Twenty-four healthy crossbred (Landrace × Large White) gilts were purchased from a specialized producer (PIC Deutschland GmbH) and housed in a commercial Austrian piglet-producing farm free of PRRSV, as confirmed by regular serological monitoring. All gilts were vaccinated against porcine parvovirus 1 in combination with Erysipelothrix rhusiopathiae, swine influenza A virus, and porcine circovirus type 2, as previously described (38). Prior to insemination (142 and 114 days prior to infection) and during mid-gestation (31 days prior to infection), twelve randomly selected gilts were vaccinated with a PRRSV MLV vaccine (ReproCyc^® PRRS EU, Boehringer Ingelheim Vetmedica GmbH, Ingelheim am Rhein, Germany) according to the instructions of the manufacturer. Vaccinated and non-vaccinated gilts were housed separately but under identical housing conditions. At day 77/78 of gestation, vaccinated and non-vaccinated gilts were relocated to a biosafety level 2 unit of the University of Veterinary Medicine Vienna on two consecutive days. All gilts were randomly allocated into six groups: 1. non-vaccinated and non-infected, No.Vac_No.Chall; 2. vaccinated and non-infected, Vac_No.Chall; 3. non-vaccinated and infected with low virulent (LV) strain, No.Vac_Chall_LV; 4. vaccinated and infected with low virulent (LV) strain, Vac_Chall_LV; 5. non-vaccinated and infected high virulent (HV) strain, No.Vac_Chall_HV; 6. vaccinated and infected with high virulent (HV) strain, Vac_Chall_HV (n = 4/group). Each group was housed in individual rooms with isolated airspaces. After one-week of acclimation, experimental infection was performed as described previously (39). Eight gilts (4 vaccinated and 4 non-vaccinated) were inoculated intranasally and intramuscularly (50% IN, 50% IM), with an infectious dose of 3 × 10⁵ TCID₅₀, with either one of two different PRRSV-1 field isolates (LV or HV) or sham-inoculated with cell culture medium (DMEM, Thermo Fischer Scientific, Carlsbad, CA, United States) at day 84 of gestation. An overview of the six groups is given in Table 1 . All experiments were approved by institutional ethics and animal welfare committee (Vetmeduni Vienna) and the national authority according to §§26ff. of Animal Experiments Act, Tierversuchsgesetz 2012 – TVG 2012 (GZ 68.205/0142-WF/V/3b/2016).

Two European PRRSV-1 field isolates with a documented history of reproductive pathogenesis, as communicated by veterinarians in the field, were used. The PRRSV-1 field isolate 720789 (Genbank Accession number OP529852, kindly provided by Christoph Keller, Boehringer Ingelheim Vetmedica GmbH), further referred to as the ‘low virulent strain (LV)’, was propagated in MARC-145 cells for seven passages. The PRRSV-1 field isolate AUT15-33 (GenBank Accession number MT000052), further referred to as the ‘high virulent strain (HV)’, was propagated for three passages in porcine alveolar macrophages as described before (9). Titers were determined on the respective cell line (MARC-145, MA-104 derived African Green monkey kidney cell line) or cells (porcine alveolar macrophages, PAMs) used for propagation.

Approximately 21 dpi (21 ± 2, gestation day 105 ± 2), gilts and their litters were anesthetized by intravenous injection of Ketamine (Narketan^® 100 mg/mL, Vetoquinol Österreich GmbH, Vienna Austria, 10 mg/kg body weight) and Azaperone (Stresnil^® 40 mg/mL, Elanco GmbH, Cuxhaven, Germany, 1.5 mg/kg body weight) and subsequently euthanized via intracardial injection of T61^® (Intervet GesmbH, Vienna, Austria, 1 mL/10 kg body weight). To retrieve samples, the abdomen of the gilts was opened, and the uteri removed, placed into a trough, and rinsed with tap water to remove maternal blood. The uteri were incised and opened at the anti-mesometrial side. The position of each fetus, from the left and right uterine horn, was recorded as previously described (38, 40). Fetal preservation status for each individual fetus was assessed and categorized as viable (VIA), meconium-stained (MEC), decomposed (DEC), and autolyzed (AUT) as previously described (39). For investigations on immune cell populations at the maternal-fetal interface, two fetuses per gilt were randomly selected and removed with their umbilical cord, placenta, and a portion of the uterus adjacent to the umbilical stump. A 1 × 1 cm piece of the maternal-fetal interface, was embedded in Tissue-Tek^® O.C.T compound (Sakura Fintek, Alphen aan den Rijn, The Netherlands) and immediately frozen in liquid isopentane whilst placed on dry ice and stored at –80 °C until further processing. The myometrium was trimmed off and the maternal endometrium (ME) and fetal placenta (FP) were mechanically separated with two forceps without contaminating either side. Once separated, 40 g of ME and 60 g of FP were collected in sterile collection cups (Greiner Bio-One, Frickenhausen, Germany) filled with medium (RPMI-1640 with stable L-glutamine supplemented with 100 IU/mL penicillin and 0.1 mg/mL streptomycin (PAN-Biotech, Aidenbach, Germany)). In addition, tissue pieces from the ME and FP for viral load quantification were snap-frozen in liquid nitrogen and stored at –80 °C until further processing.

The procedure for the isolation of immune cells from the porcine maternal-fetal interface has been described previously (37). In brief, ME and FP tissues were cut into small pieces and incubated in tissue digestion medium [RPMI-1640 supplemented with 2% (v/v) heat-inactivated fetal calf serum (FCS; Sigma-Aldrich, Schnelldorf, Germany), 25 U/mL DNase type I (Thermo Fischer Scientific), 300 U/mL Collagenase type I (Thermo Fisher Scientific), 100 IU/mL penicillin (PAN-Biotech), and 0.1 mg/mL streptomycin (PAN-Biotech)] for 1 h at 37 °C and constant mixing. Remaining larger pieces of tissue and dead cells were removed by draining the cell suspensions through a coarse-meshed sieve and subsequent filtering through a layer of cotton wool. Suspensions were centrifuged (350 × g, 10 minutes, 4°C), resuspended in 40% Percoll (13 mL, Thermo Fisher Scientific), underlaid with 70% Percoll (13 mL, Thermo Fisher Scientific), and subjected to density gradient centrifugation (920 × g, 30 minutes, room temperature). Isolated leukocytes were washed four times (phosphate-buffered saline (PBS, 2x), RPMI-1640 + 5% FCS (1x), and RPMI-1640 + 10% FCS (1x)) and immediately used for immune phenotyping.

The extraction of PRRSV RNA from the ME and FP, and quantification of the viral load in these tissues has been described elsewhere (38). Briefly, tissues were homogenized in lysis buffer (QIAzol^® lysis reagent, QIAGEN GmbH, Hilden, Germany) with three stainless steel beads using a TissueLyser II instrument (QIAGEN GmbH). The homogenates were centrifuged, chloroform was added, and the tubes were vigorously vortexed and subsequently spun (13 000 × g, 5 minutes) to ensure phase separation. The aqueous phase was collected, and viral RNA was obtained using the Cador Pathogen Kit (QIAGEN GmbH) in a QiaCubeHT device (QIAGEN GmbH) following the manufacturer’s instructions. An ORF7-specific reverse transcription quantitative polymerase chain reaction (RT-qPCR) for the LV and HV strain, primers and probes listed in Table 2 , was performed using the Luna Onestep RT PCR kit (New England Biolabs GmbH, Frankfurt am Main, Germany). The viral load, expressed as genome equivalents (GE), was determined based on the serial dilution of SP6 transcripts, specific to the PRRSV-1 isolates that were cloned into a pGEM-T vector (pLS69, Promega GmbH, Walldorf, Germany) and amplified. The cloned product was digested with DNaseI (New England Biolabs GmbH) and viral SP6 RNA was purified with the RNeasy kit (QIAGEN GmbH). Hereafter, a Quantus fluorometer and RNA-specific fluorescent dye (Promega) were used to determine the RNA concentration. The RNA concentration was multiplied with Avogadro’s number and divided by the molecular mass of the PRRSV-1 specific SP6 transcripts to determine the absolute quantity of GE.

Mononuclear immune cells (1.5 × 10⁶ cells per isolation) from ME and FP, were transferred into a 96-well round bottom microtiter plate (Greiner Bio-One) and stained in a 5- or 6-step procedure. An overview of the primary monoclonal antibodies (mAbs) and secondary reagents used per panel is given in Table 3 . All incubation steps (20 minutes, 4°C) were followed by two washes with cold PBS supplemented with 10% (v/v) porcine plasma (in-house preparation) or as specified. Surface antigens were stained with mAbs listed in Table 3 followed by incubation with secondary reagents. Free antibody sites of the isotype-specific secondary antibodies were blocked with 2 µg whole mouse IgG (ChromPure, Jackson ImmunoResearch, West Grove, PA, United States) and subsequently washed with PBS. Thereafter, a mixture of directly conjugated primary mAbs, streptavidin conjugates, and the Fixable Viability Dye eFluor 780 (Thermo Fisher Scientific) was applied. The BD Cytofix/Cytoperm kit (BD Biosciences, San Jose, CA, USA) was used to fix and permeabilize the cells. This was followed by a staining for intracellular antigens using directly conjugated mAbs. All samples were measured on a FACSCanto II flow cytometer (BD Biosciences) equipped with three lasers (405, 488, and 633 nm), and a minimum of 1 × 10⁵ lymphocytes per sample were recorded. Single-stained samples were prepared and recorded for automatic calculation of compensation, using FACSDiva software version 6.1.3 (BD Biosciences). The obtained data was analyzed with FlowJo software version 10.8.1 (BD Biosciences) and a consecutive gating strategy was applied ( Supplementary Figure 1 ). A time gate was applied and based on the light scatter properties [forward scatter area (FSC-A) vs. side scatter area (SSC-A)] lymphocytes were identified. A 2-step doublet discrimination was performed and subsequently cells with high auto fluorescent signal were excluded using a 530/30 nm bandpass filter in the excitation line of the violet laser. Dead cells were excluded by a high signal for the Fixable Viability dye eFluor 780.

Tissue from the maternal-fetal interface was sectioned using a Leica CM1950 microtome (Leica Biosystems Nussloch GmbH, Nussloch, Germany). Sections were loaded onto a slide, air-dried at room temperature for 1 h, and fixed with methanol/acetone (1:1) for 30 minutes at -20°C. Slides were blocked with PBS + 5% goat serum (Vector Laboratories, Inc., Burlingame, CA, U.S.A.) for 30 minutes at room temperature. Mouse anti-PRRSV-NP mAb (IgG2a, clone P11/d72-c1, in-house, 1:2) was diluted in PBS and applied overnight (4°C). Thereafter, secondary goat anti-mouse IgG2a AlexaFluor488 (Thermo Fisher Scientific, 1:500) was diluted in PBS and applied for 40 minutes at room temperature. This was followed by a 2 h incubation with a rat anti-human/mouse Cytokeratin 8 mAb (1:500; IgG2a, clone TROMA-1, Merck KGaA, Darmstadt, Germany) and visualized by secondary goat-anti-rat IgG (H+L) AlexaFluor647 (1:500; Thermo Fisher Scientific), for 40 minutes, both at room temperature. After each incubation step, slides were washed three times in PBS for five minutes. Nuclei were stained with DAPI (Sigma Aldrich) for 3 minutes in the dark and the slides were washed twice with PBS. Finally, slides were washed once with dH₂O and covered with mounting medium (Mowiol^®4-88, Polysciences Europe GmbH, Germany) and a cover glass. Tissue sections were scanned using an Axioimager Z.1 microscope (Carl Zeiss Micro imaging GmbH, Germany) equipped with TissueFAXS hardware and software (TissueGnostics GmbH, Austria).

The frequencies of major immune cells lineages (NK, γδ, B, CD4 T, and CD8β T cells), as a measure within viable lymphocytes, were exported into Microsoft Excel (Office 2016, Microsoft, Redmond, WA, United States) and corrected for CD45 expression as previously described (37). Also, frequencies of immune cell subsets and myeloid phenotypes were exported into Microsoft Excel and imported into GraphPad Prism version 9.2.0 (GraphPad Software Inc., San Diego, CA, United States) for the graphical presentation highlighting animal-to-animal variation. Statistical analysis was performed with R version R v4.0.2 (41).

The viral load in tissues from the maternal-fetal interface was determined using RT-qPCR for PRRSV ORF7, with primers specific for the LV or the HV strain. Since no viral RNA for any strain could be detected in the ME and FP from gilts in the No.Vac_No.Chall and the Vac_No.Chall group (data not shown) only the challenged groups (No.Vac_Chall_LV, Vac_Chall_LV, No.Vac_Chall_HV, and Vac_Chall_HV) are displayed in Figure 1 . Our analysis revealed that the emmeans for the viral load were significantly higher for the No.Vac_Chall_HV group as compared to the No.Vac_Chall_LV, Vac_Chall_LV, and Vac_Chall_HV group within ME and FP, respectively ( Figures 1A, B ). Of note, at the maternal-fetal interface from fetuses originating from Vac_Chall_LV gilts no viral RNA could be detected ( Supplementary Figure 2 ). For Vac_Chall_HV gilts, viral RNA could be detected in the ME from a few fetuses from two different litters (gilts 15 and 16) and for gilt 15 in affected fetuses the virus was transmitted to the FP ( Supplementary Figure 2 ), highlighting vertical transmission. Furthermore, there was a substantial negative impact on the fetal preservation status in the No.Vac_Chall_HV gilts ( Supplementary Figure 2 ). Only 56% of these fetuses were designated as viable whereas in the other groups the vast majority (>90%) of fetuses were viable (data for the No.Vac_No.Chall and Vac_No.Chall group not shown). A clear difference in impact on the fetal preservation status between the two PRRSV-1 field isolates (LV and HV) was observed and demonstrated a divergence in virulence.

Since PRRSV infects cells of the myeloid lineage, we sought to investigate their phenotype at the maternal-fetal interface. Following FCM staining, myeloid cells at the maternal-fetal interface were identified based on their CD172a expression and were subsequently divided into CD14^pos and CD14^neg subsets ( Figures 2A, B ). No significant differences were observed for the CD14^pos cells within total CD172a^pos cells in the ME whereas a significant decrease was observed in the FP of No.Vac_Chall_HV group as compared to No.Vac_No.Chall group ( Figure 2C , top panel). In addition, a high degree of variation between individual fetuses, especially within the ME, was identified ( Figure 2C , bottom panel, scatterplots). Both CD14-defined subsets were further analyzed for their co-expression of CD163 and CD169, both molecules involved in viral entry, and CD163^highCD169^pos mononuclear phagocytes (MPCs) were identified ( Figures 2A, B ). The abundance of these CD163^highCD169^pos MPC phenotypes in the respective CD14-defined populations in ME was rather low ( Figures 2A, D left) and no differences in emmeans for either macrophage phenotypes was observed between the six groups ( Figure 2D left, top panel). A high abundance of CD163^highCD169^pos phenotypes within CD14^pos and CD14^neg CD172a^pos cells was found in the FP, especially in the No.Vac_No.Chall and the Vac_No.Chall groups ( Figures 2B, D right, bottom panel). A significant drop for CD163^highCD169^pos cells within CD14^pos CD172a^pos MPCs was seen in the FP of fetuses from the No.Vac_Chall_HV as compared to the No.Vac_No.Chall, Vac_No.Chall, and No.Vac_Chall_LV groups ( Figures 2B, D right, top panel). For CD163^highCD169^pos cells within CD14^neg MPCs in the FP a similar drop was observed for the No.Vac_Chall_HV as compared to the No.Vac_No.Chall group ( Figures 2B, D right, top panel). These significant contrasts for both MPC subsets in the FP ( Figure 2D right) prompted us to investigate a correlation with viral load. For this purpose, a spearman correlation was performed for all challenged groups and both anatomic locations ( Figure 2E ). A strong negative correlation (R = –0.76, p = 0.03) between both MPC phenotypes and the viral load was revealed in the FP from No.Vac_Chall_HV fetuses. Furthermore, virus infected cells, as identified with a monoclonal antibody targeting PRRSV-NP, were predominantly detected in the FP ( Supplementary Figure 3 ).

Next to MPCs, major lymphocyte subsets were investigated by flow cytometry and the applied gating strategy is illustrated in Supplementary Figure 1 . A CD3^negCD8α^posCD16^posCD172a^neg phenotype was used to identify NK cells. During steady state conditions (No.Vac_No.Chall), total NK cells were present in similar frequencies within total lymphocytes in both the ME and FP ( Figure 3 , Scatter plots). With regards to PRRSV-mediated changes, no significant contrasts were detected in the ME, possibly due to the high degree of animal-to-animal variation. Significant higher emmeans for total NK cells could be observed in the FP from No.Vac_Chall_HV fetuses. Significant contrasts for the FP were found between No.Vac_Chall_HV vs No.Vac_No.Chall, No.Vac_Chall_HV vs Vac_No.Chall, and No.Vac_Chall_HV vs Vac_Chall_HV ( Figure 3B ). Porcine γδ T cells at the maternal-fetal interface were identified with a monoclonal antibody targeting a T-cell receptor γδ-specific CD3ϵ chain (clone PPT16) (54). Emmeans for the total γδ T cells in the ME were lower for both non-vaccinated challenged groups (No.Vac_Chall_LV and No.Vac_Chall_HV) as compared to the No.Vac_No.Chall group. Furthermore, the vaccination seemed to have prevented this loss in total γδ T cells in the Vac_Chall_HV group. These differences in emmeans were also visible in the scatterplots showing the percentages of γδ T cells within lymphocytes ( Figure 3A ). Total γδ T cells were significantly reduced in the FP of fetuses from the No.Vac_Chall_HV group as compared to No.Vac_No.Chall, Vac_No.Chall, and vaccinated counterpart (Vac_Chall_HV) ( Figure 3B ). Total B cells at the maternal-fetal interface were identified using the pan-B cell marker CD79α. For this phenotype, no PRRSV-associated changes were observed neither in the ME nor in the FP ( Figure 3 ). CD4 and CD8 T cells, were characterized by gating on total CD4 and total CD8β expressing T cells ( Supplementary Figure 1 ). No significant PRRSV-induced contrasts for both T cell phenotypes in the ME could be identified by our statistical model. However, for the total CD8β T cells a high degree of animal-to-animal variation was observed for most groups except for the Vac_Chall_HV group. A significant reduction in total CD4 T cells could be observed in the FP from the No.Vac_Chall_HV group as compared to the No.Vac_No.Chall and Vac_No.Chall groups. For CD8β T cells, we only observed a significant increase in the No.Vac_Chall_LV group in comparison to the No.Vac_No.Chall group ( Figure 3B ).

Total CD3^negCD8α^posCD16^posCD172a^neg NK cells in the FP were further investigated for their expression of NKp46. Three NK cell subsets, NKp46^neg, NKp46^pos, and NKp46^high were identified. Representative pseudocolor plots for one fetus from the No.Vac_No.Chall and No.Vac_Chall_HV group are shown in Figure 4A . Considering that the relative frequencies of the three NKp46-defined NK cell subsets are interdependent, a univariate CoDa was performed. Therefore, to correct for this interdependence our data was transformed to centered log ratios (clr). The output of our model, showed a significant increase in NK cells with a NKp46^pos phenotype in the No.Vac_Chall_HV as compared to the Vac_No.Chall group ( Figure 4B ). For the other two NKp46-defined NK cell phenotypes, no significant changes were observed. When considering the raw frequency data, however, a visual reduction in the NKp46^neg NK cells in the FP from the No.Vac_Chall_HV group could be observed. Notably, considerable variation between individual fetuses was observed. Data on NKp46-defined NK cell phenotypes in the ME are not shown, since no significant changes were observed.

Total γδ T cells were analyzed for their expression of CD8α which enabled us to identify a CD8α^neg/dim and CD8α^high expressing subset in the ME and FP. Representative pseudocolor plots for the two investigated anatomic locations are shown in Figures 5A, B . CD8α^high expressing γδ T cells were the main phenotype in the ME ( Figure 5A ) whereas the CD8α^neg/dim expressing γδ T cells were more abundant in the FP ( Figure 5B ). For the statistical analysis, only γδ T cells with a CD8α^neg/dim phenotype were included since the effect size of the CD8α^high γδ T cells is dependent on the CD8α^neg/dim phenotype. In the ME no significant difference was found for the CD8α^neg/dim phenotype ( Figure 5A ). Nonetheless, a significant reduction for this phenotype and thus an increase in CD8α^high γδ T was observed in the FP of fetuses from the No.Vac_Chall_HV group as compared to the No.Vac_No.Chall, Vac_No.Chall, and No.Vac_Chall_LV group ( Figure 5B ).

Total CD4 T cells at the maternal-fetal interface were investigated for their expression of CD8α and CD27 ( Figure 6 ). This enabled us to delineate three subsets with a CD8α^negCD27^pos naive, CD8α^posCD27^pos early effector or central memory (Tcm), and CD8α^posCD27^neg late effector or effector memory phenotype (Tem) ( Figures 6A, B ), representative pseudocolor plots are shown). Since the three CD8α/CD27-defined CD4 T cell subsets are interdependent on each other, the components of the compositions were clr transformed before hypothesis testing to deal with the constant sum constraints. No significant changes in the CD8α/CD27-defined CD4 T cell subsets were observed in the ME ( Figure 6A ). In the FP, however, a significant decrease in CD8α^negCD27^pos naive CD4 T cells and a concurrent increase in CD8α^posCD27^pos Tcm cells was observed in the No.Vac_Chall_HV group as compared to the No.Vac_No.Chall, Vac_No.Chall, and No.Vac_Chall_LV group ( Figure 6B ). Of note, the five FP tissues with the highest CD8α^posCD27^pos percentages tested PRRSV positive in this tissue (fetuses G22 L7, G22 R10, G23 L5, G23 R11, and G24 L2, Supplementary Figure 2 ) and showed a reduced number of CD163^highCD169^pos MPCs ( Figure 2D ). Furthermore, the significant loss in CD8α^negCD27^pos naive CD4 T cells was also observed as compared to the Vac_Chall_HV. However, in this case the increase in CD8α^posCD27^pos Tcm cells in the No.Vac_Chall_HV compared to the Vac_Chall_HV was not significant.

As CD8 T cells are major effector cells in many viral infections, we sought to investigate their phenotype at the maternal-fetal interface. Therefore, the expression of CD8α and CD27 on the identified CD8β T cells was evaluated. CD8β T cells with a CD8α^posCD27^pos, CD8α^posCD27^dim, and CD8α^posCD27^neg phenotype were identified ( Figures 7A, B , representative pseudocolor plots are shown) and represent CD8β T cells with a naive, early effector, and late effector phenotype, respectively (55, 56). The interdependency between the three CD8β T cell phenotypes was corrected for with CoDa. Several significant contrasts were identified in both investigated anatomic compartments. In the ME a significant loss of CD8β T cells with a CD8α^posCD27^pos naive phenotype and an accompanying increase of CD8α^posCD27^dim early effector phenotype was observed from No.Vac_Chall_HV fetuses as compared to the No.Vac_No.Chall and Vac_No.Chall group ( Figure 7A ). A similar increase of CD8α^posCD27^dim early effector CD8β T cells was observed for the ME from No.Vac_Chall_LV group as compared to the No.Vac_No.Chall and Vac_No.Chall groups ( Figure 7A ). In addition, significant contrasts for CD8β T cells with a CD8α^posCD27^dim early effector phenotype were observed between the non-vaccinated challenged groups, No.Vac_Chall_HV and No.Vac_Chall_LV, and their vaccinated counterparts, Vac_Chall_HV and Vac_Chall_LV, respectively ( Figure 7A ). Furthermore, a significant but limited increase in CD8β T cells with a CD8α^posCD27^dim early effector phenotype was observed in the Vac_Chall_HV and Vac_Chall_LV groups as compared to the No.Vac_No.Chall and Vac_No.Chall groups ( Figure 7A ). In the FP, a significant loss of CD8α^posCD27^pos naive CD8β T cells in the No.Vac_Chall_HV group as compared to the Vac_No.Chall group concurred with a strong increase in CD8β T cells with a CD8α^posCD27^dim early effector phenotype ( Figure 7B ). Also for No.Vac_Chall_LV group a significant increase of CD8β T cells with a CD8α^posCD27^dim early effector phenotype was observed ( Figure 7B ). Similarly to the ME, significant contrasts in the FP were observed between the non-vaccinated challenged groups, No.Vac_Chall_HV and No.Vac_Chall_LV, and their vaccinated counterparts, Vac_Chall_HV and Vac_Chall_LV, respectively ( Figure 7B ). Compared to the other investigated lymphocyte subsets, CD8α^posCD27^dim early effector CD8β T cells showed the strongest response to infection with the two PRRSV-1 strains.