The Cdh5-CreERT2 transgene causes conditional Shb gene deletion in hematopoietic cells with consequences for immune cell responses to tumors

The tamoxifen-responsive conditional Cdh5-CreERT2 is commonly used for endothelial cell specific conditional deletion of loxP-flanked gene sequences. To address the role of endothelial cell Shb gene for B16F10 melanoma immune responses, tamoxifen-injected Cdh5-CreERT2/WT and Cdh5-CreERT2/Shbflox/flox mice received subcutaneous tumor cell injections. We observed a decrease of tumor myeloid cell Shb mRNA in the tamoxifen treated Cdh5-CreERT2/Shbflox/flox mice, which was not present when the mice had undergone a preceding bone marrow transplantation using wild type bone marrow. Differences in CD4+/FoxP3+ Tregs were similarly abolished by a preceding bone marrow transplantation. In ROSA26-mTmG mice, Cdh5-CreERT2 caused detectable floxing in certain bone marrow populations and in spleen cells. Floxing in bone marrow could be detected two months after tamoxifen treatment. In the spleen, however, floxing was undetectable two months after tamoxifen treatment, suggesting that Cdh5-CreERT2 is operating in a non-renewable population of hematopoietic cells in this organ. These data suggest that conditional gene deletion in hematopoietic cells is a potential confounder in experiments attempting to assess the role of endothelial specific effects. A cautious approach to achieve an endothelial-specific phenotype would be to adopt a strategy that includes a preceding bone marrow transplantation.

The Shb gene 1 plays a role in tumor biology in numerous settings [2][3][4][5][6][7][8] . Many of the findings point to an endothelial cell involvement 2,3,9,10 , but the Src-homology 2 domain protein B (SHB) also has an impact on immune or hematopoietic cell behaviour [11][12][13][14] . Shb is required for vascular endothelial growth factor-A (VEGFA) dependent angiogenesis and vascular leakage in endothelial cells 9 and these effects appear to be mediated via regulation of focal adhesion kinase 2,15 . T-cell receptor activation 12 also requires Shb and in the absence of Shb, T cells exhibit an augmented Th2-response 14 . Absence of Shb decreases hematopoietic stem cell proliferation causing a reduced ability of myeloid cells to repopulate after bone marrow replacement 13 . Our recent finding that CD8+ cell infiltration into B16F10 melanomas was influenced by the Shb gene raised the possibility that endothelial cells exert an influence on immune responses to tumors in a manner that could be of relevance to tumor expansion and metastasis 4 , and we decided to investigate this further by crossing the Cdh5-CreERT2(1Rha) transgene Cre-recombinase onto the Shb flox/flox background. This transgene is considered the gold standard for endothelial specific conditional deletion of loxP target genes 16 and has not been reported to generate inactivation of hematopoietic cells in adult mice unlike the Tie2-Cre transgene which efficiently causes gene deletion in hematopoietic cells 17 or the constitutive Cdh5-Cre transgene that causes recombination in embryonic hematopoietic cells 18 . One report suggests Cdh5-Cre-dependent recombination in hematopoietic cells in adult mice but that study used a different Cdh5 promoter fragment to drive Cre expression 19 . A third transgenic Cdh5-CreERT2 mouse was generated that perturbed angiogenesis but was not further investigated in detail with respect to its capacity to cause non-endothelial cell gene inactivation 20 . Herein we observed that the Cdh5-CreERT2 transgene 16 causes conditional gene deletion in certain hematopoietic cells with functional consequences that can be avoided by implementing protocols utilizing a preceding bone marrow transplantation.

Results and Discussion
Mice (Cdh5-CreERT2/WT or Cdh5-CreERT2/Shb flox/flox pretreated with tamoxifen) with B16F10 melanomas were investigated for endothelial-dependent alterations in Treg immune cells as a consequence of Shb gene inactivation. Immune organs (thymus, inguinal lymph nodes, spleen, bone marrow and blood) were collected and subjected to immune profiling by FACS staining to detect CD4/FoxP3 double-positive Tregs. Absence of Cdh5-CreERT2 induced deletion of Shb in endothelial cells reduced a tumor-induced increase of CD4+/FoxP3+ Tregs in local lymph nodes and accentuated that cell population in bone marrow (Fig. 1a). Next, these analyses were supplemented with bone marrow transplantation experiments using wild type bone marrow to Cdh5-CreERT2/Shb flox/flox or Cdh5-CreERT2/WT mice. The shift in the CD4+/FoxP3+ population caused by absence of Shb disappeared (Fig. 1b) suggesting that the effects were cell autonomous to Tregs. Isolated endothelial (CD31+) cells from tumors showed a 75% reduction of Shb mRNA by qPCR (Fig. 2a). Unexpectedly, an identical reduction of Shb mRNA in CD11b+ cells was noted (Fig. 2a) and such a reduction was observed regardless of whether compared with wild type mice (Fig. 2a) or Shb flox/flox mice (S. Fig. 1A). The cell populations were highly enriched for VE-cadherin/Cdh5 and CD11b (Itgam) (S. Fig. 1B,C), respectively, excluding endothelial cell contamination as an explanation for the reduction in Shb mRNA in CD11b+ cells. The reduction in tumor CD11b+ Shb mRNA was also reverted by wild type bone marrow (Fig. 2b). This suggests conditional deletion of Shb in hematopoietic cells by the Cdh5-CreERT2 transgene as responsible for the effects.
We next investigated floxing of Shb flox/flox in bone marrow CD31+ and CD11b+ cells by genotyping and detected gene inactivation in both cell types at day 4 after the cessation of the tamoxifen injections in adult To directly demonstrate Cdh5-CreERT2 activity in hematopoietic cells, the ROSA26-mTmG reporter mouse was used together with the transgene. This reporter expresses under basal conditions tdTomato and begins expressing GFP after floxing. However, the tdTomato signal is stable and may remain for weeks after Cre recombination 24 . The GFP/FITC signal was intermediate (10 3 -10 4 ) in bone marrow cells (S. Figs 3, 4) and consequently there was no distinct difference between GFP-positive cells and autofluorescence. For that reason, Cre-negative controls were always used in parallel under identical conditions. In the bone marrow, lineage-positive cells lacked recombination that showed statistical significance seven days after tamoxifen injections (Fig. 3). In cKit + Sca1 + Lineage− (K + S + L−) stem and multi-potent progenitor cells a low but detectable degree of recombination occurred that was consistently higher than that of the Cdh5-CreERT2 negative cells using the identical gating for GFP+ (Fig. 3, S. Fig. 3). Another bone marrow population with detectable recombination was a hitherto uncharacterized cKit + Sca1 + Lineage+ cells (Fig. 3, S. Fig. 4). We could not detect significant endothelial cell contamination in the K + S + L− population by FACS analysis (S. Fig. 3) since there were no or very few GFP hi -cells (>10 4 ) present in that population, which is in contrast to what was observed in lung endothelial cells in which the GFP/FITC FACS-signal is >10 4 (S. Fig. 5) when analyzed under identical conditions. The low degree of endothelial cell contamination could be a consequence of bone marrow crushing without digestion during preparation and the gating for high cKit expression. Standard protocols for bone marrow endothelial cell isolation include enzymatic digestion 25,26 and endothelial cell cKit expression is lower than that of cKit + Sca1 + Lineage− hematopoietic stem/multipotent progenitor cells 22,27 .
In spleen, myeloid cells (CD11b+/Gr1+), B cells (CD19+/B220+) and T cells (CD4+/CD8+) all exhibited recombination of the ROSA26-mTmG locus and the effect became more apparent by fixing the cells and staining for GFP (Fig. 3, S. Fig. 6). The expression of the ROSA26-mTmG reporter was significantly lower in hematopoietic (CD45+/CD31+) cells than in endothelial cell when flow cytometry was performed under identical conditions in the same FACS analysis (S. Fig. 5, results not shown). Lung CD45/CD31-positive hematopoietic cells also showed detectable recombination (S. Fig. 5) comparable to that of spleen myeloid cells (S. Fig. 6). When ROSA26-mTmG recombination was analyzed two months after tamoxifen treatment, floxing was detected in the bone marrow populations, whereas floxing could not be detected in the spleen (Fig. 4). This suggests that floxing in spleen occurs in a non-renewable hematopoietic population in the spleen. www.nature.com/scientificreports www.nature.com/scientificreports/ www.nature.com/scientificreports www.nature.com/scientificreports/ In summary, the Cdh5-CreERT2 transgene results in conditional deletion in hematopoietic cells. The effect may seem modest as suggested by the ROSA26-mTmG reporter data ranging between 3-20% but could play a significant role in certain settings. The data are based on two separate Cdh5-CreERT2 sublines excluding the possibility that unique drifting in phenotypic properties has occurred during local breeding. Firstly, differences between different hematopoietic cell types are noted. A partial explanation may lie in differences in Cdh5-CreERT2 expression but also in how susceptible the locus is to recombination in different cells 28,29 . In addition, different loci may show different susceptibility to floxing 29 . Another issue is whether there is positive or negative selection for the hematopoietic cell with a gene inactivation causing altered cellular properties. Finally, the pathological condition at which the study is conducted may influence the efficacy of conditional deletion. Currently in tumor CD11b+ cells, 75% of these appear to have deleted their Shb gene based on mRNA content and this had phenotypic consequences since tumor-induced alterations in Tregs were observed.
There are several reasons why Cdh5-CreERT2 16 activity in hematopoietic cells has remained hitherto undetected. The tdTomato signal is very stable so one must gate tdTomato/GFP double positive cells and due to the relatively weak GFP signal these are only moderately above background fluorescence as depicted by the Cre-negative cells. In addition, the population most active in the bone marrow is an uncharacterized lineage+/Sca1+/cKit+ population probably reflecting a progenitor subtype. Finally, conditional deletion is inititally (after one week) readily detected in spleen and undetectable two months later.
The potential of conditional deletion in hematopoietic cells can easily be avoided by performing a preceding bone marrow transplantation but that has potential adverse effects as well. Besides the simple fact that the mice will be of older age since complete hematopoietic restitution takes three months and causes changes in the relative proportions of different hematopoietic subpopulations 30 , the irradiation has direct adverse effects on endothelial cells 31 .

Conclusion
The Cdh5-CreERT2 transgene may conditionally delete target genes in certain hematopoietic cells and in order to avoid possibly confounding results, a strategy employing a preceding bone marrow transplantation can be implemented.
All animal experiments were approved by the Regional State Administrative Agency for Southern Finland and by the local animal ethics committee at the Uppsala County Court. EU regulations and guidelines for housing and care of laboratory animals were followed (DIRECTIVE 2010/63/EU on the protection of animals used for scientific purposes). www.nature.com/scientificreports www.nature.com/scientificreports/ Cdh5-CreERT2/WT (wild type) or Cdh5-CreERT2/Shb flox/flox (5 mice each) as described 4 . The mice were maintained for 3 months before further experimentation.
Tumor studies. B16F10 melanoma cells (2 × 10 5 ) were injected subcutaneously as described 4 in mice of [8][9][10][11][12] weeks of age that were injected for five consecutive days with 2 mg tamoxifen dissolved in peanut oil ten days after the last tamoxifen injection. Alternatively, mice subjected to a bone marrow transplantation three months prior with wild type bone marrow were used. When tumors reached a size of 0.5-1 cm 3 mice were sacrificed and relevant organs collected.
Immune cell profiling. Thymi, spleens, inguinal lymph nodes, bone marrows and blood were collected and cells isolated that were stained for CD4 and FoxP3 and analyzed by fluorescence activated cell sorting (FACS) as described 5 . Cell isolation by magnetic beads. Tumors were excised and digested as described 4 . Alternatively, iliac and femur bones were crushed and cells were collected by washing with PBS. Spleens were disrupted with a 1 ml syringe and cells collected in PBS. Blood was drawn immediately before sacrifice. Bone marrow, spleen and blood was subjected to red blood cell lysis by incubation for 15 minutes on ice in red blood cell lysis buffer (150 mM NH 4 Cl, 10 mM NaHCO 3 , 1 mM EDTA).  www.nature.com/scientificreports www.nature.com/scientificreports/ Real-time reverse transcriptase PCR (qPCR). Gene expression was determined using the QuantiTect TM SYBRGreen real-time RT-PCR kit (Qiagen) as described 4,5 . For Shb, 5′-TTT GAT GCC AAG AGC GAC CT and 5′-GAG TCT GAG TCC ACG CTC TG primers were used. For bone marrow VE-cadherin/Cdh5 mRNA, the following primer pairs were used: 5′-TTG CCC TGA AGA ACG AGG AC; 5′-ACT GCC CAT ACT TGA CCG  TG and 5′-AGC AGT GGA TGC AGA TGA CC; 5′-GCC TGT TTC TCT CGG TCC AA. Beta-actin values were subtracted and relative expression was determined according the formula 2 −∆Ct compared with relevant control.
Flow cytometry. The transgenic mice were injected for five consecutive days with tamoxifen (2 mg/mouse) and maintained for seven days before sacrifice. Bone marrow cells were stained with purified antibodies recognizing lineage markers (CD3, CD4, CD8, Gr1, CD11b, B220, CD19 and Ter119) followed by an anti-rat secondary antibody (PerCP-Cy5.5, Biolegend). Subsequently, cells were stained for cKit (APC-Cy7, eBioscience) and Sca1 (APC, eBioscience). Spleen and blood cells were stained for lineage (Gr1 and CD11b; B220 and CD19; CD4 and CD8) followed by anti-rat secondary antibody. Stained cells were subjected to flow cytometry by FACS essentially as described 13 . Spleen cells were fixated in 4% paraformaldehyde subsequently to initial flow cytometry and stained for Alexa488-anti-GFP prior to a second FACS. Antibodies used are listed in Supplemental Table 1 and FACS data were analyzed in Flowjo after flow cytometry on a BD Fortessa.
Statistics. Means ± SEM for the number of observations are given where one mouse is one n. The number of experiments, i.e. occasions at which a number of mice (of both genotypes) were analysed, are also given. Comparisons of chance differences were done by Students' t-test when compared with corresponding control or by ANOVA followed by Sidak's or Dunnett's multiple comparisons test as indicated using Prism 8 software.