Gap gene network and positional information

Perkins, Jaeger, Reinitz and Glass (2006) modelled four trunk gap genes: hb, Kr, kni and gt. Expression is set up during cleavage cycles 13 and 14A, a span of about 70 minutes.³ At the start of cycle 13 there is some hb in the anterior, from maternally deposited mRNA. The other three genes are not unambiguously detectable.³ By the end of cycle 14A there are six main peaks, two each for hb and gt and one each for Kr and kni.³ The authors treated bcd, cad and tll as given inputs and did not model them.³

Perkins and colleagues built their last two models on logical rules. A gap protein is produced if at least one activator exceeds its threshold and no repressor exceeds its own.² This implements the plan of broad activation by maternal gradients, with local repression by other gap genes.¹ Lobo and Levin (2015) place this work in a wider setting. They note that most tools for inferring networks from expression data yield models without spatial information. They name the gene circuit method and later automated approaches as successes: they recovered the gap gene network from spatial data.⁴

Pietak and colleagues (2019) describe planarian factors such as β-Catenin, Wnt and ERK as gradients with characteristic polarities along the anterior-posterior axis. They state that experimental interventions on these factors give heteromorphoses, consistent with positional information regulating the body plan.⁵ Their model and new experiments tested pre-existing gradients against net nerve direction. In small fragments without the ventral nerve cord, the axis rotated 90 degrees. They concluded that axon polarity is the dominant driver of polarity.⁶ Fields and Levin (2022) note that in C. elegans the first division sets the axis.⁹

McMillen, Walker and Levin (2022) give a caution from the segmentation clock. Delayed mutual information between cells along the axis suggested that posterior cells caused the state of anterior neighbours. Ablating the tailbud cells left the hairy wave intact, so the causal inference was wrong.⁷ McMillen and Levin (2024) report that single cells moved to another axial domain change their expression to match neighbours. Groups of cells moved the same way keep their original expression.⁸ They read this as collectives holding positional memory longer than single cells.⁸