Homologs of Ancestral CNNM Proteins Affect Magnesium Homeostasis and Circadian Rhythmicity in a Model Eukaryotic Cell

Firstly, to identify homologs in the Ostreococcus tauri proteome, the Pfam DUF21 consensus sequence was used to BLAST against its proteome. Using this method, two proteins were identified in the Ostreococcus tauri proteome: OtCNNM1 (ostta05g01490) and OtCNNM2 (ostta11g01030) (Figure 1A). In addition to the DUF21 and CBS-pair domains, CNNM1 presents a cyclic nucleotide–monophosphate binding domain (cNMP bd) commonly found in ion channels and kinases, both in eukaryotes and prokaryotes [20]. CNNM2 has a C-terminal CorC-HlyC ion transport domain.

We included reference proteomes of representative species for other eukaryotic clades. After removing all redundant isoforms, we built a phylogenetic tree with all identified sequences. This analysis revealed two clearly distinct phylogenetic clades (Figure 1B). Clade A includes OtCNNM1 and all human proteins (Homo sapiens CNNM1-4), plus at least one protein from every eukaryotic proteome, including the previously characterized CNNM proteins from Saccharomyces cerevisiae (MAM3) and Medicago truncatula (MtCBS1) (Figure 1B). On the other hand, Clade B includes OtCNNM2 and exclusively other proteins from photosynthetic species, or prokaryotic species. Unlike members of Clade A, protein sequences in Clade B have a CorC-HlyC ion transport domain.

Given that Mg²⁺ cannot travel freely across biological membranes [3], the activity of the Mg²⁺ transport system directly determines how Mg²⁺ is allocated and accumulated within cells. Conflicting reports about the roles of CNNM proteins in Mg²⁺ homeostasis have been published [21]. Given the fact that Ostreococcus tauri has a single homolog in each of the two clades, it represents an excellent model to study the contributions of both types of CNNM proteins to Mg²⁺ homeostasis. We cloned full-length gene versions under the control of the constitutive HAPT promoter [22] and that containing a C-terminal STREPII tag. We then transfected these overexpressing vectors, validated transgene expression by Western blot, and selected lines with high protein levels for CNNM1 (C1.1, C1.4, C1.6) and CNNM2 (C2.2, C2.5, C2.6. Figure S1). We then explored the effect of overexpressing CNNM proteins on overall cellular Mg²⁺ content using a well-described in vitro luminescent assay to measure Mg²⁺ [3] in whole-cell extracts. Interestingly, our results reveal a drastic reduction of 10–15% of overall Mg²⁺ content in the overexpressing lines compared with the parent line (Figure 2A). This confirms a role for CNNM proteins in [Mg²⁺] accumulation consistent with outward transport.

However, whole-cell Mg²⁺ content is dictated by transport over the plasma membrane between the cytosol and extracellular space. To reveal differential transport from the cytosol into intracellular organelles, we differentially lysed the plasma membrane with the non-ionic detergent digitonin, releasing the cytosolic Mg²⁺ content [23]. Upon extracting Mg²⁺ from the resulting pellet, containing all intracellular-membrane-bound compartments, we found that the overexpression of CNNM proteins led to an increase of up to 20% in the accumulation of Mg²⁺ (Figure 2B). These results indicate that while increased CNNM activity lowers the overall cellular content of Mg²⁺, a larger percentage of it is in the organelles, in turn meaning that the cytosolic concentrations must be depleted. Although structural changes such as differential organelle size could affect these measurements, our results indicate that cells might compensate against low cytosolic Mg²⁺ by storing more in the organelles. Regardless, our results confirm that the abundance of representative proteins from both clade A and clade B of CNNM proteins directly affect subcellular Mg²⁺ homeostasis.

Our previous work established that intracellular Mg²⁺ levels are important for the circadian rhythmicity of cells [3]. We then sought to test whether the observed changes in Mg²⁺ levels upon overexpression of CNNM1 or CNNM2 proteins translated directly to changes in circadian gene expression. The background line we used to generate CNNM overexpression lines expressed a fusion protein of the TTFL clock gene Circadian Clock Associated 1 (CCA1) with firefly luciferase (LUC) [22]. The lines enabled us to study the effect of CNNM overexpression on the parameters of TTFL gene expression rhythms. Our results show that under constant light conditions (LL), CCA1-LUC rhythms are significantly lengthened upon the overexpression of either CNNM1 (Figure 3A–D) or CNNM2 (Figure 3E–H). This indicates that both CNNM proteins directly impact on circadian timekeeping.

Previous studies have established cell treatments that inhibit or alter Mg²⁺ transport, such as Cobalt(III)hexammine (Co(NH₃)₆²⁺, CHA) [3,15]. CHA is known to interfere and specifically block Mg²⁺ transport, leading to severe phenotypes as abolished Mg²⁺ rhythmicity [3]. To test whether the circadian defects observed in overexpressing lines of CNNM proteins (Figure 3) are directly caused by the differential accumulation of Mg²⁺ in these lines (Figure 2), we next treated the CNNMox lines with a wide range of [CHA] to test the sensitivity of the clock compared with the parent line. CHA treatments led to dose-dependent circadian period lengthening in the parent line (Figure 4A), consistent with previous results [3], as well as in the overexpression lines (Figure 4B–D). The effect of CHA on circadian rhythms in a line overexpressing CNNM1 was similar to that observed in the parent line. However, the effect of the overexpression of CNNM2 on circadian period was strikingly synergistic with CHA treatments. The synergy of CHA and CNNM2 overexpression on lengthening circadian rhythms implies that both treatments influence the clock through two separate cellular mechanisms. This result indicates that differential subcellular localization and/or the different domain architecture of Clade A and B of CNNM proteins leads to a differential sensitivity to the Mg²⁺ transport blockage.

Given the observed changes to subcellular Mg²⁺ homeostasis and the circadian clock, we hypothesized that the overexpression of CNNM proteins could alter physiologically relevant properties of cellular life. We therefore tested whether the known deleterious effects of Mg²⁺ deprivation on cell proliferation were differential in CNNM overexpression versus parent lines. Cells were subjected to a four-order magnitude range of extracellular Mg²⁺([Mg²⁺]_e), and the growth rate was measured by cell counting for 8 days. Interestingly, when compared with the parent line, the overexpression of either CNNM1 or CNNM2 can protect against extremely low extracellular Mg²⁺ (micromolar range, Figure 5A–E). However, only the overexpression of CNNM1 led to increased cell proliferation at any of the Mg²⁺ concentrations tested. The observed phenotypes in cell proliferation were not linked to clock rhythms; CCA1 rhythms were the same for all lines in all conditions (Figure S2). Although these data are consistent with the Mg²⁺ transport activity of CNNM proteins, it contradicts the hypothesis that CNNM proteins mediate Mg²⁺ extrusion out of the cells only: an efflux protein would exacerbate the effect of low extracellular Mg²⁺ on growth. This could mean that these proteins can mediate both Mg²⁺ efflux and influx, depending on the physiological and environmental status and the direction of gradients over a biological membrane.

Homologs of characterized CNNM proteins were searched for in the predicted proteomes of model species representative of various eukaryotic taxonomic groups, including Ostreococcus. To identify homologs of characterized CNNM transport proteins, we manually used the BLAST webpage interface (NCBI) [35]. BLAST searches with the human CNNM2 protein sequence (Uniprot protein Q9H8M5↗), using BlastP with default search parameters against the reference proteins database (refseq_protein) by specifying the organism queried. Searches used the blastP (protein–protein BLAST) algorithm. We did not download any proteins with E-values lower than the threshold, and removed all redundant isoforms and partial sequences. Proteins were renamed for phylogenetic analysis to include the organism name and a protein number or protein name if previously published (Table S1). Additionally, the CorA and CorC proteins from Methanoculleus thermophilus and Gloeomargarita lithophora, and the CBS1 protein from Medicago truncatula, were added to this list. All curated protein sequences were aligned using MUSCLE in MEGA 11.0.11 using default settings [36]. Protein alignments were used to build maximum likelihood (ML) trees (100 bootstraps), using the Jones–Taylor–Thorton (JTT) substitution model and the nearest neighbor interchange (NJJ) substitution model.

Ostreococcus cultures were grown for 6–7 days in artificial sea water (ASW) under 12/12 h light/dark cycles at 20 °C, under enriched blue light (183 Moonlight Blue Filter), as previously described [37]. New transgenic lines overexpressing OtCNNM1 (ostta05g01490) and OtCNNM2 (ostta11g01030) genes were PCR-amplified using specific primers incorporating the STREP tag sequence (Table S2). Then, fragments were digested with AvrII (NEB) and ligated into the pOTOX vector [22]. Genomic transformation was conducted as previously published [38] and performed under the CCA1-LUC background line [22]. Positive colonies were selected with ClonNAT antibiotic (Nourseothricin, Jena Bioscience) and protein samples were validated by immunoblotting using an anti-STREP antibody (Strep-II antibody, abcam ab184224).

Seven-day-old Ostreococcus cultures were diluted at a 1:3 dilution rate and mixed with 0.2 mM D-luciferin. Then, 90 µL of this solution was added to wells in a 384-well plate (Greiner) and imaged in a TriStar2 luminescent plate reader (Berthold) under 2 μmols m⁻² s⁻¹ of blue light (183 Moonlight Blue Filter; Leefilters) under LL. To test the effect of CHA (Sigma), wells were filled with 81 µL of diluted culture and mixed with 9 µL of each 10x [CHA] (or water). For washout treatments for low-Mg concentrations, 80–90 µL media from each well was carefully replaced with 80–90 µL of treated media with luciferin (or ASW + luciferin), avoiding disturbing the cells. Results were analyzed and plotted using GraphPad Prism v9 or with BioRender. Circadian parameters were quantified with BioDare2 [39].

Cell pellets were harvested from 1-week-old cultures and resuspended in Buffer 1 (50 mM Hepes-HCl pH 8, 150 mM NaCl). For whole-cell samples, cell suspensions were mixed while vortexing with equal volumes of 2x Buffer 1 + 1% Triton X-100; then, the supernatant was collected after centrifugation at maximum speed. For organelle fractions, suspensions were mixed with 2x Buffer 1 + 0.01% Digitonin (Sigma). After 5 min incubation at 4 °C and centrifugation at 4000 rpm for 15 min, the organelle fraction was obtained after discarding the resulting supernatant and resuspending pellets in Buffer 1 + 0.5% Triton X-100.

The quantification of intracellular Mg²⁺ was performed using a luminescent plate assay, as previously described [3], with a slight alteration: an assay buffer with 50 mM Hepes-HCl pH 8, 1 mM luciferin, 0.05 mg·mL QuantiLum (Promega), and 1 mM ATP was used [3]. Data from at least 3 technical replicates for each time point were plotted.

Ostreococcus growth rates were quantified by counting changes in the cell number using a hemocytometer with a light microscope at 40× magnification. The same starting cell density from 1-week-old cultures was resuspended and incubated in different low-Mg media under LL conditions, and the cell number was counted every 2 days. Data from at least 5 technical replicates for each time point were plotted.

Not applicable.