HIF‐1α targeted deletion in myeloid cells decreases MDSC accumulation and alters microbiome in neonatal mice

The newborn's immune system is faced with the challenge of having to learn quickly to fight off infectious agents, but tolerating the colonization of the body surfaces with commensals without reacting with an excessive inflammatory response. Myeloid‐derived suppressor cells (MDSC) are innate immune cells with suppressive activity on other immune cells that regulate fetal‐maternal tolerance during pregnancy and control intestinal inflammation in neonates. Until now, nothing is known about the role of MDSC in microbiome establishment. One of the transcription factors regulating MDSC homeostasis is the hypoxia‐inducible factor 1α (HIF‐1α). We investigated the impact of HIF‐1α on MDSC accumulation and microbiome establishment during the neonatal period in a mouse model with targeted deletion of HIF‐1α in myeloid cells (Hif1a loxP/loxPLysMCre+). We show that in contrast to wildtype mice, where an extensive expansion of MDSC was observed, MDSC expansion in neonatal Hif1a loxP/loxPLysMCre+ mice was dramatically reduced both systemically and locally in the intestine. This was accompanied by an altered microbiome composition and intestinal T‐cell homeostasis. Our results point toward a role of MDSC in inflammation regulation in the context of microbiome establishment and thus reveal a new aspect of the biological role of MDSC during the neonatal period.


Introduction
Infections are one of the main causes of perinatal morbidity and mortality [1]. Newborns and especially preterm infants are many times more susceptible to infection than older children and adults. This increased susceptibility to infection is due to the fact that the neonatal immune system reacts differently to infectious agents compared to the adult immune system and only matures to the functional state of the adult immune system during the first postnatal months [2][3][4].
Most neonatal infections involve mucosal surfaces. For both, neonatal sepsis and necrotizing enterocolitis (NEC), which are the most important infections in preterm infants, it has been shown that alterations in the intestinal microbiome precede disease onset and that causative pathogens often descend from the intestinal flora of the infected infant itself [5][6][7][8][9], highlighting specific defects in mucosal immunity.
During pregnancy, the fetus develops in a largely sterile environment, protected from pathogens through maternal immunity. Maternal and fetal tolerance mechanisms prevent the rejection of the semi-allogeneic fetus by maternal immune cells [10]. After birth, the newborn becomes progressively colonized with microbes, directly exposing the neonatal immune system to potential pathogens and concomitantly inducing mucosal immune system development [11,12]. During this transitional period, newborns are highly susceptive to infections. It is supposed that the same mechanisms causing vulnerability to infections are otherwise needed to allow undisturbed microbiome establishment [13].
Myeloid-derived suppressor cells (MDSC) are myeloid cells with suppressive activity on other immune cells that accumulate during pregnancy in the maternal and fetal organism and contribute to maternal-fetal tolerance [14][15][16][17][18]. In neonates, MDSC influence T-cell and monocyte functions [19,20] and appear to play a role in inflammation control during the neonatal period [21]. MDSC can be sub-grouped in two populations-monocytic and granulocytic MDSC (MO-MDSC, GR-MDSC)-depending on their phenotype, with GR-MDSC probably playing the predominant role during the neonatal period.
The transcription factor hypoxia-inducible factor 1α (HIF-1α) plays an important role in MDSC accumulation and activation [14,22,23]. Our group has shown that a lack of HIF-1α in myeloid cells (Hif1a loxP/loxP LysM Cre+) leads to a decreased MDSC accumulation during pregnancy and affects suppressive activity of MDSC [24]. We now aimed to investigate how Hif1a loxP/loxP LysM Cre+ influences MDSC homeostasis in newborn mice. We show that (1) in spleens of newborn wildtype (WT) mice MDSC and especially GR-MDSC expand massively on the first postnatal day, then rapidly decrease until day seven; (2) MDSC and GR-MDSCexpansion is strongly reduced in spleens of neonatal Hif1a loxP/loxP LysM Cre+ mice; and (3) the reduced levels of GR-MDSC in Hif1a loxP/loxP LysM Cre+ mice also affect the intestinal mucosa and are accompanied by an altered intestinal microbiome composition with reduced diversity and altered balance as well as an altered intestinal T-cell homeostasis.

GR-MDSC accumulate in neonatal mice
First, we quantified MDSC levels in splenocytes from adult WT mice in comparison to newborn mice on P1 and P7. We found that spleen leucocytes of newborn mice at P1 contained significantly higher percentages of MDSC than those of adult animals while levels at P7 did not differ significantly from those of adult animals (median 76.0% for P1, 21.5% for P7 and 5.5% for adult mice, n = 4-5, p < 0.01 for P1 compared to adult mice) (Figure 1A and B). Analysis of MDSC-subpopulations revealed that MDSC-accumulation in newborn mice mainly concerned the subpopulation of GR-MDSC (percentages of GR-MDSC in spleen leucocytes median 59.1% for P1, 8.5% for P7 and 3.5% for adult mice, n = 5-6, p < 0.01 for P1 compared to adult mice) (Figure 1C and D). MO-MDSC percentages were also elevated in newborn mice but to a significantly lesser extent than GR-MDSC (percentages of MO-MDSC in spleen leucocytes median 12.1% for P1, median 11.6% for P7 and median 2.3% for adult mice, n = 4-5, p < 0.05 for P1 and P7 in comparison to adult mice) ( Figure 1D and E). In comparison to adult animals, neonatal mice exhibited significantly fewer B-cells and T-cells ( Figure S1). To confirm that the cells we defined as MDSC were indeed ctive suppressively, we analyzed their capacity to inhibit CD4 + T-cell proliferation in a proliferation assay. We found that GR-MDSC isolated from splenocytes on P1 inhibited CD4 + T-cell proliferation in a concentrationdependent manner to 47.5 ± 7.6% (ratio GR-MDSC:T-cells 1:2, n = 6, p < 0.01), 51.6 ± 6.8% (ratio GR-MDSC:T-cells 1:4, n = 6, p < 0.01) and 65.2 ± 13.4% (ratio GR-MDSC:T-cells 1:8, = 6, p = 0.2) in comparison to the proliferation of CD4 + T-cells without addition of GR-MDSC ( Figure 1F+G).

Deletion of HIF-1α in myeloid cells leads to decreased accumulation of GR-MDSC in neonatal mice
Since we have recently shown that the accumulation of GR-MDSC during pregnancy was impaired in mice with targeted deletion of HIF-1α in myeloid cells, we now asked whether this was also true for newborn mice. We found that on P1, Hif1a loxP/loxP LysM Cre+ animals had significantly lower percentages of MDSC and GR-MDSC in spleen leucocytes than WT animals (median 33.0% versus median 71.0%, n = 9-11, p < 0.01 for MDSC and median 26.6% versus median 57.5%, n = 9-11, p < 0.05 for GR-MDSC), while levels had aligned by P7 (median 12.5% versus 10.3%, n = 11, p = 1.0 for MDSC and median 14.9% versus 8.9%, n = 11, p = 0.4 for GR-MDSC) (Figure 2A-C). There were no differences in MO-MDSC expression between spleens of newborn WT vs. Hif1a loxP/loxP LysM Cre+ mice. Considering the other immune cell populations in spleens, there were increased levels of B-and T-cells and decreased levels of monocytes in Hif1a loxP/loxP LysM Cre+ animals in comparison to WT animals at P1 while immune cell composition was similar in WT and Hif1a loxP/loxP LysM Cre+ mice at P7 ( Figure 2E and Supplementary Figure 2A-C). Having shown in a previous work that MDSC from Hif1a loxP/loxP LysM Cre+mice have reduced suppressive activity compared with wildtype MDSC, we now examined the expression of the effector molecules arginase 1, inducible NO synthase (iNOS), and indolamin-2,3-dioxygenase (IDO) in in vitro generated MDSC from Hif1a loxP/loxP LysM Cre+ and wildtype animals. Here, we found a slight but not statistically significant reduction in the expression of all three effector enzymes in MDSC from Hif1a loxP/loxP LysM Cre+ ( Figure S2D-F).

Deletion of HIF-1α in myeloid cells leads to altered intestinal microbiome composition in neonates
As one of the biggest challenges to the neonatal immune system is tolerance to the establishing microbiome, we next investigated whether the decreased accumulation of GR-MDSC in newborn Hif1a loxP/loxP LysM Cre+ mice can also be detected in the intestinal mucosa and if it is accompanied by an altered microbiome composition. For this purpose, adult female and male Hif1a loxP/loxP LysM Cre+ and WT mice were kept together for at least 6 to 8 weeks before mating to achieve microbiome alignment in parent animals. Newborn mice were then analyzed at P1 and P7 for expression of GR-MDSC in intestinal mucosa and for intestinal microbiome composition at P7. We found decreased numbers of GR-MDSC also in intestinal lamina propria leucocytes of Hif1a loxP/loxP LysM Cre+ animals at P1 (median 1.5% versus 2.9%, n = 9-10, p < 0.01) and to a lesser extent at P7 (median 1.8% versus 2.7%, n = 5, p = 0.14) ( Figure 3A and B). In parallel, we found an altered microbiome composition with a reduced diversity and an altered balance of the intestinal microbiome in Hif1a loxP/loxP LysM Cre+ animals at P7 ( Figure 3C-E). Of the most abundant taxa detectable in all animals, WT mice had a higher proportion of Proteobacteria, Bacteroidetes, and Actinobacteria and a lower proportion of Firmicutes ( Figure 3F-I).

Discussion
The neonatal period is a very special phase of life, particularly from an immunological point of view. Like all other organs of the newborn, the immune system has to change from an intra-to an extrauterine state. While in utero, the most important requirement for the fetal immune system is to avoid being rejected by maternal immune cells, postnatally it must learn quickly to effectively fight off infectious agents. At the same time, however, it must tolerate billions of microorganisms colonizing body surfaces to form the microbiome. In the present study we show that a targeted deletion of the transcription factor HIF-1α in myeloid cells led to a disturbed accumulation of immune-suppressive MDSC and especially GR-MDSC in neonatal mice both systemically and locally in the gut. This disturbed MDSC-accumulation was accompanied by an altered intestinal microbiome composition and changes in intestinal immune homeostasis, especially in the activation state of intestinal T-cells.
Our first finding of increased MDSC-levels in newborns compared to adults has already been described by us and others in humans [18,[36][37][38]. From both, term and preterm human neonates it is known that levels of MDSC remain increased over the first four to six postnatal weeks, which suggests that they contribute to the increased susceptibility to infection on the one hand, but possibly also play a role for immune regulation in the context of microbiome establishment on the other hand [37,39]. Corresponding to our results, others have shown an accumulation of MDSC and especially GR-MDSC in mice during the first three postnatal weeks [21]. The latter study further showed that MDSC seem to play a role in controlling inflammation during NEC and hypothesized that this may be due to a suppressed inflammatory response to microbiome establishment [21]. However, the impact of MDSC on microbiome establishment per se has not yet been investigated.
We also found that the accumulation of GR-MDSC observed in newborn WT mice was significantly less pronounced in neonatal Hif1a loxP/loxP LysM Cre+ mice. Similar observations were made under tumor conditions where HIF-1α activation mediated an arrest in MDSC differentiation, leading to an MDSC accumulation [40]. Conversely, blockade of HIF-1α led to decreased MDSC accumulation in a tumor microenvironment and resulted in successful eradication of hepatocellular carcinoma in combination with an anti-PD-L1 therapy [41]. Furthermore, expression of HIF-1α is relevant to functional activation of MDSC under cancer conditions [22,23,42]. Our group showed that Hif1a loxP/loxP LysM Cre+ led to a diminished accumulation of GR-MDSC in the spleens and uteri of pregnant mice and impaired their suppressive capacity significantly thereby confirming a role of HIF-1α in MDSC accumulation and function under physiological perinatal conditions [17]. Since the fetus develops in an environment with low oxygen pressure [43,44] and HIF-1α is one of the most important transcription factors regulating the cellular response to hypoxia [45], it is obvious that immune cells whose accumulation and function is crucially triggered by HIF-α are impaired in newborn Hif1a loxP/loxP LysM Cre+ mice and that after birth, when hypoxia is normally no longer present, a rapid convergence to the numbers observed in WT animals occurs. In a very recent study we showed in vitro that HIF-1α accumulation was strongly diminished in neonatal immune cells hinting toward a general impairment of HIF-1α regulation during the neonatal period, probably influenced by the physiologically low oxygen levels in utero [46]. In qPCR analyses of spleens from newborn wild-type mice, we detected mRNA expression of hypoxia-dependent proteins glucose transporter 1 (GLUT1) and carbonic anhydrase IX (CAIX). For GLUT1 we found no difference in mRNA expression between P1 and P7, for CAIX a slightly decreased expression at P7 compared to P1 ( Figure S4), which might indicate that immediately after birth more hypoxia-activated signaling pathways are activated. However, further investigations are necessary to draw definite conclusions here. It is important to mention that in addition to hypoxia also alternative pathways have been described to activate HIF-1α [47]. It has been shown that activation of the proinflammatory NF-kB pathway can stimulate transcription of HIF-1α [48]. Because proinflammatory activation of the immune system appears to occur around birth [49], this pathway, in addition to the hypoxia present intrauterine, may be partly responsible for the HIF-1α-mediated accumulation of MDSC in the neonatal period. The same is true for the sex hormone estrogen, which can also activate the HIF-1α pathway [50] and has been described as a mediator of MDSC accumulation [51]. The observation that MDSC are not completely absent in neonatal Hif1a loxP/loxP LysM Cre+ mice could be explained by the fact that signaling pathways independent of HIF-1α may be involved and that adaptation to hypoxia may be related to the chronically hypoxic environment of the fetus. Regarding the mechanism for the decreased MDSC accumulation in newborn Hif1a loxP/loxP LysM Cre+ mice, we found in a previous study an increased apoptosis rate in MDSC of Hif1a loxP/loxP LysM Cre+ mice during pregnancy. A survival-promoting effect of HIF-1 has also been widely described in the context of tumor disease [52][53][54][55], suggesting that alteration of apoptosis properties by lack of HIF-1α also affects MDSC homeostasis in the newborn.
A limitation of our work is that we could not compare the suppressive activity of MDSC isolated from newborn WT and newborn Hif1a loxP/loxP LysM Cre+ animals. In our experience, the strength of the inhibitory effect of MDSC in a T-cell proliferation assay is highly dependent on the experimental conditions and so a good comparison between two MDSC sources is only possible when the proliferation assays are set up in parallel on the same target cells. Unfortunately, this was not possible because WT and Hif1a loxP/loxP LysM Cre+ offspring were not born at the same time point. However, in a previous work, we were able to show that in vitro generated MDSC from adult Hif1a loxP/loxP LysM Cre+ mice had a significantly poorer ability to inhibit T cell proliferation compared to MDSC generated from WT mice [17].
We found cells with phenotypic characteristics of GR-MDSC also in the intestinal mucosa, however, due to the low numbers, we could not perform functional experiments with these cells. Numbers of GR-MDSC in intestinal mucosa were low compared to those in spleen. However, also in intestines, newborn Hif1a loxP/loxP LysM Cre+ animals exhibited lower numbers of GR-MDSC than WT animals. The healthy intestine is characterized by hypoxia, especially at the tips of the villi [56] and it has been shown that HIF-1α regulates myeloid cell infiltration to the intestinal mucosa [57,58]. Corresponding to our results of decreased intestinal MDSC numbers in Hif1a loxP/loxP LysM Cre+ animals, Kim et al. found increased numbers of neutrophilic cells in mice overexpressing myeloid HIF-1α and decreased numbers of neutrophilic cells in mice Hif1a loxP/loxP LysM Cre+ during DSS induced colitis [57].
In our study, the decreased MDSC numbers in neonatal Hif1a loxP/loxP LysM Cre+ mice were accompanied by an altered microbiome composition. As mentioned before, we are not aware of other studies investigating an impact of MDSC on intestinal microbiome composition. However, it has been shown that a myeloid knockout of HIF-1β leads to altered microbiome composition in mice exposed to hypoxia [59]. Conversely, metabolites produced by the intestinal microbiome were shown to influence the expression of HIF-1α; for example, short chain fatty acids (SCFA) cause HIF-α stabilization by inhibition of prolylhydroxylases (PHDs) [60].
We observed increased relative abundances of Firmicutes in the microbiome of neonatal Hif1a loxP/loxP LysM Cre+ mice compared to neonatal WT mice, while relative abundance of Bacteroidetes and Proteobacteria were decreased. Increased Firmicutes:Bacteroidetes ratio has been described in obese patients and animals compared to normal-weight individuals [61,62], however the mechanisms underlying these correlations are only incompletely understood [62]. To our knowledge, there are no data on weight gain or microbiome composition of adult Hif1a loxP/loxP LysM Cre+ animals that may indicate that the changes in microbiome composition we here observed may alter the metabolic profile of the animals in the longer term. Further studies are needed to analyze the relationship between MDSC in the neonatal period, microbiome composition, and long-term outcome. The decreased diversity and balance observed in neonatal Hif1a loxP/loxP LysM Cre+ mice may indicate the presence of dysbiosis in these animals. Dysbiosis during early life has been linked to chronic diseases in later life such as asthma, allergy, diabetes, and impaired neurocognitive outcome [63]. In our work, we show an association between MDSC numbers in the neonate and intestinal dysbiosis. Whether there is really a causal relationship here remains unclear. It also remains unclear whether the relationship between MDSC and microbiome is direct or indirect. In addition to altered MDSC numbers, we observed altered Tcell populations and T-cell activation in the gut of Hif1a loxP/loxP LysM Cre+ mice. It has already been shown by other groups that, for example, Tregs play an important role in the establishment of the microbiome [64,65]. For example, there might be an indirect effect of MDSC on the microbiome via the induction of Tregs. More studies are needed to further elucidate this association and its underlying mechanisms. If MDSC during the neonatal period do indeed help establish a healthy microbiome, this could lead to new preventive approaches for a variety of diseases.
Lastly, we found an altered immune cell composition in intestines of neonatal Hif1a loxP/loxP LysM Cre+ compared to neonatal WT mice at P7. While numbers of intestinal macrophages were similar in both mouse strains, Hif1a loxP/loxP LysM Cre+ mice exhibited significantly higher numbers of intestinal DCs and IMCs. A previous study showed that under tumor conditions HIF-1α mediates suppressive activity in macrophages [22]. Even if relative macrophage numbers in our study did not differ between Hif1a loxP/loxP LysM Cre+ and WT animals, a decreased suppressive capacity of macrophages under the hypoxic conditions of the intestinal mucosa could interact with decreased MDSC numbers to form a less suppressive environment in Hif1a loxP/loxP LysM Cre+ animals. For DCs, a lack of HIF-1α led to increased inflammation in a model of DSS-induced colitis model [66]. Since targeted deletion of HIF-1α in our model was coupled to the promotor LysM expressed in all myeloid cells the differences observed for myeloid cells are probably directly due to the HIF-1α knockout and not to changes in MDSC homeostasis caused by it. Together, the observed alterations in the myeloid cell compartment in these Hif1a loxP/loxP LysM Cre+ mice point toward an increased inflammatory readiness compared to WT animals.
With regard to non-myeloid cells, there were increased numbers of total T-and NK-cells in Hif1a loxP/loxP LysM Cre+ intestines, while numbers of B-cells were similar to those in WT intestines. While intestinal CD4 + and CD8 + T-cell subpopulations did not differ between Hif1a loxP/loxP LysM Cre+ mice and WT mice, we found increased numbers of DN T-cells in Hif1a loxP/loxP LysM Cre+ animals. The role of DN T-cells in the periphery has not yet been clearly elucidated [67]. However, in a recently published work, DN T-cells in murine intestine were shown to be commensal and cytokine-responsive T cells that can contribute to intestinal pathogenesis in the context of barrier breakdown [68]. Furthermore, we found decreased numbers of regulatory Tcells and increased expression of the activation marker CD44 on intestinal T-cells from Hif1a loxP/loxP LysM Cre+ mice. As lymphocyte HIF-1α expression is not affected by the targeted deletion in our model the changes in lymphocyte composition and activation must be mediated indirectly by the altered state of the myeloid cells. Tregs play an important role for regulating immune homeostasis and inflammation during microbiome establishment [69] and increased numbers of intestinal Tregs have been shown to protect from inflammatory diseases like asthma or colitis, especially if induced during the neonatal period [70,71]. We and others showed that Tregs are induced by MDSC [14,20,72]. In the data presented here, it remains unclear whether the decreased Treg numbers in neonatal Hif1a loxP/loxP LysM Cre+ mice are due to decreased MDSC numbers or the altered microbiome. The performance of stool transplantation experiments could provide information in this regard. However, in combination with the increased CD44 expression on T-cells, which is a marker for effector T-cells, our results show that beyond changes in myeloid cells, the intestinal lymphoid compartment of Hif1a loxP/loxP LysM Cre+ mice also shows increased proinflammatory readiness. Anti-inflammatory effects of HIF-1α have been described several times [73][74][75] often involving nucleotide metabolism [76,77] and the neuronal guidance molecule netrin-1 [75,78,79]. Thus, one possible mechanism by which lack of HIF-1α in myeloid cells could lead to decreased Treg numbers would be via the adenosine signaling pathway. Adenosine has previously been shown to be important for the accumulation and function of Tregs in the context of pulmonary inflammation [73,74]. In addition, MDSCs are known to express ectonucleotidases that contribute to the formation of adenosine in a HIF-1α dependent pathway [80]. Moreover, very recently, netrin-1 was shown to increase the suppressive activity of MDSC [81], so that decreased netrin-1 levels in Hif1a loxP/loxP LysM Cre+ mice could potentially have an effect on MDSC functionality. However, these considerations are pure speculation here and require further investigation.
Taken together, our data suggest that the presence of MDSC in the newborn plays a role in inflammation regulation in the context of microbiome establishment. Similar observations have also been made for CD71 + erythroid cells, another suppressive cell population that also appears to control inflammation in the gut of newborn mice during microbial colonization. At the same time, Elahi et al. demonstrated that the presence of CD71 + cells contribute to the increased susceptibility to infection of newborn mice [13]. Whether this is also true for MDSC is not yet clear. Data from adult mice show that absence of HIF-1α in myeloid cells reduces inflammation in the setting of sepsis and improves survival [82]. However, the role of MDSC was not investigated in this study. Further work is needed to better understand the role of MDSC in neonatal immune regulation to potentially develop targeted therapies.

Tissue collection and single cell preparations
Spleens were removed form neonatal mice on postnatal days 1 (P1) and 7 (P7) and from adult mice. Tissue was pushed once through a 40μm filter (Greiner bio-one, Frickenhausen, Germany) using a syringe plunger. Whole intestines were removed in toto and cut longitudinally. Stool samples were collected from the colon and immediately frozen at -80°C. Small intestines were washed three times with ice-cold PBS. Afterward they were incubated in PBS with 5mM EDTA and 10% FCS to separate intestinal epithelial cells (IECs) for 30 minutes. Intestines were then sectioned into 2mm segments and incubated with 0.5mg Liberase (Roche, Basel, Switzerland) for one hour. Afterward, cells were centrifuged, resuspended in 40% Percoll (Cytiva, Marlbourogh, USA), and layered on 70% percoll. After density gradient centrifugation for 20 minutes at room temperature the lamina propria leucocytes were collected from the interface between the two percoll layers. At P1 small intestines from two mice were pooled to obtain enough cells for analysis. All cell suspensions were then adjusted to 1-4×10 6 cells/ml in PBS.

Cell isolation and flow cytometry
For isolation of GR-MDSC from murine splenocytes, cells were labeled with Gr-1 Biotin-antibody and isolated over Streptavidin microbeads followed by a second isolation step using Ly6G Biotin-antibody and Anti-Biotin microbeads (modified protocol of MDSC Isolation Kit mouse, Miltenyi, Bergisch-Gladbach, Germany). Purity of GR-MDSC after separation was >90%, as determined by flow cytometry.
For isolation of CD4 + T-cells from splenocytes, cells were labeled with T-cell Biotin-antibody cocktail followed by two sequential anti-Biotin magnetic bead separation steps (Miltenyi Biotec, Bergisch-Gladbach, Germany) according to the manufac-turer´s instructions. Purity of CD4 + T-cells after separation was For analysis of effector enzyme expression by in vitro generated MDSC, cells were first stained extracellularly with CD11b FITC and Gr-1 PE-Cy7. Afterward, cells were fixed and permeabilized using the BD Cytofix/Cytoperm and BD Perm/Wash kit (BD BioScience, Heidelberg, Germany) according to manufacturer's instruction. Antibodies used for intracellular staining were Arginase-1 PE (R&D Systems, Minneapolis, USA), iNOS PE (CXNFT, eBioscience, San Diego, USA), and IDO PerCP-eFluor 710 (mIDO-48, eBioscience, San Diego, USA).

RNA Isolation and cDNA synthesis
Spleens were removed from neonatal mice at P1 and P7. Tissue was frozen in liquid nitrogen and homogenized using mortar and pestle. RNA was isolated using the NucleoSpin ® RNA kit (Macherey-Nagel, Düren, Germany), and cDNA synthesis was performed with the ProtoScipt II First Stranc cDNA Synthesis Kit (New England BioLabs, Frankfurt am Main, Germany) according to manufacturer's instructions.

DNA extraction, library preparation and 16S rRNA sequencing
DNA was extracted from stool samples using the ZymoBIOMICS TM DNA Miniprep kit (ZymoResearch, Feiburg i. Br., Germany) according to manufacturer´s instructions. Genomic DNA was quantified with a Qubit dsDNA BR/HS Assay Kit (Thermo Fisher Scientific, Waltham, MA, USA) and normalized to 50ng Input for library preparation. The first step PCR was performed in 25μl reactions including KAPA HiFi HotStart ReadyMix (Roche), 515F [27], and 806R [28] (≈350 bp fragment of the 16S V4 region) and template DNA (PCR program: 95°C for 3 min, 28x (98°C for 20 s, 55°C for 15 s, 72°for 15 s), 72°C for 5 min). First PCR products were purified using 28μl AMPure XP beads and eluted in 50μL 10mM Tris-HCl. Indexing was performed in the second step PCR including KAPA HiFi HotStart ReadyMix (Roche), index primer mix (Illumina Nextera XT Index Kit v2), purified first PCR product as template (PCR program: 95°C for 3 min, 8x (95°C for 30 s, 55°C for 30 s, 72°C for 30 s), 72°C for 5 min). After another bead purification (20μl AMPure XP beads, eluted in 30μL 10mM Tris-HCl) the libraries were checked for correct fragment length on an agarose gel, quantified with a Qubit dsDNA BR Assay Kit (Thermo Fisher) and pooled equimolarly. The pool was sequenced on an Illumina MiSeq device with a v2 sequencing kit with 2×250 bp read length and a depth of 50-80k reads per sample.
The information about abundances of taxa at the appropriate levels (genus, order, family, class or phylum) were extracted from plain text file using R scripts (R version 4.2.0). Next, data were normalized to get tables containing the relative abundances via R code. Diversity indices were provided by application of the diversity function implemented in the vegan library (version 2.6.2). Balances were calculated according to [35]. For the calculation of the balances only taxa were considered that had no zero counts in the abundance tables. In each of the case groups WT and Hif1a loxP/loxP LysM Cre+ the medians of taxas´abundances were calculated. The difference between these determinates the classification of taxa to denominator or numerator in the formula for the balance calculation. The formulas given in [35] were applied. These were plotted applying ggplot from ggplot2 library (version 3.3.6).

Statistical analysis
Statistical analysis was done using GraphPad Prism 9.1.2 (Graph-Pad Software, La Jolla, CA). Comparisons between more than two groups of unpaired and not normally distributed data were analyzed using the Kruskal-Wallis test and Dunn´s multiple comparison test. Comparisons between more than two groups of paired and not normally distributed data were analyzed using the Friedman test and Dunn´s multiple comparison test. Comparisons between two groups of unpaired and not normally distributed data were evaluated using the Mann-Whitney test. A pvalue < 0.05 was considered statistically significant.