All animal studies were approved by the local government of Lower Franconia, Germany (approval numbers 55.2-2532-2-836 and 55.2-2532-2-1129) and performed in strict accordance with the guidelines for animal care and experimentation of the German Animal Protection Law and the DIRECTIVE 2010/63/EU of the EU. The animals were housed in IVC cages under standardized lighting conditions and had ad libitum access to food and water. All experimentations with anonymized and non-traceable human cord blood was approved by the ethics committee of the University Wuerzburg (approval number 20191212 02).

We included female and male NSG (NOD.Cg-Prkdc^scid Il2rg^tm1Wjl /SzJ) and NSG-SGM3 (NOD.Cg-Prkdc^scid Il2rg^tm1Wjl Tg(CMV-IL3,CSF2,KITLG)1Eav/MloySzJ) mice from the Jackson Laboratories (Bar Harbor, ME, USA) and female Balb/c mice (BALB/cJRj, Janvier labs, Le Genest-Saint-Isle, France) in all experiments. The human CD34+ hematopoietic stem cells were isolated from human cord blood by magnetic separation (EasySep™ Human Cord Blood CD34 Positive Selection Kit II, STEMCELL technologies, Cologne, Germany) following the manufacturer’s protocol. The quality of the cell preparation was controlled by staining for hCD34+ and hCD3+ cell markers by flow cytometry. Only preparations with > 85% hCD34+ purity and < 1% hCD3+ cell content were used for humanization purposes. Humanized NSG (huNSG) and NSG-SMG3 (huSGM3) mice were generated by engrafting 100,000 hCD34+ cells (of a donor mix) at the age of 6 – 8 weeks, similar to the procedures described in earlier publications (15–17, 27). Briefly, after whole-body irradiation with a sub-lethal dose of 2 Gy, mice were injected intravenously with the human hematopoietic stem cells. Murinized NSG-SGM3 mice (muSGM3) received a similar treatment than huNSG and huSGM3 mice, but 100,000 bone marrow cells from a Balb/c donor instead of human CD34+ cells were injected. The peripheral blood of humanized mice was analyzed every two weeks after engraftment for the presence and frequency of murine CD45+, as well as human CD45+, CD66b+, CD3+ and CD20+ cells by flow cytometry.

Every two weeks during the course of the humanization and at the end of the bacterial thigh muscle infection, peripheral blood was collected and mixed with EDTA as anticoagulant (pluriSelect Life Science, Leipzig, Germany). The hemoglobin content (Hämoglobin, Diaglobal GmbH, Berlin, Germany) and the red blood cell count (Erythrozyten Gower’s Reagenz, Bioanalytic GmbH, Umkirch/Freiburg, Germany) was measured by fluorometric assays following manufacturer’s instructions. During the humanization period, we furthermore determined the reticulocyte numbers by streaking blood samples on glass slides and counted the number of reticulocytes among red blood cells after staining (Brillant cresyl blue, Bioanalytic GmbH, Umkirch/Freiburg, Germany).

HuSGM3 mice were infected intramuscularly (i.m.) at 12 weeks and the huNSG mice at 18 weeks post hCD34+ stem cell injection as described previously (16). At these points human CD45+ cell numbers reached robust numbers in blood and the number of hCD3+ T cells and hCD19+ B cells were at comparable levels. Balb/c, wild-type NSG-SGM3 and murinized NSG-SGM3 mice were infected at the age of 18 weeks in order to match the age of the humanized NSG-SGM3 mice (engraftment at the age of approximately 6 weeks and 12 weeks of humanization). Briefly, Methicillin-resistant S. aureus (MRSA) LAC* lux (28) was pelleted after overnight shaking at 37°C in B medium and resuspended in 0.9% NaCl solution. The left thigh of each mouse was then shaved, disinfected and injected with 1 x 10⁸ CFU bacteria in a total volume of 50 µL (16, 29). Besides huSGM3 and huNSG mice, we used age-matched murinized NSG-SGM3, wild-type NSG-SGM3 and Balb/c mice as controls. The wellbeing of each mouse was inspected and scored every 12 hours p.i. and the weight measured every 24 hours. Those mice that did not reach the humane end point as defined by the score sheet were either sacrificed on day 2 or on day 7 p.i. Peripheral blood, the infected thigh muscle, kidneys, liver, spleen, heart, lung and bone marrow from tibia and femur were then harvested. The spleens were halved and one part homogenized by pressing through a cell strainer for flow cytometry and bacterial burden determination, while the second half was processed for histological examination. The thigh muscle, kidneys, liver, lung and heart were homogenized in 0.9% NaCl and serial dilutions were plated on B agar plates in order to determine the bacterial burden. Bone marrow was harvested by flushing both femurs and tibias with sterile 0.9% NaCl solution, followed by filtration through a cell strainer.

Histological sections and immunohistochemical stainings of splenic tissue were performed using formalin-fixed and paraffin-embedded (FFPE) tissue slides according to standard protocols. Briefly, spleens were fixed overnight in 10% neutral-buffered formalin solution, embedded in paraffin and cut to 5 µm slices. After deparaffinization, samples were stained with H&E and immunohistochemistry staining was performed with anti-human CD45 primary antibody (Agilent Dako, Waldbronn, Germany) at the Institute of Pathology at the University Clinics Wuerzburg according to the appropriate protocols within an automated immunostainer (Benchmark Ultra; Ventana/Roche, Tucson AZ, USA). Specimens were then inspected by an experienced pathologist (FK) for the presence and organization of human immune cells in the spleens.

The peripheral blood of humanized mice was examined by flow cytometry every two weeks after stem cell administration. Blood samples were therefore stained with anti-human CD45/CD3/CD19/CD66b and anti-mouse CD45. The rate of humanization was calculated as: hCD45+ cells/(hCD45+ cells & mCD45+ cells), as applied previously (16). The peripheral blood, the spleen homogenate and bone marrow were interrogated for the presence of immune cells on day 2 p.i. with combinations of anti-human CD45/CD3/CD19/CD14/CD66b and anti-mouse CD45/Ly6C/Ly6G antibodies after red blood cell lysis. All antibodies were supplied by Miltenyi Biotec (Bergisch Gladbach, Germany). Flow cytometric measurements were performed on a MACSQuant flow cytometer and analyzed with MACSQuantify software 2.6 (Miltenyi Biotec, Bergisch Gladbach, Germany).

In order to determine cytokine levels in the thigh muscle, homogenate was centrifuged at 3,000 x g for 5 minutes and the supernatant stored at -80°C until further processing. The peripheral blood samples were agglutinated overnight at 4°C, then centrifuged at 15,000 x g for 15 minutes. The serum was harvested and stored at -80°C until cytokine measurement. Levels of human or murine CCL-2, IL-1β, IL-6, IL-10, IL-17A, and TNF-α as well as human IL-8 were measured in the infected thigh muscles or in peripheral blood with a custom-mixed Luminex assays from Bio-Techne (Wiesbaden, Germany) following the manufacturer’s manual. In order to test the specificity of the individual assay, we measured the cytokine standards of the human kit with the mouse kit and vice versa and could not detect values above background levels.

All statistical analyses were performed with GraphPad Prism (9.1.2) and p < 0.05 was considered as significant. The applied statistical tests can be found in the respective figure caption.

We generated humanized NSG (huNSG) and NSG-SGM3 (huSGM3) mice by sublethally irradiating the animals and intravenously injecting human cord blood-derived CD34+ hematopoietic stem cells. The mice were then continuously monitored and weighed over a period of 18 weeks. A divergence between the two groups in terms of body weight change ( Figure 1A ) became obvious. While huNSG mice continually gained weight, huSGM3 mice initially gained weight, too, but then started to stagnate from week 12 to 18.

Overall, huNSG and huSGM3 mice showed similar developments and levels of humanization rates and individual immune cell populations than described earlier (16, 30). HuSGM3 animals had throughout the whole course of humanization a higher rate of humanization and stronger increase in human immune cells ( Figures 1B–F ). In particular, the amount of hCD66b+ cells was at all times strongly enhanced compared to huNSG mice.

Since symptoms of secondary hemophagocytic lymphohistiocytosis (HLH) and/or macrophage activation syndrome (MAS) were described for NSG-SGM3 mice after engraftment with human CD34+ stem cells (30–32), we measured the numbers of erythrocytes, reticulocytes and the level of hemoglobin in the blood of the humanized mice every two weeks post engraftment ( Figures 1H–J ). While all three parameters remained at a comparable level for the huNSG mice, we saw a drop in erythrocyte numbers and hemoglobin levels as well as an increase in reticulocyte numbers in huSGM3 mice during prolonged course of humanization. The development and extend of these changes fits to earlier description and started at 8 to 14 weeks post engraftment (30–32). Of note, this was accompanied by a strong acceleration of hCD3+ T cell numbers in blood. The reticulocytosis implies that this is caused by a decimation or increased usage of erythrocytes, not a dysfunctional production (30, 31).

Based on the results from the humanization phase, we decided to infect huNSG mice at 18 weeks and huSGM3 at 12 weeks after engraftment. HuSGM3 mice showed at this point an overall enhanced level of human immune cells compared to huNSG mice, with slightly reduced levels of hemoglobin and erythrocytes and increased numbers of reticulocytes. Both humanized mouse groups were infected locally with 1 x 10⁸ CFU S. aureus LAC* lux in the left thigh muscle, with wild type NSG-SGM3, murinized NSG-SGM3 (which were treated equally to humanized NSG-SGM3, but received murine bone marrow cells instead of human CD34+ stem cells) and wild type Balb/c mice as controls. This type of infection causes the formation of large deep tissue abscesses in wild-type mice as described earlier (29). Since S. aureus utilizes various mechanisms and virulence factors to acquire iron from hemoglobin (33, 34), we decided to track the levels of erythrocytes and hemoglobin in the humanized mouse groups during the course of infection. The overall amounts were, for both factors, lower in huSGM3 mice than in huNSG mice at the start of infection ( Figures 2A, B ). But the level of erythrocytes in the blood of huSGM3 mice increased in the first 48 hours of infection, reaching the levels of huNSG.

Mice were weighed and inspected every 12 hours throughout the infection experiment and received a score based on weight loss and signs of disease. The humanized mouse groups proved to be much more susceptible to the deep tissue infection with S. aureus than the control groups ( Figure 2C ), which fitted to earlier results in this model (16). Nonetheless surprising was the extreme vulnerability of next-generation humanized (huSGM3) mice which showed early in the experiment very strong signs of disease and weight loss, causing a strongly decreased survival rate compared to all other groups, even significantly inferior to huNSG.

The strongly reduced survival rate of huSGM3 mice raised the question of the cause for the enhanced vulnerability, particularly in comparison to huNSG mice. In a first attempt to define the systemic consequences of a local infection with S. aureus in the thigh muscle, we analyzed the bacterial burden in various organs on day 2 p.i. Later dates were not accessible since the number of surviving huSGM3 mice was not sufficient for analysis. No difference in terms of bacterial burden could be seen at the primary site of infection ( Figure 3A ). The analysis of bacterial burden in kidneys, liver, lung, spleen and heart revealed, that humanized mice showed a tendency towards higher numbers of S. aureus in the respective organs, although the differences were rather small and the results varied ( Figure 3B ). The strongest difference could be measured in the lungs with humanized mice displaying significantly higher bacterial numbers than Balb/c mice ( Figure 3C ). Balb/c mice showed the weakest bacterial spreading and infection of inner organs, while it was similar for all other groups ( Figure 3D ). Overall, a clear correlation of bacterial burden on day 2 p.i. and survival during the course of infection could not be established.

Since the number of bacteria during infection did not coincide with the severity of infection, we next analyzed immune cell types and effector molecules in order to identify differences which might explain the high vulnerability of huSGM3 mice. First, we compared major murine immune cell populations in the blood, spleen and bone marrow of all mice ( Figure S1 ). Both humanized mouse groups had similar numbers of murine CD45+ cells, granulocytes and monocytes, but significantly less than wild type or murinized mouse groups ( Figure S2 ). Immunophenotyping of human immune cells in the blood of huSGM3 and huNSG mice at the start of the infection and on d2 p.i. showed overall higher numbers of all investigated cell types in huSGM3 mice, namely T cells, B cells, granulocytes and monocytes ( Figure 4A ). The numbers remained stable in huNSG within the first 2 days of infection, but a clearly distinguishable pattern became visible for the huSGM3 mice. While the number of hCD66b+ granulocytes increased between d0 and d2 p.i., we could see significant drops of B and T cell numbers in the blood. The pattern of higher human immune cell numbers in huSGM3 mice was interestingly mirrored in the spleen ( Figure 4C ) but not the bone marrow ( Figure 4B ) on d2 p.i. The number of human cells in general, as well as of B cells, monocytes and granulocytes in particular, were similar in the bone marrow of huSGM3 and huNSG mice. In contrast to this, we could measure five-fold more hCD45+ cells in the spleens of huSGM3 mice than in huNSG mice, with all investigated human immune cell types represented significantly stronger. Histological examination of the spleens from huNSG and huSGM3 mice showed similar hCD45+ cell patterns with strong accumulation in lymphoid follicles at the periarteriolar lymphoid sheath and associated lymph follicles as well as in the parenchyma ( Figure 5 ).

Next, the levels of murine and human cytokines/chemokines CCL2, IL-1β, IL-6, IL-10 and TNFα, as well as of human IL-8 were determined in the infected thigh muscle and in the blood on d2 p.i., since we hypothesized that the higher immune cell numbers in huSGM3 might be accompanied by higher levels of effector molecules and an elevated inflammatory state. The levels of murine cytokines were similar for all groups in both compartments, indicating that the infection with S. aureus did not lead to different activation of the murine immune cells ( Figure S3 ). The examination of human cytokine levels revealed that while the levels at the primary site of infection were similar for huSGM3 and huNSG mice, strongly increased levels of CCL2, IL-6, IL-8 and TNF-α could be measured by the Luminex assay in the blood ( Figure 6 ). This indicates a systemic response of the human immune system in huSGM3 mice against the S. aureus infection and might furthermore reflect the response to systemically spreading bacteria.

The decreased survival of huSGM3 mice following local S. aureus infection in the thigh muscle compared to huNSG mice was accompanied by higher human immune cell numbers in blood and spleen, as well as with increased levels of immune effector molecules in the blood, but not with increased bacterial burden. Thus, we hypothesized, that the vulnerability does not originate from the pathogen itself but rather from the human immune system. This leads in consequence to the assumption, that a higher rate of humanization might cause higher susceptibility against S. aureus. Comparing the rate of humanization (prior infection) to the time at which each individual mouse reached the humane end point (or survived until the end of the experiment on day 7 p.i.), showed a clear negative correlation between both parameters ( Figure 7 ). In consequence, high numbers of human immune cells prove to be detrimental for the mice during bacterial infection with S. aureus.