Genetic Ablation of a Female-Specific Apetala 2 Transcription Factor Blocks Oocyst Shedding in Cryptosporidium parvum

Little is known about the transcriptional control of life cycle progression in Cryptosporidium and the factors that orchestrate the commitment to, and the execution of, different developmental cell fates. However, the C. parvum genome encodes numerous AP2 DNA binding proteins, homologs of which have been linked to stage-specific gene expression control in the related parasite Plasmodium (19). Analysis and experimental manipulation of these factors thus may provide insight into the Cryptosporidium life cycle. Initial PCR screening at different times of culture suggested that some of these factors themselves may underlie transcriptional control across the life cycle (29, 30). We recently used engineered reporter parasites to isolate specific stages of Cryptosporidium and carried out comparative transcriptomics by transcriptome sequencing (RNA-seq) (2). This analysis suggested five AP2s to be preferentially transcribed in sexual stages (cgd1_3520, cgd2_3490, cgd4_1110, cgd6_2670, and cgd8_810). We initially focused on cgd4_1110, which appeared to be transcribed in female gametes, and tagged the gene to yield C-terminal translational fusion with either an mNeon-green reporter, or a triple hemagglutinin (HA) epitope tag introduced into the native locus by Cas9-mediated homologous recombination (31) (see Fig. S1 in the supplemental material for maps and PCR validation). cgd4_1110-3xHA parasites were used to infect HCT-8 cultures, which were subsequently fixed at 24 and 48 h postinfection, stained with an HA-specific antibody, and counterstained with Vicia villosa lectin, which recognizes glycan epitopes found on all C. parvum stages. No HA staining was observed at 24 h, a time point at which only asexual forms are found. However, at 48 h, when differentiation to gametes had occurred, 60.7% (±4,4%) of parasites were HA positive (Fig. 1A and B). Using cgd4_1110-mNG parasites, we similarly found label only at the 48-h time point. Relying on nuclear morphology to distinguish stages (2, 13) at this time point, we detected the reporter in female gametes, but not in asexual meronts or in male gamonts (Fig. 1C). Consistent with the presumptive function of a DNA-binding protein, the label coincided with the single nucleus of the female gamete (arrowhead, Fig. 1C). To further validate this finding using a molecular marker, we infected cells with cgd4_1110-HA parasites and stained them with antibodies to HA and the meiotic recombinase DMC1, which was recently shown to be restricted to female gametes (15). We observed HA staining exclusively in DMC1-positive cells (Fig. 1D).

Previous transcriptional profiling work by us and others suggested that, due to the timing of expression, cgd6_2670 might also be a female-specific AP2 (2, 30). To further test this, the locus was tagged with a triple HA epitope or an mCherry reporter, and cultures infected with the resulting parasite strains were fixed at different time points. Upon staining, we observed fluorescence exclusively at the 48-h time point, as expected (Fig. 2A); however, compared to cgd4_1110, this staining was limited to a much smaller proportion of the overall parasite population (Fig. 2B). Analysis of the mCherry strain showed nuclear fluorescence exclusively in parasites with 4 or 8 round nuclei, which appeared clustered into a rosette at the center of intracellular parasite stages (Fig. 2C). No labeling was associated with identifiable asexual stages or with female gametes, but faint cytoplasmic staining was notable in mature male gamonts. The temporal pattern of cgd6_2670 transcription matches that of the male fusion protein HAP2 (Fig. 2D). We therefore studied parasites engineered to express an HA-tagged version of HAP2 (2, 32). HAP2 is found as a characteristic dot at the apex of mature male gametes and is perinuclear in immature gamonts, likely reflecting protein in the endoplasmic reticulum en route through the secretory pathway (Fig. 2E). Importantly, these immature stages showed centrally rosetted nuclei. We quantified the nuclear area of these stages (1.46 ± 0.32 μm²) and found it similar to that of cgd6_2670-positive rosette stages (1.829 ± 0.33 μm²) but significantly different from that of asexual meronts (Fig. 2F; 4.12 ± 0.83 μm²; P < 0.0001). The nuclear morphology of these stages is reminiscent of developing male gamonts recently observed by time lapse imaging (15). Next, we engineered parasites to express a tdNeon reporter under the control of the cgd6_2670 promoter to mark all stages in which this gene is transcribed. In contrast to our previous experiments, we did not fuse the reporter to the AP2 protein itself, thus releasing it from posttranslational control through development (see Fig. 2H). We introduced this construct into the HAP2 locus, adding an HA tag to specifically identify males. When studying the resulting transgenic strain, we again found fluorescence to be restricted to the 48-h time point, but now observable in higher numbers, matching those previously observed for male gamonts (Fig. 2G to J) (2). Importantly, essentially all parasites expressing tdNeon were also positive for the male marker HAP2-HA (Fig. 2K and L). We conclude that the two AP2 proteins characterized here are expressed in a female (cgd4_1110) and early-male-specific (cgd6_2670) fashion and thus named them AP2-F and AP2-M, respectively.

To understand the role of gamete-specific AP2 factors in Cryptosporidium life cycle progression, we next set out to disrupt their loci. An attempt to disrupt AP2-M was unsuccessful (data not shown). We next attempted to insert a drug marker into the N-terminal AP2-F DNA binding domain. This was unsuccessful in five attempts using three different guide RNAs. Parallel control experiments inserting an epitope tag without altering the coding sequence readily resulted in transgenic parasites (we note that we were able to disrupt the C-terminal DNA binding domain; this did not impact parasite growth or development, and we thus did not study this mutant any further; see Fig. S2). In C. parvum transgenesis, selection occurs in animals, and recovery of modified parasites requires the formation of oocysts, following the completion of the full life cycle (31). While failure to disrupt the locus is consistent with the essentiality of a gene, it is not proof, and we thus next sought to develop an experimental model to ablate genes in C. parvum conditionally. Multiple approaches have been adapted to apicomplexans for this purpose, including modulating transcription initiation (33), affecting the stability of the targeted mRNA or protein (34–37), or using recombinases to excise portions of the genome (38, 39).

Here, we used a modified version of Cre-recombinase (DiCre) that, upon rapamycin induction, assembles an active enzyme from two fragments to excise gene segments flanked by loxP recognition sequences (38, 40). To enable the introduction of loxP sites into the coding sequence of genes without changing their protein products, we explored the use of introns. We searched the C. parvum genome for short introns and through preliminary experimentation settled on the 73-bp intron from cgd1_1320, which is expressed across the life cycle (32). To test the functionality of this intron in trans, we inserted it into the nanoluciferase (Nluc) coding region (Fig. 3A). Transient transfection of this reporter into sporozoites followed by infection of HCT-8 cells resulted in luciferase activity indistinguishable from that obtained with a continuous Nluc coding sequence (Fig. 3B). Faithful splicing was critical for this, as the activity was lost when the 5′ splice donor site of the intron was ablated (Fig. 3B; P < 0.0001). Next, loxP sequences were introduced at different positions of the intron, and luciferase activity was measured. We identified multiple positions that resulted in activity comparable to that of the native intron (Fig. 3C; see Fig. S3 for detail). We conclude that the intron is faithfully spliced when introduced in trans, and its splicing remains intact when loxP sites are incorporated.

Next, we constructed a plasmid that expresses two fragments of Cre recombinase under the control of the C. parvum enolase promoter, each fragment fused to FKBP12 or FRB, respectively, and separated by a viral translational skip sequence (40). To test whether this system results in rapamycin-inducible Cre activity in C. parvum, we designed a transient reporter assay. In our test construct, Nluc expression was disrupted by a floxed stop sequence containing a yellow fluorescent protein (YFP) reporter (Fig. 3D). Cre-mediated excision should remove the stop sequence and result in Nluc expression. Transfection of this floxed reporter, or DiCre alone, did not result in luciferase activity above a background control. However, when parasites were simultaneously transfected with both constructs and grown in culture medium containing 100 nM rapamycin, robust activity was detected, and this activity depended on the presence of rapamycin (Fig. 3E; P < 0.0001). Please note that rapamycin does not impede Cryptosporidium growth at the concentration used here. We tested a range of concentrations and determined the 50% inhibitory concentration (IC₅₀) to be 8.3 μm (see Fig. S4).

Next, we tested the ability of the system to excise floxed sequences from the parasite genome. We targeted the well-characterized nonessential thymidine kinase (TK) locus (41, 42), replacing the last 222 bp of the native sequence with a recodonized version preceded by an artificial intron carrying a loxP site (Fig. 4A). We also appended a 3XHA epitope tag to be able to detect the protein along with cassettes for expression of DiCre and the Nluc-Neo selection marker. PCR mapping confirmed successful modification of the TK gene (Fig. S1), and Sanger sequencing of the amplified insert demonstrated that the targeted region was flanked by loxP sites. Using primers flanking the floxed region, we detected two amplicons representing the floxed locus, prior to and following Cre-mediated excision (note that all primers used in this study are listed in supplementary Data Set S2). The initially faint smaller 743-bp band became more noticeable upon continuous passage of the strain, suggesting that stable expression of DiCre recombinase in this strain resulted in some leaky activity in the absence of rapamycin.

We next infected HCT-8 cells with TK DiCre parasites, and cultures were harvested after 24, 36, and 48 h with or without rapamycin, and genomic DNA was isolated and analyzed by PCR for the presence of the floxed region (Fig. 4B). Rapamycin treatment resulted in progressive loss of the full-length band beginning at 24 h postinfection (hpi). We next scored parasites in immunofluorescence experiments using anti-HA antibody to detect the (floxed) epitope and Vicia villosa lectin (VVL) as counterstain for all parasites. As expected from the PCR mapping, we detected both tagged and untagged parasites in cultures grown in the absence of rapamycin. When rapamycin was added to cultures, we detected significant loss of HA staining (Fig. 4C and D; control versus rapamycin, 12.53% ± 4.67% after 48 h of rapamycin compared to 40.72% ± 6.5% in controls; P < 0.01). We also directly measured the biochemical activity of TK using 5-ethynyl-2′-deoxyuridine (EdU) incorporation. EdU incorporation relies on TK to activate the probe to the nucleotide level (31, 43). Wile-type (WT) parasites and a mutant lacking TK (31) were used as positive and negative controls, respectively. A total of 53.64% ± 6.5% of the WT parasites were EdU-positive, while no staining was observed in the TK knockout (KO) (Fig. 4E and F). In the TK DiCre strain 24.98% ± 8.7% of the untreated parasites were found to be EdU-positive, and this number dropped to 4.7% ± 5.94% upon rapamycin treatment (Fig. 4F, P < 0.05). We conclude that C. parvum genes tolerate a floxed intron retaining normal transcription and translation, and that treatment with rapamycin induces loss of the targeted gene and its encoded protein and activity. We also noted some leaky DiCre activity when targeting a nonessential gene.

We next sought to impose rapamycin-inducible ablation on the likely essential AP2-F gene. Our initial attempts to flox the entire gene while concurrently introducing cassettes for DiCre and selection marker failed. In our experience, targeting cassettes exceeding 5 kb in size often fail to recombine efficiently. We thus decided to flox the promoter and initiation codon, a much shorter, but nonetheless essential sequence (see schematic in Fig. 5A). To avoid undue recombination, we typically recodonize homologous sequences that are localized internally within the targeting cassette. This is not feasible for a promoter, and we therefore replaced the native AP2F promoter with a floxed promoter with a matching transcriptional profile. We chose the promoter of the Cryptosporidium oocyst wall protein 1 (COWP1), which is active exclusively in female gametes (2), and we also added an N-terminal 3XHA epitope tag. Transfection with this construct produced viable drug-resistant parasites, and PCR mapping and Sanger sequencing confirmed the modification, indicating that the promoter swap was tolerated (Fig. S1). HCT-8 cells were infected with the AP2-F floxed strain and incubated for 24, 36, and 48 h in the presence or absence of 100 nM rapamycin. PCR analysis of genomic DNA isolated from these cultures, using primers spanning both loxP sites, found a single band in parasites grown in the absence of rapamycin (Fig. 5B). Growth in the presence of rapamycin resulted in the appearance of a second smaller band, indicating promoter excision. To assess the impact of rapamycin treatment on the level of AP2-F protein, infected HCT-8 cultures were stained with antibodies to HA (AP2-F) and DNA meiotic recombinase 1 (DMC1, a marker for female gametes [15]; Fig. 5C). Rapamycin treatment resulted in significant loss of AP2-F staining, 7.62 ± 3.38% compared to 60.22% ± 18% in untreated controls (Fig. 5D, P < 0.01), but did not impact parasite growth in culture (Fig. 5G). As shown in Fig. 5E and F, loss of AP2-F did not result in loss of female parasites, suggesting that AP2-F does not act as a female commitment factor.

To test whether AP2-F affects gene expression and to identify genes modulated by the factor, we next performed mRNA sequencing experiments. Three biological replicates of WT and AP2-F-DiCre parasites were grown in HCT-8 cells with or without rapamycin for 48 h, RNA was extracted, and cDNA libraries were prepared for Illumina sequencing. Comparison of WT parasites treated with and without rapamycin yielded little difference in parasite gene expression (Fig. 5H), indicating that rapamycin alone did not induce a significant transcriptional response. Upon addition of rapamycin to the AP2-F-DiCre strain, we noted a reduction of transcripts for AP2-F itself by greater than 2.5-fold (Fig. 5I, highlighted in red). A total of 15 additional genes showed significantly decreased expression, and 17 genes increased in RNA abundance (see Fig. 5I and supplemental Data Set S1). Of the 16 genes with reduced expression upon treatment, 14 showed tight specificity for female gametes in a previous study (2), while only 6 of 17 with increased expression were restricted to female gametes. Most prominent among the female-specific transcripts was a cluster of crystalloid body proteins (all proteins linked to this sporozoite-specific organelle in a recent proteomic study [A. Guérin and B. Striepen, unpublished observation] are highlighted in magenta in Fig. 5H and I). Two proteins involved in meiosis, DMC1 and NIMA kinase 5, also showed a reduction, as did a DNA polymerase. We note that other abundant classes of female proteins, such as those that make up the oocyst wall proteome (highlighted in purple), remained unchanged, suggesting specificity for a subset of female genes.

Our initial KO experiments suggested AP2-F to be an essential gene, but while we noted changes in female transcription in culture upon rapamycin treatment, these changes did not result in obvious changes in either growth or differentiation over the development observable in HCT-8 cell culture (fertilization, meiosis, and oocyst formation and release do not occur in this system [2]). We thus considered that the processes governed by AP2-F, including the requirement for the crystalloid body, may manifest later in female development, and at a time point not well represented by in vitro culture. To test this, we conducted in vivo infection experiments in mice. Groups of three gamma interferon KO (IFN-γKO) mice were treated daily by gavage with 100 μL vehicle or 10 mg/kg rapamycin, beginning 2 days prior to infection with 10,000 AP2-F-DiCre strain oocysts (Fig. 6A). Nluc activity was monitored in mouse feces, and we observed significant reduction of oocyst shedding in mice treated with rapamycin compared to the vehicle control (Fig. 6B). We found this to be true in four independent biological experiments using groups of three mice each. Fig. 6C shows the area under the curve normalized for each matched group of treated and untreated mouse samples for all experiments (P < 0.0001). Mice were euthanized on day 6 in two of these experiments, intestines were resected, and Nluc activity was measured for three ileal punch biopsies per mouse. In contrast to fecal shedding, intestinal Nluc readings were not reduced but were moderately higher in treated mice (Fig. 6D). To measure the effects of rapamycin treatment alone, we used the TK-DiCre strain as a control (Fig. 6E to G; note that the TK gene is dispensable). We found rapamycin to be of no detriment to infection with these parasites, and luciferase readings were consistently higher upon treatment, regardless of whether we measured fecal or intestinal luciferase. This is likely due to the immunosuppressive effects of this drug (44). We conclude that AP2-F is critical to the formation and/or shedding of oocysts from infected animals but that it is not required for the growth of parasites in the intestinal tissue.

IFN-γ^–/– mice (stock no. 002287) were purchased from Jackson Laboratory and bred in-house. Mice of both sexes were used (typically ranging from 6 to 8 weeks in age). Mice were treated with antibiotics, infected, and handled as detailed in reference 60.

HCT-8 cells were purchased from ATCC (CCL-224TM) and maintained in RPMI 1640 medium (Sigma-Aldrich) supplemented with 10% Cosmic calf serum (HyClone), 1% penicillin-streptomycin (Gibco), 1% l-glutamine (Gibco), and 1× amphotericin B (GoldBio). For HCT-8 infection medium, serum concentrations were reduced to 1%.

If not indicated specifically, reagents were purchased from Sigma-Aldrich. Rapamycin (LC Laboratories) was dissolved in 95% ethanol at a stock concentration of 50 mg/mL. For in vivo studies, rapamycin was further diluted in water. Mice were weighed and treated daily by oral gavage with 100 μL of drug solution adjusted to deliver 10 mg/kg body weight (note that the ethanol concentration must not exceed 5% and was 2.5% in most of our experiments). A vehicle control was established from 95% ethanol by equivalent dilution.

Primary antibodies used in this study include rat anti-HA (Roche clone 3F10), VVL-FITC (1:1,000, Vector FL1231); DMC1 antibody (1:10, a kind gift of Christopher Huston, University of Vermont [15]), rabbit anti-TrpB (31) and mouse anti-alpha tubulin (a kind gift of Jacek Gaertig, University of Georgia), both used at 1:300; and secondary antibodies carrying a variety of Alexa Fluors (Abcam) at 1:1,000.

The DiCre cassette was built using a P. falciparum construct (40) kindly shared by Tobias Spielmann (Bernhard Nocht Insitut Hamburg, Germany). Guide oligonucleotides (Sigma-Aldrich) were introduced into the C. parvum Cas9/U6 plasmid16 by restriction cloning. See reference 61 for a detailed discussion of guide design for C. parvum. Transfection plasmids were constructed by Gibson assembly using NEB Gibson assembly master mix (New England Biolabs). File S2 provides the sequences of all primers used in this study. Note that constructing plasmids that contained both the DiCre expression cassette and the floxed segment often resulted in premature excision of the floxed segment in bacteria due to Cre activity. We thus engineered the two pieces independently and then generated our linear targeting constructs by fusing PCR amplicons containing the Dicre cassette and the floxed segments, respectively, by Gibson assembly (we recommend a 15- to 20-bp overlap between amplicons for ensuring fusion). The final assembled product was then PCR amplified using terminal primers and used in transfection experiments.

We generated and purified transgenic parasite lines using Cryptosporidium parvum Iowa II strain oocysts (Bunchgrass Farms, Deary, ID) following previously detailed protocols (31, 60, 61).

The Nano-Glo luciferase assay system (Promega) and Glomax 3000 (Promega) were used to measure luciferase activity in parasites, following the manufacturer’s protocol unless otherwise noted. For fecal samples, 19 to 21 mg of feces was processed as described previously in reference 61. For in vitro cultures, monolayers were directly resuspended in the NanoGlo substrate solution, and luminescence measured. For intestinal samples, gut punches were weighed and lysed in fecal lysis buffer (61) by vortexing for 10 min with glass beads. Samples were mixed 1:1 with NanoGlo substrate solution and assayed in triplicate.

We followed the protocols for transient transfection of parasites as described (31). Briefly, 1 × 10⁷ C. parvum oocysts (Bunchgrass Farms, Deary, ID) were excysted and transfected with 10 μg each of DiCre and floxed plasmids using an Amaxa 4D electroporator (Lonza). The transfected parasites were diluted in HCT-8 infection medium and divided equally to infect three 24-well plates of subconfluent HCT-8 cells. Nanoluciferase activity was measured 48 h after infection.

HCT-8 cells were infected with 50,000 oocysts in HCT-8 infection medium with and without 100 nM rapamycin. After different times of infection, DNA was isolated from the cells according to the quick-start protocol from the DNeasy blood and tissue kit (Qiagen). The PCR primers used are listed in supplemental Data Set S2.

HCT-8 cells were grown on coverslips in either 24- or 96-well plates and 20,000 of prepared oocysts in HCT-8 infection medium, respectively. For DiCre activation experiments, parasites were also grown in the presence of 100 nM rapamycin. Cells were fixed with 4% paraformaldehyde (Electron Microscopy Science) in phosphate-buffered saline (PBS) and then permeabilized with PBS containing 0.25% Triton X-100. The cells were blocked with 3% bovine serum albumin (BSA, Sigma-Aldrich) solution, followed by incubation with primary antibodies. Cells were washed with PBS and then incubated with appropriate fluorophore-conjugated secondary antibodies and counterstained with DAPI (4′,6-diamidino-2-phenylindole; Thermo Fisher). Coverslips were then mounted on glass slides with Fluoro-Gel (Electron Microscopy Science) mounting medium. Superresolution structured illumination microscopy was conducted using a Carl Zeiss Elyra (UGA Biomedical Microscopy Core) or a GE OMX (PennVet Imaging Core) microscope. Wide-field microscopy was performed using a Leica LAS X microscope (PennVet Imaging Core), and images were processed and analyzed using Carl Zeiss ZEN v.2.3 SP1, GE Softworx, and NIH ImageJ software.

HCT-8 cells were infected with 100,000 oocysts of the TK Flox Dicre strain. EdU (Thermo Fisher) was added to cultures 36 h after infection to a final concentration of 10 μM, and cells were fixed 12 h later. A Click-iT EdU Alexa-Fluor 594 kit (Thermo Fisher Scientific) was used to label incorporated EdU. Parasites were counterstained with anti-HA antibody and fluorescein-conjugated Vicia villosa lectin (Vector Laboratories).

Female IFN-γ^−/− mice (4 to 6 weeks old) were treated with 10 mg/kg of rapamycin (LC Laboratories) via oral gavage daily, starting 2 days prior to infection and continuing 5 days postinfection. Mice were sex and age matched between treatment groups. Following 2 days of pretreatment with rapamycin, mice were gavaged with 10,000 oocysts in PBS. Bulk fecal samples were collected, and intestinal punches were harvested on day 6. Punches were stored at –80°C prior to processing for nanoluciferase measurement.

HCT-8 cells were infected with 100,000 oocysts of wild-type (nontransgenic) C. parvum or AP2-F DiCre strains with or without 100 nM rapamycin. Cells were lysed and RNA was extracted using the RNeasy minikit (Qiagen, Hilden, Germany) according to the manufacturer’s protocol. The Illumina stranded mRNA prep ligation kit (20040534) was used for cDNA generation and library preparation. Total RNA and libraries were quality controlled and quantified using the TapeStation 4200 (Agilent Technologies, Santa Clara, CA) and Qubit 3 (Thermo Fisher Scientific, Waltham, MA) instruments. Samples were pooled, and single-end reads were generated on an Illumina NextSeq 500 sequencer.

Raw reads were mapped to the C. parvum Iowa II reference (VEuPathDB, release 50) using Kallisto v.0.45.0 (62). All subsequent analyses were carried out using the statistical computing environment R v.4.0.3 in RStudio v.1.3.1093 and Bioconductor. In brief, transcript-level quantification data were summarized to genes using the tximport package, and data were normalized using the trimmed mean of M values method, which was implemented in EdgeR (63). Only genes with more than 10 counts per million in at least 3 samples were used for analysis. Precision weights were applied to each gene based on the mean-variance relationship using the VOOM function in limma (64). Linear modeling and Bayesian statistics carried out in limma were used to identify differentially expressed genes with an FDR-adjusted P value of ≤0.01 and an absolute log₂-transformed fold change of ≥1 after adjustment with Benjamini-Hochberg correction. Genes were identified as oocyst wall and crystalloid body based on analyses derived from the C. parvum Hyper-LOPIT proteomic data set (A Guérin, K Strelau, K Barylyuk, B Wallbank, L Berry, OM Crook, KS Lilley, RF Waller, B Striepen, submitted for publication). Gene set enrichment analysis was carried out using Gene Set Enrichment Analysis (GSEA) software v.4.3.2 (65). Oocyst wall and crystalloid body signatures obtained from the Hyper-LOPIT data set were used for GSEA with 1,000 permutations, a weighted enrichment statistic, and the Diff_of_Classes metric for ranking genes. All other parameters were kept at the default settings. All code is found in supplemental Data Set S3.

GraphPad PRISM was used for all statistical analyses. When measuring the difference between the two populations, we used a standard Student’s t test. No statistical tests were used to predetermine sample size, and no animals were excluded from the results. Analysis of variance (ANOVA) was used to compare the means of multiple groups, followed by Tukey’s post hoc test for pairwise comparisons.

All the protocols for animal experimentation were approved by the Institutional Animal Care and Use Committee of the University of Georgia (protocol A2016 01-028-Y1-A4) and/or the Institutional Animal Care and Use Committee of the University of Pennsylvania (protocol number 806292). No statistical tests were used to predetermine the sample size of mice used for experiments. Mice were not randomized, and investigators were not blinded before any of the experiments.

The RNA sequencing data generated in this study are available from the GEO database repository under accession number GSE216844. All code used in these analyses is available in Data Set S3 and on GitHub (https://github.com/katelyn-walzer/Ablation_of_AP2-F_Manuscript).