Pf - Expression reprogramming during liver development

Expression is highly variable during liver development: The switch is concentrated at hepatocyte entry and the sporozoite-to-trophozoite conversion; the mid-to-late liver transcriptome is comparatively stable by pairwise testing.

A more sensitive time-course model on the liver samples alone (quadratic in time, library batch blocked) still finds 811 genes (20% of tested) changing progressively across days 2–6, so late development is fine-tuning rather than a second wholesale switch — consistent in spirit with the paper's maSigPro result of 1,510 temporally regulated genes.

Peak-stage assignment of the 2,477: sporozoite 1,410, day 2 18, day 4 47, day 5 322, day 6 680.

Expression variation during liver development - Sporozoites

8-26 Claude Science: Yes — for P. falciparum it rests on one dataset, not on single-cell data

There is no single-cell transcriptomics of P. falciparum liver stages. Every single-cell liver-stage atlas is in another species: P. berghei (PMID 36352220, Nature 2022, 66 citations) and P. vivax (PMID 35443155, Cell Host Microbe 2022, 101 citations; PMID 36093191, 58; hypnozoite RBPs PMID 37392389, 22).

For P. falciparum the only genome-wide, time-resolved resource is bulk RNA-seq of parasites recovered from human liver-chimeric mice: GSE220039 (PMID 39891010, Nat Microbiol 2025, 20 citations; preprint PMID 36711670, 11) — days 2, 4, 5, 6 post-infection, three biological replicates each, with salivary-gland sporozoites as t₀ (sporozoite reference PRJNA344838, PMID 29539423, 129 citations). A 2025 review covers the landscape and its gaps (PMID 40263025). So I re-analysed that dataset from raw counts rather than taking the published figures on trust.

(a) PCA of parasite transcriptomes after rarefying every library to a common 9,481 parasite reads (16 libraries; the 376-read day-2 library excluded); points are libraries, the line joins stage centroids. (b) Differentially expressed genes (|log2FC| > 1, FDR < 0.05) from limma-voom on 4,891 filtered genes. (c) Stage-mean expression, z-scored per gene, of the 2,477 genes differentially expressed versus sporozoites, rows sorted by peak stage.

The quantitative case for "massive"

About half the expressed genome is reset relative to the sporozoite, and early intrahepatic development is dominated by shutdown — which reproduces the published statement that ~70% of DEGs are less abundant with 28–36% higher.

When it happens: day 2 → day 4 gives 799 DEGs, day 4 → day 5 only 28, day 5 → day 6 only 3. The switch is concentrated at hepatocyte entry and the sporozoite-to-trophozoite conversion; the mid-to-late liver transcriptome is comparatively stable by pairwise testing. A more sensitive time-course model on the liver samples alone (quadratic in time, library batch blocked) still finds 811 genes (20% of tested) changing progressively across days 2–6, so late development is fine-tuning rather than a second wholesale switch — consistent in spirit with the paper's maSigPro result of 1,510 temporally regulated genes. Peak-stage assignment of the 2,477: sporozoite 1,410, day 2 18, day 4 47, day 5 322, day 6 680.

Two things to watch in this dataset

• Depth drives the naive ordination. Day-2 libraries yield only 376–21,687 parasite reads, and in a PCA of all 17 libraries PC1 correlates with parasite read depth at Spearman rho = 0.77. Rarefying every library to a common 9,481 reads drops that to rho = 0.18 while preserving the developmental ordering along PC1 (sporozoite −145 → day 2 −58 → day 4 +72 → day 5 +104 → day 6 +106). Panel (a) uses the rarefied version.
• ComBat-seq with default arguments erases the effect of interest. Four of the five sporozoite libraries are the external PRJNA344838 batch, so batch correction without a stage covariate removes the sporozoite-versus-liver signal: the same contrasts then return 7–211 DEGs instead of 1,344–1,746. Handling batch in the design recovers the published magnitudes. Both variants are in the summary table.

Caveats: bulk data, so these are population averages with no single-cell or spatial resolution for P. falciparum; one host model and parasite line (FRG huHep, NF54); days 3 and 7+ unsampled, so the switch is bracketed only between days 2 and 4; day-2 numbers are the least reliable. Model assumption diagnostics were not assessed.

• Pf_liverstage_reprogramming_report.md — dataset survey, methods, full results and caveats
• Pf_liverstage_DE_summary.csv — DEG counts for all contrasts under three analysis variants
• liverstage_reprogramming_genes.csv — the 2,477 reprogrammed genes with stage z-scores and peak stage
• Pf_liverstage_dataset_landscape.csv — candidate datasets with PMIDs, DOIs and citation counts