Some signals are more equal than others

We take it for granted that administering growth factors such as EPO or GM-CSF causes a selective increase in patients' hematocrit or granulocyte/monocyte counts, respectively. These selective effects of cytokines are in part due to lineage-restricted expression of their cognate receptors on late hematopoietic progenitors. But since multiple cytokine receptors are coexpressed on early committed progenitors and stem cells, it is conceivable that these cells can distinguish among specific incoming signals originating from different receptors and translate this information into activation of differentiation programs. If so, we would of course like to break the code in which this information is encrypted.

Although very attractive, the concept of specific instructive signaling by cytokine receptors has remained controversial. Hisakawa and colleagues (page 3618) provide evidence largely in support of an alternative permissive model. They examined the effects of a functional human GM-CSF receptor (hGM-CSFR) ectopically expressed on erythroid progenitors of transgenic mice. Due to species specificity, activation of the transgenic hGM-CSFR is entirely dependent on exogenous human ligand. Furthermore, to exclude the possibility of cross-talk with the endogenous mouse EPO receptor (EPOR), hGM-CSFR mice were crossed with EPOR knockout mice, and fetal liver cells deficient for EPOR but expressing thehGM-CSFR transgene were derived. Under these stringent conditions, hGM-CSF promoted fully differentiated CFU-Es and BFU-Es in vitro. Furthermore, increased expression of the same adult hemoglobins as in EPO-treated wild-type control cells was observed. These results argue that the EPO and hGM-CSF signals are equal and interchangeable. But detailed analysis of yolk sac cells at embryonic day 8 revealed subtle differences: only hGM-CSF was able to induce expression of adult hemoglobins in primitive erythropoietic cells, while both receptors promoted erythroid-colony formation. Thus specific signaling information may be detectable only within a particular cellular context, and hence both the instructive and permissive models may be partially correct. The next question is, What are the molecular features defining the cellular context?