non-human primates appear to be permitted only if the work could not be carried out in any other species. Thus, although tissues from primates are often essential to 10 achieve relevant results, there are, nevertheless, considerable problems in their use.
Arginine vasopressin (AVP), an anti-diuretic hormone, is released, not only into the blood stream, but also into the central nervous system where, in mammals, it has been shown to be important for stress coping, aggression, courtship behavior, learning, bonding, and various socio-sexual behaviors (Walum & Young, 2018). AVP is produced mainly by 15 magnocellular neurosecretory neurons in the supraoptic nucleus (SON) and paraventricular nucleus (PVN) of the hypothalamus (Buijs et al., 1983). The AVP gene encodes a precursor containing AVP, AVP-associated neurophysin II (NPII), and a glycopeptide copeptin (Breslow, 1993;Davies et al., 2003;Land et al., 1982). The expression and release of AVP by magnocellular neurosecretory neurons in the SON and PVN are regulated by physiological 20 conditions, including plasma osmotic pressure and blood pressure (Koshimizu et al., 2012).
The magnocellular axons project primarily via the internal layer of the median eminence to the posterior pituitary where they release AVP into the systemic circulation. In addition, some parvocellular neurons in the PVN produce AVP and project into extrahypothalamic areas where the AVP and/or other co-packaged molecules regulate brain function as neuromodulators (Morris, 2020).
Corticotrophin-releasing factor (CRF) is a strong stimulator of adrenocorticotrophic hormone (ACTH) secretion from the anterior pituitary when released onto portal capillaries in the median eminence in response to the stress (Vale et al., 1981). A population of 5 parvocellular neurons in the anteromedial part of the PVN produces CRF and may, depending on the extent of stress, also produce AVP. Both peptides released into the hypothalamohypophysial portal circulation play an important synergistic role in stress resilience (Fellmann et al., 1984;Itoi et al., 2014;Mouri et al., 1993;Whitnall et al., 1985). Because intense AVP-and CRF-immunoreactivity have been observed in the external layer of the 10 macaque median eminence, the peptides are probably co-released into the portal circulation to amplify ACTH release from the primate anterior pituitary (Otubo et al., 2020;Zimmerman et al., 1973).
In rodents, the presence of glutamate-immunoreactivity in magnocellular neuroendocrine cells of the SON suggests that AVP neurons also produce glutamate as a 15 neurotransmitter (Meeker et al., 1991;Meeker et al., 1989). Within the neurosecretory endings of the posterior pituitary, glutamate immunoreactivity is specifically localized to electron-lucent microvesicles with no overlap onto the dense-cored neurosecretory vesicle (dcv) population in rats (Meeker et al., 1991). Immunocytochemical co-localization of CRF and the vesicular glutamate transporter 2 (VGLUT2) in rats suggests that the co-release of 20 CRF and glutamate may function to regulate postsynaptic targets (Valentino et al., 2001). It is currently unclear whether glutamate has a similar or other functions in the primate AVP/CRF system.
Tissues fixed in formaldehyde without glutaraldehyde have long been thought to be inappropriate for use in electron microscopic analysis, especially in the brain (Maunsbach, 25 1966a(Maunsbach, 25 , 1966b. Here we report the immunoelectron microscopic characterization of AVPproducing neurons in the primate hypothalamo-pituitary axis tissue fixed with formaldehyde and stored at -25ºC for several years. Special attention was paid to the size of dcv in AVPproducing magno-and parvocellular neurons and to the colocalization of CRF with AVPrelated gene products in the dcv. We show that immunoelectron microscopy of formaldehyde-5 fixed tissue can confirm the size difference in dcv between magno-and parvocellular AVP neurons in Japanese macaque monkeys. Furthermore, we show that, in macaque monkeys, AVP/CRF and the VGLUT2/glutamate are co-localized in both the magnocellular endings of the posterior pituitary and the parvocellular endings in the external layer of the median eminence. 10

posterior pituitary
Full details of all the antibodies used in this study are shown in Table 1. We first validated by Western blot analysis the specificity for the VGLUT2 and NPII antibodies in Japanese macaque monkeys. Western blot analysis demonstrated the specificity for the guinea pig polyclonal antibody against VGLUT2 and the expression of VGLUT2 at the protein level 20 in the posterior pituitary; a single strong immunoreactive band was detected at the expected molecular weight of ~56 kDa (Balaram et al., 2013) (Fig. 1). The mouse monoclonal antibody against AVP-neurophysin II (NPII) also detected a single major band at the expected molecular weight of ~12 kDa on the blots of posterior pituitary (Kawakami et al., 2021;Otubo et al., 2020) (Fig. 1). 25

Anatomy of AVP-producing neurons in the macaque hypothalamus
Immunostaining for NPII was performed to study the localization of AVP neurons in the hypothalamus of Japanese macaque monkeys. Numerous magnocellular (large-sized) AVP neurons were observed in the PVN ( Fig. 2A) and SON (Fig. 2B). Parvocellular (small-5 sized) AVP neurons were also observed in the PVN (Fig. 2C). The median eminence can be divided into an internal and an external layer. The internal layer is the zone through which the axons of magnocellular neurons densely project into the posterior pituitary ( Fig. 2D) (Buijs et al., 1983;Morris, 2020;Zimmerman et al., 1973). The external layer of the median eminence is where the axons of many parvocellular neurosecretory neurons, including parvocellular 10 AVP/CRF neurons of the PVN, end on hypothalamo-hypophysial portal vessels and release anterior pituitary-hormone-regulating hormones (Fig. 2D) (Buijs et al., 1983;Morris, 2020;Zimmerman et al., 1973).

Size characterization of dcv
In this study, immunoelectron microscopic characterization of AVP-producing neurons was performed in macaque hypothalamo-pituitary axis tissues fixed with 4% formaldehyde and stored at -25ºC for several years. We found that both ultrastructure and 25 immunoreactivity are sufficiently preserved in macaque brain tissues stored long-term for light microscopy. Approximately 200 dcv in the neurosecretory axons in the posterior pituitary (magnocellular), the internal layer of the median eminence (magnocellular), and the external layer of the median eminence (parvocellular) were selected and their major axis diameters were measured. The average of major axis was 196 ± 2 nm in the posterior 5 pituitary, 197 ± 2 nm in the internal layer of the median eminence, but significantly smaller (93 ± 1 nm) in the external layer of the median eminence ( Fig. 5; Table 2).
Next, we measured the diameter of dcv in the perikarya of randomly selected SON and PVN AVP neurosecretory neurons (at x5,000 magnification; Table 3). In the SON (7 cell bodies), the average diameter of dcv contained in the cell body was 204 ± 12 nm (11 vesicles), 10 confirming that, as expected, all AVP neurons in the SON are magnocellular ( Fig. 6; Table 3).
The 18 AVP neurons in the PVN selected for analysis formed two distinct groups: those with larger dcv (180 ± 19 nm; 11 vesicles; Fig. 7A

A subpopulation of parvocellular AVP neurons in the PVN co-expresses CRF
In the PVN, of the 18 AVP neuronal cell bodies analyzed, 6 were identified as magnocellular on the basis of vesicle size; none of these showed any CRF-immunoreactivity.
Of the 12 PVN perikarya identified as parvocellular on the basis of vesicle size, a subpopulation (especially in the parvocellular part of the PVN) co-expressed CRF (Fig. 8) 20 and such neurons were more intermingled with magnocellular neurons in the macaque monkey than in rodent PVN as reported previously (Foote & Cha, 1988;Kawata & Sano, 1982;Otubo et al., 2020). The terminal regions of parvocellular AVP/CRF neurons in the external layer of the median eminence were next analyzed (Fig. 9). In these cells, the major axis diameter of the sparse dcv was 106 ± 3 nm (16 vesicles) (Fig. 10) for the vesicles in the 25 CRF-immunoreactive terminals, comparable to the 113 ± 9 nm (12 vesicles) in the cell body ( Fig. 11). In two putative parvocellular neurons (#24 and #25), the mean major axis of the dcv appeared slightly larger which might relate to the maturity of the vesicles (Table 3) (Cannata & Morris, 1973).

Discussion
It is often assumed that glutaraldehyde, a cross-linking fixative, is essential for the preservation of fine ultrastructure (Maunsbach, 1966a(Maunsbach, , 1966b and that formaldehyde-only fixed tissues are inappropriate for electron microscopic analyses. However, the use of glutaraldehyde often results in a reduction of immunoreactivity and, for this reason, formaldehyde-fixed tissues without glutaraldehyde are used in most cases for 5 immunohistochemical analysis at the light microscopic level (Maunsbach, 1966a(Maunsbach, , 1966b.
The current study demonstrates that, in the macaque brain, interpretable ultrastructure and adequate ultrastructural immunoreactivity are both preserved in brain tissue even after longterm storage at -25ºC for conventional light microscopy, provided that the fixed tissues are post-fixed with glutaraldehyde after the long-term storage. Thus, formalin-fixed macaque 10 brains stored for a long period, at least 5-years, can be used for immunoelectron microscopic analysis, suggesting that not only macaque brain, but also postmortem formalin-fixed human brain tissue could be amenable to immunoelectron microscopic analysis. Furthermore, the ultrastructure of tissues from rare animals such as endangered species might be studied in this way even if they had been prepared for light microscopic immunohistochemistry and stored 15 at -25ºC in antifreeze solution.
The methodology used in this study has some obvious advantages in the ultrastructural analysis of the nervous system, in which tissue specimens are necessarily small, particularly where the brain is large, as in macaque monkeys and humans. After confirmation of the specificity of antibodies at the light microscopic level, post-embedding 20 immunoelectron microscopic analysis can readily be performed on adjacent sections. Furthermore, because identification of brain nuclei and/or cells is easier at the light microscopic level, they can then be precisely targeted for post-embedding immunoelectron microscopy.
It has been reported that rodent AVP neurons also use glutamate as a neurotransmitter (Meeker et al., 1991;Meeker et al., 1989;Zhang et al., 2020;Ziegler et al., 2002). In this study, we examined the expression of glutamate and VGLUT2 in the macaque posterior pituitary. We first demonstrated by Western blotting that VGLUT2 was expressed in the macaque posterior pituitary, suggesting that magnocellular AVP (and also oxytocin) neurons 5 use glutamate as a neurotransmitter in Japanese macaque monkeys. Immunoelectron microscopic analysis revealed that the AVP/CRF-immunoreactivity was restricted to the dcv, but the VGLUT2-immunoreactivity was associated with the smaller, electron-lucent microvesicles (Fig. 14). Similarly, immunoreactivity for glutamate itself was essentially restricted to microvesicles in the axonal endings of AVP neurons in both the posterior 10 pituitary and external layer of the median eminence (Fig. 14). Because the antibody for glutamate, an amino acid, is raised against the cross-linking site of glutaraldehyde as the antigen, it is thought to be essential that glutaraldehyde is included in the fixative prior to glutamate immunocytochemistry (Meeker et al., 1991;Meeker et al., 1989;Sun et al., 2007).
In our study, although glutaraldehyde was not included in the primary perfusion fixative, 15 post-fixation of the formaldehyde-fixed tissue with glutaraldehyde enabled us to detect glutamate by immunoelectron microscopy even after prolonged storage. Therefore, the subcellular localization of amino acids appears to be preserved for a long-term even with conventional formaldehyde fixation. The immunoelectron microscopic detection of amino acids including glutamate and GABA should also be possible in conventional formaldehyde-20 fixed and stored human brain tissue.
Our results in macaque tissue show that glutamate as an excitatory neurotransmitter is in microvesicles both in magno-and parvocellular AVP neurons. Although the localization of glutamate receptors has been reported in the neural lobe, median eminence and anterior pituitary, a clear role for glutamate secretion at these sites remains unknown (Brann, 1995;25 Hrabovszky & Liposits, 2007). Salt loading has been shown to produce robust increases in the VGLUT2 mRNA of AVP neuron perikarya and in VGLUT2-immunoreactivity in the rat posterior pituitary (Hrabovszky et al., 2006), suggesting that an osmotic challenge could produce an increase in glutamate release in the primate magnocellular AVP system (Fig. 14).
Similarly, a role in the anterior pituitary of glutamate released from parvocellular AVP/CRF 5 neurons into the hypothalamo-hypophysial portal veins remains unclear. Functional implications for glutamate release at the terminals of both magno-and parvocellular AVP neurons remain to be explored.
The size of dcv contained in the terminals of AVP neurons was quantified and compared among the posterior pituitary and internal/external layers of the median eminence. 10 This showed, also for the macaque, the expected significant difference in size of the dcv (Fig.   14). It is known that the posterior pituitary and the internal layer of the median eminence contain the axons of magnocellular AVP neurons of the PVN and SON, whereas the external layer of the median eminence is the terminal region of parvocellular AVP neurons of the PVN (Buijs et al., 1983;Morris, 2020;Zimmerman et al., 1973) and other parvocellular neurons 15 influence the anterior pituitary. In the dcv of the internal layer of the median eminence through which the magnocellular axons pass en route to the posterior pituitary, the AVP precursor should be processed sufficiently to reveal both prominent NPII-and copeptinimmunoreactivity in almost all of the axons (Kawata & Sano, 1982;Otubo et al., 2020;Zimmerman et al., 1973). In the macaque hypothalamus, quantifying only the size of the dcv 20 in the cell body or its axon allows us to determine the magnocellular or parvocellular identify of the cell. Finally, our study confirms at the electron microscopic level that certain axonal endings in the external layer of the median eminence of macaques are derived from parvocellular AVP neurons of the PVN.

Conclusions
We report the immunoelectron microscopic characterization of AVP-producing neurons in primate hypothalamo-pituitary systems by the use of formaldehyde-fixed tissues stored at -25ºC for several years. The size difference in dcv between magno-and 5 parvocellular AVP neurons in Japanese macaque monkeys was determined (Fig. 14). We also demonstrate by immunoelectron microscopy that both magno-and parvocellular AVP neurons are glutamatergic in primates (Fig. 14). These results show, for the macaque brain, that interpretable neuronal ultrastructure and sufficient immunoreactivity for several relevant proteins are preserved even after a long-term storage of tissues fixed in formaldehyde for 10 light microscopy. They indicate that this methodology could be applied to human postmortem brain and, therefore, be very useful in translational research.
Macaques were maintained in a temperature-controlled (22-24ºC) room under a daily 20 photoperiod of 12:12 hour light/dark cycle (lights off at 8:00 p.m.). These animals were checked and shown to be free of specific pathogens. Food and water were available ad libitum. All animals were kept in individual cages. The housing and experimental protocols followed the guidelines of the Ministry of Education, Culture, Sports, Science, and Technology (MEXT) of Japan, and were in accordance with the Guide for the Care and Use 25 of Laboratory Animals prepared by Okayama University (Okayama, Japan), by Tokyo Medical and Dental University (Tokyo, Japan), and by National Institute for Physiological Sciences (Okazaki, Japan). All efforts were made to minimize animal suffering and reduce the number of animals used in this study.

Tissue processing
Hypothalamic sections and posterior pituitaries were obtained from four macaque monkeys (2 females and 2 males) as described previously (Otubo et al., 2020). Briefly, macaques were deeply anaesthetized with an overdose of sodium pentobarbital (~100 mg/kg body weight), and transcardially perfused with physiological saline followed by 4% 10 paraformaldehyde (PFA) in 0.1 M phosphate buffer (PB, pH 7.4). After perfusion, brains and pituitaries were immediately removed and immersed in the same fixative for ~16 hours at 4ºC.
Coronal 30-µm-thick sections of the hypothalamus were prepared with a Linear-Slicer (PRO10, Dosaka EM, Kyoto, Japan). Preparations were rinsed with PB and then immersed in antifreeze solution (30% glycerol, 30% ethylene glycol, 0.36% NaCl, and 0.05% NaN3 in 40 15 mM PB) for storage at -25ºC until use. For Western blot analysis, three female macaques were sacrificed by blood loss under deep pentobarbital anesthesia (see above). Posterior pituitaries were quickly removed, frozen immediately in powdered dry ice, and stored at -80ºC until use.
The information for antibodies used in this study is shown in Table 1.

Post-embedding immunoelectron microscopy 20
The stored, formaldehyde-fixed sections were rinsed with PBS five times for 10 minutes each, and re-fixed in 4% PFA + 0.1% glutaraldehyde in 0.1 M PB for 4-6 hours at for 1 hour at room temperature. After incubation with the primary antibodies, the sections 10 were washed with PBS, then incubated with a goat antibody against rabbit IgG conjugated to 5 or 10 nm gold particles (1:50 dilution, BBI Solutions, Cardiff, UK) and a goat antibody against mouse IgG conjugated to 15 nm gold particles (1:50 dilution, BBI Solutions) and/or a goat antibody against guinea pig IgG conjugated to 5 nm gold particles (1:50 dilution, Nanoprobes, Yaphank, NY, USA) for 1 hour at room temperature. The Abcam antibody to 15 glutamate was raised in rabbit against L-glutamate conjugated to glutaraldehyde as the immunogen. The manufacturer's information shows no measurable cross-reactivity detected against glutamate in peptides or proteins but modest cross-reactivity against D-glutamate. The specificity of this antibody was confirmed by immunostaining in the rat retina (Sun et al., 2007). To intensify the detectability of the immunoreaction for CRF in the cell body, a 20 streptavidin-biotin intensification kit (Nichirei, Tokyo, Japan) was used. Tissues were first incubated with the biotinylated goat anti-rabbit IgG antibody for 10 minutes at room temperature, followed by incubation in avidin-biotin-horseradish peroxidase (HRP) complex solution for 5 minutes at room temperature. The sections were then washed with PBS, incubated with the goat antibody against HRP conjugated to 6 nm gold particles (1:50 25 dilution, Jackson ImmunoResearch Laboratory, West Grove, PA, USA) for 1 hour at room temperature.
Triple immunoelectron microscopy with antibodies against CRF, NPII, and glutamate was performed by using the front and back of ultrathin sections mounted on nickel grids without a supporting film. First, immunocytochemistry with a pair of primary antibodies 5 (CRF and NPII) was performed on one side of the section and detected using 15 nm (mouse) and 10 nm (rabbit) colloidal gold particles (BBI Solutions), respectively. Next, immunocytochemistry with the other primary rabbit antibody (against glutamate) was performed on the other side of the section and detected by use of 5 nm colloidal gold particles (BBI Solutions). 10 Finally, the sections were contrasted with uranyl acetate and lead citrate and viewed using an H-7650 (Hitachi, Tokyo, Japan) or JEM-1010 (JEOL, Tokyo, Japan) electron microscope operated at 80 kV. The information for antibodies used in this study is shown in Table 2. Signal Solution 1 (TOYOBO). The blotted membranes were rinsed three times with 0.05% Tween 20 in Tris-HCl-buffered saline (TBST) and incubated with horseradish peroxidaseconjugated goat polyclonal antibody against mouse IgG (Bio-Rad Laboratories) or guinea pig IgG (Bioss Inc., Woburn, MA, USA) 1:10,000 dilution in Can Get Signal Solution 2 (TOYOBO) for 1 hour at room temperature. After washing for three times with TBST, blots 5 were visualized by Immun-Star WesternC Chemiluminescence Kit (Bio-Rad Laboratories).
The information for antibodies used in this study is shown in Table 2.

Statistical analysis
All data values were expressed as mean ± standard error of the mean (SEM). 10 Statistical significance was determined as P < 0.05 using one-way analysis of variance (ANOVA). When significant difference was found by ANOVA, the post hoc Tukey-Kramer test was performed. All the various analyses in this study were conducted "blind". Nomura for their technical support. We are grateful to Dr. Wylie Vale for the supply of an antiserum against CRF.

Conflicts of Interest:
The authors declare no conflict of interest.  Corticotropin-releasing factor is preferentially colocalized with excitatory rather than inhibitory amino acids in axon terminals in the peri-locus coeruleus region.
Neuroscience, 106 (2)    vesicles distributed in the posterior pituitary (black), internal layer of the median eminence (ME) (red), and external layer of the ME (green) in macaque monkeys. The major axis of dense-cored vesicles in the posterior pituitary and the internal layer of the ME (~200 nm) was significantly larger than that of the vesicles in the external layer of the ME (~100 nm). *P < 0.05, vs. posterior pituitary and internal layer of the ME. 10   factor (CRF) (B) and vasopressin-associated neurophysin (NPII) in the macaque median eminence (ME). Immunoreactivity against copeptin (A; green) or CRF (B; green) and NPII (magenta) is merged in each right-hand panel (overlap; white). The outlined areas in merged images are enlarged. Arrowheads in (A) indicate tissue doubly immunopositive for copeptin and NPII. Scale bars, 50 µm, and 10 µm in enlarged images. InL, internal layer of the ME; 10 ExL, external layer of the ME.     vesicles (dcv) between magnocellular (~200 nm) and parvocellular (~100 nm) vasopressin (AVP)-and/or corticotrophin-releasing factor (CRF)-producing neurons in the hypothalamus of the Japanese macaque monkey. We also demonstrate that both magno-and parvocellular AVP/CRF neurons are glutamatergic in primates. AP, anterior pituitary; ME, median eminence; MV, clustered microvesicles; PP, posterior pituitary; PVN, paraventricular 10 nucleus; SON, supraoptic nucleus.  corticotrophin-releasing factor; External ME, external layer of the median eminence. *P < 0.05, vs. posterior pituitary and internal ME. All data values are presented as mean ± standard error of the mean (SEM).