Donor DNA (IDT) and gRNA (IDT; designed via crispr.mit.edu) (as a Crispr Cas9 RNP) (21, 22) were co-microinjected into 1- to 2-cell mouse embryos, implanted into pseudo-pregnant females, and allowed to come to term (sequences are shown in Supplementary Table S1 ). Out of 84 pups, 12 were positive by genomic PCR genotyping (using 5’ and 3’ primer pairs as indicated in Figure 1 , Supplementary Table S1 ), and 5 males were chosen to be further analyzed. One did not transmit via the germ line, and out of the remaining 4 lines we chose 2 for the extended analysis below. These were designated ‘EKLF-CRE’ mice.

To assess tissue-specificity of CRE activity, male EKLF-CRE mice were crossed with female R26R-YFP mice (kind gift from N Dubois (23)), all at 4-6 weeks of age.

Genotyping was performed on 20 ng genomic DNA isolated from tails by PCR analysis using the diagnostic oligonucleotide primers in Supplementary Table S1 . These assessed the 5’ (F0.5/CreR) and 3’ (CreF/3A) junctions between the knock-in donor DNA and the EKLF gene ( Figure 1 , Supplementary Figure S1A ). Use of the ‘outer’ primer pair (F0.5/3A) yields a small product in the absence of the knock-in ( Figure 1 , Supplementary Figure S1A ). Subsequent experiments show that multiplex PCR will also work with these oligos (not shown), thus enabling single-lane monitoring of knock-in and WT alleles for hetero- and homo-zygosity. Expression levels of endogenous EKLF RNA do not decrease, but rather appear to be stabilized by the knock-in ( Supplementary Figure S1B ).

Fetal livers were dissected from embryonic day E13.5 embryos or adult bone marrow and mechanically dispersed into single cells for fluorescence activated cell sorting (FACS). Briefly for bone marrow (BM), femurs were removed and the marrow was flushed by gently inserting a 23-gauge needle fitted with 3 ml syringe into the opening of the bone using 1 ml cold PBS/10%FBS. Fetal liver or bone marrow dissociated cells were filtered through a 70uM cell strainer (Falcon # 352350), and collected by centrifugation for 5 min at 1000rpm, and finally washed with PBS/10%FBS.

Cells to be analyzed for macrophage included the FWV peptide during the isolation procedure to help prevent aberrant macrophage/erythroid cell associations when used in combination with effective dispersal and singlet enrichment (17, 24).

Suspended cells were stained with the following antibodies: anti-mouse F4/80-PE (eBiosciences #12-4801-80), CD71-PE (eBioscience #12-0711-81), Ter119-APC (eBioscience #17-5921-81), CD44-FITC (Biorad #MCA89FT), anti-spectrin b1 (Santa Cruz #sc-374309) conjugated with AlexaFluor 647 (Invitrogen #Z11235), Sca1 (Ly6A/E)-PE (BD #BDB561076), B220 (CD45R)-PE (BD #BDB561878), CD3-PE, BD #BDB561799). Flow cytometry data was analyzed by FCS Express software, and gates were drawn based on unstained and single-color compensation controls from the same samples, using the same dyes and within the same experiment.

Erythroblastic island clusters were enriched from dispersed adult BM using a serum gradient (17).

For immunofluorescence of BM single cells or island clusters, cytospins were incubated with anti-spectrin b1 (Santa Cruz #sc-374309), anti-F4/80 (eBiosciences #12-4801-80), or anti-EKLF 7b2a (25) as described (17). Secondary antibodies included Alexa flour 568 goat anti-mouse IgG (Invitrogen #A11031) and Alexa flour 647 donkey anti-mouse IgG (Invitrogen #A-31571).

Photography was performed with a Zeiss Axio Imager.Z2(M) equipped with a Q-Imaging camera.

BM cells were obtained as above. Sca1+, CD3+, B220+ and Ter119+ BM cells were then isolated using the EasySep mouse PE positive selection kit (Cell Signaling Technologies #17656) according to the manufacturer’s instructions (detailed method available in (24)). 0.3ug of the selected PE-labeled antibodies (as used above for flow cytometry) were added for each isolation. For Ter119+ cells we used 0.3ug of anti-mouse TER-119-PE antibody (BD #553673).

RNA was isolated from selected BM cells using Trizol and purified RNA was treated with Turbo DNase (Invitrogen #AM1907). cDNA was synthesized using the SuperScript IV VILO kit (Invitrogen #11756050) and real time qPCR was performed using a Qiagen QuantiTect SYBR Green RT-PCR Kit. Primers are listed in Supplementary Table S1 .

A Easi-Crispr-cas9 approach (21, 22) was used to knock in the iCRE coding segment into the 3’ sequence of the endogenous mouse EKLF gene, fused in-frame to the final leucine amino acid ( Figure 1A ). This entailed the design of a directed Crispr gRNA along with a single-stranded donor DNA. The donor DNA included a self-cleaving T2A sequence (26) flanked by linkers, adjacent to the NLS/iCRE sequence (derived from pCAG-iCRE (27)). The ultimate design (based on (28)) relies on the endogenous EKLF polyA signal. Two lines were more extensively analyzed.

To test for CRE functionality, we crossed the EKLF-CRE mice with the Rosa26/stop/EYFP reporter line (29), which expresses YFP only if CRE is active and removes the flox-embedded stop sequence ( Figure 1B ). This yielded mice at the expected Mendelian ratios; we used littermate embryos for comparison. E13.5 embryos were harvested, and individual fetal livers cells were dispersed and analyzed by FACS. At this stage ~90% of cells are erythroid (30). Erythroid cells were monitored by CD71/Ter119 cell surface markers ( Figure 2A ) as these provide a good initial survey of immature to mature red cells. After proper gating and enrichment of singlets we find that the number of live cells are equivalent when comparing YPF+ to YFP- littermates ( Figure 2A i), and that the percentage of erythroid cells within this population is equivalent ( Figure 2A ii). There is an excellent signal with respect to YFP detection, which is near absent in cells without EKLF-CRE ( Figure 2A iii). Importantly, we find the YFP+ cells overlap the CD71/Ter119 expression pattern identically to that seen endogenously ( Figure 2A iv). These data indicate that CRE expression driven by the EKLF transcription unit recapitulates erythroid-restricted expression with high penetrance in developing embryos.

To monitor CRE activity earlier in erythroid differentiation, we analyzed YFP overlap with the CD44 marker ( Figure 2B ). The CD44 sialoglycoprotein is expressed highest in proerythroblasts but then decreases as maturation proceeds (32). As before, there is no effect on live cell numbers ( Figure 2B i) or the CD44/Ter119 pattern ( Figure 2B ii), and the YFP+ signal remains unique to the EKLF-CRE expressing cells ( Figure 2B iii). These analyses again show complete overlap of YFP+ with the CD44/Ter119 range of expression seen endogenously (e.g., double positive cells are similar in number; Figure 2B iv). Combining these markers along with cell scatter properties (FSC) provides an accurate assessment of maturity of erythroid cells (31). As seen in Figure 2C , the CD44+/FSC profile within the Ter119+ subset shows that CRE expression is apparent in stages from proerythroblast to orthochromatic erythroblast, distributed as seen in prior fetal liver erythroid analyses (i.e., P2>P1>P3 (25);), and highest in the least mature population (P1; colored in red). This is a direct overlap with the known EKLF RNA and protein expression patterns (33, 34).

To assess whether EKLF-CRE is also functional in the adult, we performed the analogous study using bone marrow (BM) cells as our starting material. These data again show a direct overlap of YFP+ with CD71/Ter119 ( Figure 3A top); however, in this case there are differences in detail compared to the fetal liver material. Although YFP+ overlaps with CD71+Ter119-, CD71+Ter119+, and CD71-Ter119+ cells as expected, there are a considerable number of YFP+ cells that are double negative, properties that had not been observed with fetal liver erythroid cells (likely these are F4/80+ cells, as explained below). To assess more subtle changes, we monitored subpopulations by gating on total live cells and color-coding the output ( Figure 3A bottom, i and ii). These reproduce Figure 3A top (judged by the orange color YFP+ subset within the total cell population) but additionally show that not all CD71+Ter119+ cells are YFP+ ( Figure 3A bottom, iii), observations likewise not seen with fetal liver-derived cells. Analysis of CD44/Ter119 expression shows that YFP+ cells segregate into discrete subpopulations encompassing different levels of CD44 and Ter119 ( Figure 3B i), and inclusion of FSC with selective gating (Ter119+ in combination with the CD44-lo to -hi (31)) shows that EKLF-CRE (YFP+) is most highly expressed in the least mature (ie, CD44+/Ter119+/FSC-mid to -hi) erythroid populations ( Figure 3B ii), as perhaps more readily apparent by comparison to the total live cell population (i.e., both YFP+/-; Figure 3B iii).

Collectively these data demonstrate that EKLF-CRE is active and is a robust mimic of normal EKLF erythroid expression in the developing embryo and in the adult.

EKLF is also expressed within the central macrophage of the erythroblastic island (17, 18, 20, 35), a niche that provides a physiological support environment for erythroid maturation (36–39). F4/80 is an excellent marker for this cell (20). Using E13.5 fetal liver cells as our source we find that CRE is expressed in the F4/80+ cell population ( Figure 4A ); however, as seen in earlier studies, not all cells are positive, but rather 39 ± 3% are, which matches that previously seen by marked GFP+ (36% (17)). These results, using an independent approach from those used previously (16–20), provide additional support for EKLF’s expression in the F4/80+ central macrophage of the erythroblastic island.

One issue not hitherto addressed is whether EKLF is also expressed in the adult BM F4/80+ macrophage cell. As with our published fetal liver analyses (20), we included the FWV peptide to interfere with erythroid/macrophage interactions (17, 24) and used stringent forward- and side-scatter sorting prior to analysis ( Supplementary Figure S2 ). Using YFP positivity as a surrogate marker for EKLF expression, we find that YFP expression is expressed in almost all (>90%) BM F4/80+ cells ( Figure 4B ). In this case it is also noteworthy that the total number of F4/80+ cells is ~17%, which is higher than that seen in the fetal liver, and serves to identify most of the large percentage of non-erythroid cells (i.e., double negative for CD71/Ter119) observed before (in Figure 3 ). To further verify that the YFP signal is a true mimic of endogenous EKLF expression, we performed immunofluorescent analysis of BM cells and find that indeed EKLF protein is colocalized in single F4/80+ cells ( Figure 4C ; cytoplasmic EKLF signal is expected (40)). These data show for the first time that, as seen during embryonic development, EKLF is also expressed in adult BM-derived erythroblastic island macrophage. The unexpected surprise, however, is that it is expressed in almost all F4/80+ cells.

We had previously found that Adducin (Add2), ß-spectrin (Sptb), and to a lesser extent Adra2b, are cell surface markers enriched in EKLF-expressing fetal liver F4/80+ macrophage (20). To assess whether this is the case in adult BM, we performed FACS analyses for each of these. Although the levels of Add2 and Adra2b expression in F4/80+ BM cells were insignificant (not shown), we find that 11% of F4/80+ BM cells are positive for both Sptb+ and YFP+ ( Figure 5A , Supplementary Figure S2 ). Immunofluorescent analyses show co-expression of EKLF/YFP, F4/80, and ß-spectrin in single cells ( Figure 5B ) and in erythroblastic islands ( Figure 5C ). These results demonstrate that expression of the Sptb surface marker in adult F4/80+ cells that express EKLF are as observed in the developing embryo.

Of relevance to the present resource, these data demonstrate that EKLF-CRE can be used to target a specific subset of macrophage, one that is highly enriched within the island niche, in both fetal liver and adult BM cells.

The level of adult BM YFP+ cells that are CD71-/Ter119- is not completely accounted for by F4/80+ macrophage cells. We investigated this more carefully, initially by immunofluorescence (IF) where we find that, as expected, not all bone marrow cells are YFP+; surprisingly however, not all YFP+ are EKLF+ ( Figure 6A ).

The onset of EKLF expression during hematopoiesis begins prior to the megakaryocyte-erythroid progenitor (MEP) stage within the multipotent progenitor (MPP) and retained in the myeloid progenitor (CMP) but not the lymphoid progenitor (CLP) (14, 15). We FACS-sorted adult BM populations by using Sca1 as a cell surface marker for HSCs, B220 for B cells, and CD3 for T cells, and verified by RNA analysis that EKLF expression resides in the Sca1-, CD3-, and B220- populations ( Figure 6B ). As a positive control, Ter119+ erythroid cells were found to be highly enriched for expression ( Figure 6B ), consistent with the earlier analyses ( Figure 3 ). In addition and importantly, using primers specific for the EKLF-CRE fusion RNA product, we find expression only within the Ter119+ cell population ( Figure 6C ). These results verify the expected tissue-specificity of EKLF RNA expression and that of the knock-in fusion variant.

These data suggest that our IF observations follow from a lineage trace that began within the first hematopoietic cell that expresses EKLF. In our knock-in mice, such early expression of EKLF-CRE will stably mark those cells and YFP expression will be retained in any progeny. We tested this idea by FACS sorting, which shows that YFP+ cells are present within a majority (~75%) of Sca1+ HSCs (or MPPs) and progeny that includes B (B220+) and T (CD3+) cell populations ( Figure 6D , Supplementary Figure S3 ). The purity of the cell types was verified by RT-qPCR analysis of the sorted populations ( Supplementary Figure S4 ).

These data fill in the contributions of EKLF-CRE expression to YFP positivity in adult BM marrow cell populations and support prior analyses that showed EKLF expression begins early within the hematopoietic hierarchy, likely within the MPP or even the HSC and thus prior to the CMP/CLP split. Even though EKLF expression is restricted to and expanded within the erythroid lineage, remnants of its early hematopoietic onset remain visible by YFP tracing.