Compartmentalization of enzymes is a cellular strategy to regulate metabolic pathways and increase their efficiency 1 . The αand β-carboxysomes of cyanobacteria contain Ribulose-1,5-bisphosphate carboxylase/oxygenase (Rubisco), a complex of 8 large (RbcL) and 8 small (RbcS) subunits, and carbonic anhydrase (CA) 2-4 . Since the proteinaceous carboxysome shell provides a barrier to the diffusion of CO2 but not HCO3 − (ref. 5), CA generates high concentrations of CO2 for carbon fixation by Rubisco 6 . The shell also prevents access to reducing agents, generating an oxidizing environment 7-9 . Formation of β-carboxysomes involves aggregation of Rubisco by the protein CcmM 10 , which exists in two forms: Fulllength CcmM (M58 in Synechococcus elongatus PCC7942) containing a CA-like domain 8 followed by three Rubisco small subunit-like (SSUL) modules connected by flexible linkers, and M35, lacking the CA-like domain 11 . It has long been speculated that the SSUL modules interact with Rubisco by replacing RbcS 2-4 . Here we reconstituted the Rubisco:CcmM complex and solved its structure. Contrary to expectation, the SSUL modules do not replace RbcS, but bind close to the equatorial region of Rubisco between RbcL dimers, linking Rubisco molecules and inducing phase separation into a liquid-like matrix. Disulfide bond formation in SSUL increases the network flexibility and is required for carboxysome function in vivo. Importantly, the formation of the liquid-like condensate of Rubisco is mediated by dynamic interactions with the SSUL domains, rather than by low complexity sequences, which typically mediate liquid-liquid phase separation in eukaryotes 12,13 . Indeed, within the pyrenoid of eukaryotic algae, the functional homologue of carboxysomes, Rubisco has been shown to adopt a liquid-like state via interactions with the intrinsically disordered protein EPYC1 14 . Understanding carboxysome biogenesis will be important in efforts to engineer CO2 concentrating mechanisms (CCM) in plants 15-19 .

In the present study, we tested the hypothesis that 17␤-estradiol (␤E2) is a neuroprotectant in the retina, using two experimental approaches: 1) hydrogen peroxide (H 2 O 2 )-induced retinal neuron degeneration in vitro, and 2) light-induced photoreceptor degeneration in vivo. We demonstrated that both ␤E2 and 17␣-estradiol (␣E2) significantly protected against H 2 O 2 -induced retinal neuron degeneration; however, progesterone had no effect. ␤E2 transiently increased the phosphoinositide 3-kinase (PI3K) activity, when phosphoinositide 4,5-bisphosphate and [32 ␥ATP] were used as substrate. Phospho-Akt levels were also transiently increased by ␤E2 treatment. Addition of the estrogen receptor antagonist tamoxifen did not reverse the protective effect of ␤E2, whereas the PI3K inhibitor LY294002 inhibited the protective effect of ␤E2, suggesting that ␤E2 mediates its effect through some PI3K-dependent pathway, independent of the estrogen receptor. Pull-down experiments with glutathione S-transferase fused to the N-Src homology 2 domain of p85, the regulatory subunit of PI3K, indicated that ␤E2 and ␣E2, but not progesterone, identified phosphorylated insulin receptor ␤-subunit (IR␤) as a binding partner. Pretreatment with insulin receptor inhibitor, HNMPA, inhibited IR␤ activation of PI3K. Systemic administration of ␤E2 significantly protected the structure and function of rat retinas against light-induced photoreceptor cell degeneration and inhibited photoreceptor apoptosis. In addition, systemic administration of ␤E2 activated retinal IR␤, but not the insulin-like growth factor receptor-1, and produced a transient increase in PI3K activity and phosphorylation of Akt in rat retinas. The results show that estrogen has retinal neuroprotective properties in vivo and in vitro and suggest that the insulin receptor/PI3K/Akt signaling pathway is involved in estrogen-mediated retinal neuroprotection.

No abstract