C57BL6 (Taconic) female mice were housed in a standard animal facility condition, Laboratory of Experimental Biomedicine at Gothenburg University, and fed phytoestrogen-free chow ad libitum (Harlan Rodent Diet, 2016).

To avoid confounding endogenous sex hormone effect and mimic a postmenopausal state, mice were ovariectomized. Through a midline incision of the peritoneum followed by flank incisions, the ovary was removed, and the skin was closed with metallic clips. The same procedure was performed for sham surgery, except removal of the ovaries. The mice were sedated with isoflurane (Baxter Medical AB) and Metacam (Boehringer Ingelheim Animal Health) was used as preoperative analgesia. The sample size was calculated using the G*power software using data from a previous study for the sialic acid effect on IgG in the OVX-vehicle and OVX-E2 groups in the immune-induced mice (15). With an alpha value of 0.05, a power of 80%, and a calculated Cohen’s d effect size of 1.66, our study would require a sample size of 7 per group to attain significance.

On termination, all mice were sedated with Ketamine (Richter Pharma) and Dexmedetomidine (Orion Pharma Animal Health), blood was collected, and the mice were euthanized by cervical dislocation. Uterus, thymus, and gonadal fat were dissected and weighed.

Mice tibia bone was dissected and fixed in 4% buffered paraformaldehyde for 48 hours followed by storage in 70% ethanol until subsequent bone mineral density (BMD) assessment. Trabecular BMD was assessed at the metaphyseal region of the tibia bone by pQCT scan analysis with Stratec pQCT XCT Research M software (software version 5.4; Norland, Fort Aktison, WI, USA) at a resolution of 70 µm as previously described (23). Cortical thickness was determined with a scan in the mid-diaphyseal region of the tibia bone with a scan of 36% from the growth plate.

Blood samples were taken before surgery, during the experiment, and on the termination. Serum was obtained by centrifugation in the serum-gel tube (Sarsted, Sweden). Total serum IgM, IgGs (Bethyl Laboratories), and IgG-subtypes IgG1, IgG2a, and IgG2b were measured using commercially available ELISA kits (Thermo Fisher, Invitrogen) and read the plate using a microplate reader at 450 nm absorbance. The total sialic acids on the IgG were measured with Lectin ELISA, as described previously (15). Briefly, anti-mouse IgG F(ab)2 fragment (Sigma Aldrich, Sweden) was used to coat 96 well plates overnight at 4°C in sodium bicarbonate buffer (0.1M, pH 9.6), then blocked for 1-2 hours using polyvinyl pyrrolidone (0.5%) in warm Tris buffer saline buffer (TBS) (Sigma Aldrich-PVP10). Incubated with serum (1:100 dilution) in TBST (0.1% tween 20) followed by biotinylated Sambucus nigra lectin (SNA) (1:4000 dilution) (Vector Laboratories, Sweden) and then streptavidin-HRP (R&D Systems, Sweden) (1:1000 dilution) were used to detect sialylation followed by substrate addition and read the plate using a microplate reader at 450-620 nm absorbance.

In a subset of ovariectomized mice from the experiment-II where an adequate volume of serum (240 µl) was available. Mouse IgG was purified using a commercially available IgG purification kit (Protein G High-Performance Spintrap™Sigma Aldrich) according to manufacturer protocol. Following purification, the concentration of IgG and protein were determined with a commercially available IgG ELISA kit (Bethyl Laboratories) and DC protein kit (Bio-rad). Purified IgG sample (30 μg protein) were reduced, Cys residues derivatized with S-Methyl methanethiosulfonate and cleaved in-solution with trypsin (1:50) in the presence of 0.5% sodium deoxycholate using standard procedures as described in (24). The samples were desalted using Pierce C18 desalting columns (Thermo Scientific) according to the manufacturer’s instructions.

An Exploris 480 Orbitrap instrument (Thermo Scientific) interfaced with an Easy-nLC-1200 chromatography system was used for the nano-LC-MS/MS analysis. The trapping column was an Acclaim Pepmap 100 C18 (Thermo Scientific) and the analytical column was in-house packed with Reprosil-Pur (75 μm × 350 mm, particle size 3 μm, Dr Maisch). The flow was set to 300 nL/min and the 90 min elution gradient was 5% B in A to 45% B in A over 78 min, then to 100% B over 2 min, and then kept at 100% B for 10 min. A was 0.2% formic acid in water, B was 80% acetonitrile, and 0.2% formic acid in water. Each sample (2 mL) was analyzed three times consecutively with different MS/MS settings, see below.

The precursor Ion MS1 spectra were collected at m/z 600-2200 at a mass resolution of 120 000. The most intense ions were subjected to MS/MS (MS2), using an isolation window of 3 m/z units, with higher-energy collision dissociation (HCD) at a normalized collision energy (NCE) of 20 for the first injection, NCE 30 for the second injection and NCE 38 for the third injection. The MS2 spectra were collected with a resolution of 30 000 with a first mass of m/z 110. The cycle time was 2 sec and an exclusion time of 12 sec was used.

The LC-MS/MS identification of trypsin-digested glycopeptides was done with the Byonic node (Protein Metrics) of Proteome Discoverer (v 2.4, Thermo Scientific) and the precursor ion intensities were extracted with the Minora Feature Detector node. Search settings included: the database was Swiss-Prot mouse (17 155 sequences); cleavage after Arg and Lys; two missed cleavages allowed; mass accuracy set to 10 ppm and 30 ppm at MS1 and MS2, respectively; fixed modification was a methylthio group at Cys; allowed modification was Met oxidation (set to rare 1); The N-glycan database (set to common 1) was selected to contain 123 complex type, oligomannose, and hybrid structures and included ones typically observed for immunoglobulins, and contained 0-3 of either sialic acid Neu5Ac or Neu5Gc glycoforms, 0-1 fucose, and 0-1 bisecting GlcNAc. The glycopeptide identities for each sample were filtered to include at least one identity from the three injections, which had a Byonic score >300. Glycopeptide hits were manually verified to contain the expected peptide+GlcNAc ion (Y1 ion) and if Fuc was suggested the peptide+GlcNAc+Fuc ion (core fucosylation), or an ion at m/z 512 (antenna fucosylation), had to be present. Further, identities containing NeuGc had to include the ions at m/z 290 and m/z 308; and correspondingly, m/z 274 and m/z 292 had to be present for Neu5Ac.

The peak intensity ratio, in percentage, for each glycoform was calculated by dividing the Minora detected precursor ion intensity with the summed intensities of the identified glycoforms that share the same peptide and presented as average values for the three injections.

Total Sialic acid (Neu5Ac + Neu5Gc) was measured in total serum using the Sialic acid assay kit (Sigma-Aldrich, Sweden) according to manufacturer protocol. Total protein was measured using a DC protein assay kit (Bio-Rad, Sweden) according to manufacturer protocol. Protein sialylation was quantified by the ratio between total sialic acid and total protein.

In experiment II, single-cell suspensions in PBS were prepared from bone marrow (BM) and spleen for flow cytometry analyses. BM cells were collected by flushing the BM in PBS followed by erythrolysis in 0.83 NH₄Cl solution (pH 7.4). Splenocytes for flow cytometry were prepared by mashing the spleen tissue using a syringe and collecting the cells through 70 µM cell strainers in PBS, followed by erythrolysis in 0.83 NH₄Cl solution (pH 7.4). Cells were counted using an automated cell counter (Sysmex Europe GmBH, Sweden). Flow cytometry analysis was performed using fluorochrome-conjugated antibodies i.e., APC anti-CD267 (TACI), PerCP-anti-CD19, BV421- anti-CD138, V500-anti-B220 and APCCy7-anti-CD3, all purchased from eBioscience (Sweden). Intracellular FITC-SNA (Vector lab, Sweden) was used to stain sialic acid. Analysis was performed using the BD FACS-verse flow cytometer and Flow Jo software (FlowJo10.6.2).

Total RNA was extracted from BM, and spleen tissue with an RNeasy mini kit (Qiagen) and reverse transcribed into cDNA using a cDNA Reverse Transcription synthesis kit (Thermo Fisher Scientific). cDNA corresponding to 2.5-5 ng RNA was used for quantitative PCR with TaqMan (Thermo Fisher) and analysis for mRNA expression was performed with the Applied Biosystem Step OnePlus Real-Time PCR System using ST6Gal1 (TaqMan Mm00486119_m1, NCBI Gene ID- 20440), B4Galt2 (TaqMan Mm00480752_m1, NCBI Gene ID- 53418), and Fut8 (TaqMan Mm00489795_m1, NCBI Gene ID- 53618) primers and probe set (Thermo Fisher) following standard TaqMan Assay protocol with a cycling condition of an initial cycle for reverse transcriptase at 50^°C for 30 minutes, DNA polymerase activation at 95^°C for 2 minutes, followed by 35-45 cycles of a 15-second denaturation at 95^°C and then 60 seconds annealing and extension at 60^°C. The housekeeping gene 18S (4310893E, Applied Biosystems) was used as an endogenous control in all analyses using the ddCT method and presented as % of the OVX-Pla group.

Statistical analyses were performed using GraphPad Prism software (version 0.1.0 (216)). Groups were compared with a one-way analysis of variance (ANOVA) for experiments- I and -II, followed by Dunnett’s multiple comparisons against the OVX-Pla group. Student´s t-test was used to compare the two groups. Grubb’s test was performed to detect significant outliers which were removed from further analysis. Data are presented as mean ± standard error of the mean (SEM) bar or as scatterplots, and p < 0.05 was considered statistically significant.

Successful ovariectomy was confirmed by a significant decrease in uteri weight compared to the sham group in all three experiments ( Table 1 ). The mice that received E2 treatment had significantly larger uteri than those in the OVX-Pla group ( Table 1 ). As expected, E2 treatment in OVX mice reduced thymus and gonadal fat weight as well as bone marrow cellularity compared to the OVX-Pla group ( Table 1 ). Both short- and long-term estrogen deficiency affected gonadal fat weight and bone marrow cellularity compared to sham-operated mice. However, thymus weight was only significantly affected in the short term. Furthermore, we assessed the impact of E2 treatment effect on trabecular BMD and cortical thickness of the tibial bone. Our results showed that E2 treatment significantly increased both trabecular BMD and cortical thickness in OVX mice ( Supplementary Figures S1A, B ).

To assess the effect of estrogen on serum IgG levels and total sialic acid attached to IgG of both the Fab and Fc domain, we conducted serum analyses with ELISA in all experiments. The total IgG levels were not affected by OVX or E2 treatment in experiments- I and -II ( Figures 1A, C ). However, in long-term estrogen deficiency (after 65 days), OVX mice had significantly decreased serum IgG levels compared to sham-operated mice ( Figure 1E ). The levels of different IgG isotypes (IgG1, IgG2a, and IgG2b) in serum from termination in all experiments showed no significant variation dependent on estrogen status ( Supplementary Figures 2A–C ). However, long-term estrogen deficiency showed a strong tendency to lower the levels of IgG1, IgG2a, and IgG2b in a similar way as total IgG from the long-term healthy post-menopausal status. The same pattern was also displayed with IgM levels, with only long-term estrogen deprived in OVX mice displaying a strong tendency to reduce levels compared to sham mice ( Supplementary Figures S3A–C ). Lectin ELISA was used to assess the degree of total IgG sialylation on both the Fab and Fc domains, but no alteration was found in sialic acid on total IgG in any of the experiments ( Figures 1B, D, F ).

Next, we performed a comprehensive investigation of the higher dose of E2 (injection) treatment in the post-menopausal normal circumstances on IgG-Fc glycosylation using a highly sensitive LC-MS/MS-based glycoproteomic analysis, which detected 19 different glycoforms on the IgG-Fc region ( Supplementary Table S1 ). Our analysis revealed that purified IgG from OVX mice from experiment II contained the highest levels of IgG2b subtype (Uniprot entry P01867, peptide sequence EDYNSTIR) and therefore we continued the investigation on this IgG subtype. OVX-E2-treated mice displayed a sharp decrease in A-galactosylated IgG, i.e., G0 glycoforms compared to the OVX-Pla group ( Figure 2A ). A strong tendency of further glycosylated IgG, carrying one and two galactose residues (G1/G2), was measured, in the OVX-E2 group ( Figure 2B ). Interestingly, the OVX-E2 group showed a significant increase in sialylated IgG-Fc GS1/GS2 glycoforms ( Figure 2C ). However, the core fucosylation of IgG remained unchanged between OVX-Pla and OVX-E2 groups ( Supplementary Table S1 ).

Next, we wanted to see if estrogen treatment affects the sialic acid levels intracellularly in the bone marrow cells and splenocytes. We characterized the total live cells, B- cells, and plasma cells in the bone marrow cells and splenocytes from the experiment-II by performing flow cytometry analysis. We assessed the sialic acid levels by measuring the mean fluorescence intensity (MFI) after staining with sialic acid binding SNA. We found that neither estrogen removal (OVX) nor estrogen treatment affected the sialic acid levels in bone marrow or spleen-derived total live cells, B-cells, or plasma cells ( Figures 3A, B ).

To determine the effect of estrogen on total serum glycoprotein sialylation, we quantified the sialic acid in total serum using ELISA in all three experiments. In healthy postmenopausal conditions, neither E2 treatment nor short- or long-term estrogen deficiency affects total glycoprotein sialylation ( Figures 4A–C ).

To determine the impact of estrogen status on the regulation of general glycosylation we analyzed the mRNA level of glycosyltransferases, sialyltransferase (St6gal1), galactosyltransferase (B4galt2), and fucosyltransferase (Fut8) in spleen tissue, and bone marrow ( Figure 5 ) from the experiment- II. E2 treatment, significantly reduced St6gal1 mRNA expression in the spleen, whereas B4Galt2 and Fut8 had no effect ( Figure 5A ). None of the glycosyltransferases were affected by E2 status in the bone marrow ( Figure 5B ).