Nostoc sp. PCC 7119 (ATCC 29151) (Adolph and Haselkorn, 1971) was grown in BG11 (Rippka et al., 1979) (containing NaNO₃ as N source) or BG11₀ (lacking any source of combined nitrogen) medium at 30 °C under standard light conditions (25 μmol photons m^-2·s^-1 from led white lamps). Cultures were maintained in an orbital shaker (100 rpm) during liquid experiments or in solidified medium (1% w/v Difco Agar). For selection of mutant strains, Streptomycin and Spectinomycin antibiotics were added to the media at a final concentration of 5 μg·mL^-1 each as described previously (Torrado et al., 2019). Heterocyst formation was induced by transferring cultures grown in BG11 to BG11₀, after centrifugation at 3000 ×g and washing of the cultures with BG11₀.

Cyt c ₆, c _6BC, c _6D and f were expressed and purified as described previously (Molina-Heredia et al., 1998; Molina-Heredia et al., 1999; Molina-Heredia et al., 2001; Albarrán et al., 2005; Reyes-Sosa et al., 2011; Torrado et al., 2017). In short, Escherichia coli strains were co-transformed with the plasmids bearing each Cyt and the support plasmid pEC86, which encodes the E. coli genes required for Cyt c maturation (Arslan et al., 1998), in LB medium supplemented with the corresponding antibiotic at 37 °C during 24 h with continuous shaking (300 rpm). After the incubation, proteins were extracted from the periplasmic fraction after three freeze-thawing cycles and subsequent purification in an ion exchange chromatography column. The fractions containing Cyt were followed and analysed spectrophotometrically and in case that the purity was inadequate, another step of chromatography was applied. The A₂₈₀/A_55X absorbance ratio was used to estimate the purity of the sample, whereas Cyt c ₆, c _6BC, c _6D and f had a maximum absorption peak of the reduced α-band of 553, 553, 552 and 556, respectively. Isolated heterocysts were purified as described in Torrado et al., 2019. Vegetative and heterocyst membranes enriched in terminal oxidases were purified as described previously (Schmetterer et al., 2001; Torrado et al., 2019). In short, isolated membranes from Nostoc sp. PCC 7119 were prepared from 800 mL of cultures at a cellular density of 3-6 µg of Chl a·mL^-1. Cells were centrifuged at 12,000 ×g for 5 minutes, resuspended in 10 ml HEPES buffer (10 mM HEPES pH 7.4 and 6 mM NaCl) with supplementation of 20% sucrose (w/v) and 10 mg of lysozyme. The suspension was incubated 30 minutes at 37°C and centrifuged at 12,000 ×g for 5 minutes. The pellet was resuspended in 10 mL of ice-cold HEPES buffer and incubated on ice for 1 hour. The suspension was supplemented with 1 mM phenylmethylsulfonyl fluoride and 0.005% (w/v) DNase I and passed through a French press three times at 11,000 psi. After this, the suspension was centrifuged at 4°C at 12,000 ×g for 10 minutes, resuspended in 5 mL of ice-cold HEPES buffer and homogenizer in a Potter, to a final concentration of 3-5 mg·mL^-1 of total protein.

Cyt samples for isothermal titration calorimetry (ITC) were oxidised (Cyt c ₆, c _6BC and c _6D) or reduced (Cyt f) with 10 µM potassium ferrocyanide or 5 mM sodium ascorbate, respectively. Subsequently, samples were extensively dialysed with 5 mM of phosphate buffer (pH 7.0) to remove the oxidising/reducing agents. ITC experiments were carried out in 5 mM phosphate buffer, pH 7.5, supplemented with 0.02% (w/v) Triton X-100, using an Auto-iTC200 instrument (MicroCal, Malvern-Panalytical) at 25 °C and a stirring speed of 1000 rpm. The reference cell was filled with distilled water. The experiments were carried on through successive additions (2 μL injections) of concentrated Cyts c ₆, c _6BC and c _6D proteins (400 µM) to the sample cell containing the binding partner Cyt f (40 µM). All solutions were degassed before titrations. Titrant was injected at appropriate time intervals to ensure that the thermal power signal returned to the baseline prior to the next injection. Control experiments of the dilution of individual injected proteins were conducted and these reference heat values were subtracted from measuring values of test titrations when appropriate. The binding isotherm (ligand-injected normalized heat per injection as a function of the molar ratio) was analysed with Origin 7 (OriginLab). In all cases the heat evolved during titrations could be well fitted to a 1:1 binding stoichiometry, and the association constant, K_A (and the dissociation constant, K_D), and the binding enthalpy (ΔH) and entropy (ΔS) values for the interaction process were estimated (Velázquez-Campoy et al., 2004). Estimated errors in the determined values were 15% for the equilibrium constants, 5% for the binding enthalpy and entropy, and 2% for the binding Gibbs energy.

The ClusPro online server (Comeau et al., 2004, https://cluspro.bu.edu/queue.php) was used to perform the protein-protein docking. Model zero was selected out of the top ten models in the balanced order. The structural model was created and visualised using Chimera 1.11 (Pettersen et al., 2004) with colours applied to label the two proteins and their heme groups to facilitate visualization and interpretation.

Multiple sequence alignment was built using ClustalW using 152, 149 and 86 sequences identified as Cyt c ₆, Cyt c _6BC or Cyt c _6D, respectively ( Supplementary Table 1 ). Sequence identification was performed using the CyanoOmicsDB database (Zhou et al., 2021). Genome comparison analysis performed using the bioinformatics tools available on the online platform Biocyc (Karp et al., 2019; Karp et al., 2020). Phylogenetic tree was constructed by maximum likelihood method with Geneious version 2023.0 created by Biomatters (https://www.geneious.com), inferred using a neighbour-joining algorithm and Jukes-Cantor as the genetic distance model, using the Bootstrap (n = 100) method (Felsenstein, 1985).

Rates of respiration were assessed by measuring the consumption of O₂ over time in an Oxygraph O₂ electrode (Hansatech, Cambridge, UK) in a double-jacket thermoregulated glass vessel as described previously (Torrado et al., 2019). The reaction mixture contained, in a final volume of 1 mL, 5 mM of HEPES buffer (pH 7.5), 2.5 mM of NaCl, 2 mM of sodium ascorbate, COX-enriched membranes equivalent to 3-5 mg·mL^-1 of total protein and 20 µM of Cyt c ₆, Cyt c _6B or Cyt c _6D.

The phylogeny of Cyt c ₆-like proteins has been the subject of study for the past 15 years (Bialek et al., 2008; Reyes-Sosa et al., 2011; Bialek et al., 2014; Howe et al., 2016; Torrado et al., 2016; Torrado et al., 2017; Slater et al., 2021). Bialek et al. (2008) performed a comparative study of the sequence of Cyt c ₆ and Cyt c ₆-like proteins, concluding that Cyt c ₆-like proteins constitutes a new branch separated from Cyt c ₆. In addition, they divided them into two groups which they called Cyt c _6B and Cyt c _6C based on a phylogenetic analysis. A new analysis performed by Slater et al. (2021) including a higher number of sequences showed that the distinction between Cyt c _6B and Cyt c _6C can occur by taxon sampling rather than by differences in function. The similarity of crystal structures, surface electrostatic potential distribution, and midpoint redox potentials of proteins from both B and C subgroups, points to the fact that Cyt c _6B and Cyt c _6C could perform a similar function and are, probably, orthologs (Zatwarnicki et al., 2014; Slater et al., 2021). However, a group of Cyt c ₆ homologous that was first found in heterocyst-forming filamentous cyanobacteria, named as Cyt c _6-3 (Torrado et al., 2017), has not been integrated in the current phylogeny. We are aware that the nomenclature of these Cyt c ₆-like proteins can result confusing, but because the function of these Cyt c ₆-like proteins is not elucidated, a clear nomenclature cannot be established. For now, we propose to keep the former nomenclature, whereas the first Cyt c ₆-like protein found in cyanobacteria will be Cyt c _6BC and the second one, the formerly called Cyt c _6-3 which only appears in filamentous cyanobacteria, Cyt c _6D.

To investigate the distribution of Cyt c ₆ of cyanobacteria and to include the presence of this new group, we mapped a phylogenetic tree of cyanobacterial sequences of Cyt c ₆-like proteins including the new group of Cyt c _6D ( Figure 2A ). The phylogenetic analysis was based on 387 sequences of Cyt c ₆, Cyt c _6BC and Cyt c _6D found in cyanobacteria from public databases. The sequences were aligned using ClustalW method and a phylogenetic tree was inferred using maximum-likelihood method (see Material and Methods). The results ( Figure 2A ) showed that the newly classified Cyt c _6D forms an independent group clearly separated from Cyt c ₆ or Cyt c _6BC. In the previous publication in which Cyt c _6D was described (Torrado et al., 2017), where the genomes and tools were more limited, Cyt c _6D was found only in heterocyst-forming filamentous cyanobacteria. A more comprehensive study involving a higher number of Cyt c ₆ protein sequences has revealed that Cyt c _6D is present in all type of filamentous cyanobacteria, independently of the heterocyst formation, but it is not present in unicellular cyanobacteria. Besides, this sequence analysis has revealed a conserved sequence pattern that is present in all Cyt c _6D ( Figure 2B ). The conserved pattern has been identified as L-X-X-Y and starts at position 40 of the consensus sequence (most common amino acid at that position; Liljas, 2013) of Cyt c _6D. Also, in Cyt c ₆ and in Pc appears a single arginine residue at position 64 in Cyt c ₆ ( Figure 2B ) that is strictly conserved in both proteins and results crucial for their interaction with PS I (Molina-Heredia et al., 2001). This residue is also strictly conserved in the position 64 in Cyt c _6BC, but it has been replaced by lysine (position 68, black triangle in Figure 2B ) in Cyt c _6D. In Nostoc, we have previously reported that Cyt c _6BC can react with PS I, but in a less efficient manner than Cyt c ₆. (Reyes-Sosa et al., 2011). However, Cyt c _6D does not react with PS I (Torrado et al., 2017).

An interesting feature found in this analysis was the position and distribution of Cyt c _6D within cyanobacteria. We analysed the genomic context of the gene encoding Cyt c _6D in filamentous cyanobacteria, finding that the position and distribution of the gene in the genome seems to follow a specific pattern in most of the studied organisms ( Figure 3 , Supplementary Table 1 ). Gene encoding Cyt c _6D is located in a conserved region, whereas the genes psbV (Cyt c ₅₅₀), petE (Pc) and petJ-3 (Cyt c _6D) are located in the same order and direction of the expression ( Figure 3 ), in contraposition to the positions of petJ (Cyt c ₆) or petJ-2 (Cyt c _6B) which are unevenly distributed ( Supplementary Figure 1 ). Cyt c ₅₅₀ is a component of the PS II complex in cyanobacteria and some eukaryotic algae, such as red and brown algae (Roncel et al., 2012), and Pc acts as soluble electron carrier between Cyt b ₆ f complex and PS I. The fact that the gene that codes for Cyt c _6D is in the same cluster as the genes that code for Cyt c ₅₅₀ and Pc may lead us to think that Cyt c _6D function is related to the photosynthetic electron transport chain. However, this function should be different from Cyt c ₆ and Pc, because as we mentioned before, Cyt c _6D is not able to reduce PS I (Torrado et al., 2017). Another gene is frequently found downstream of Cyt c _6D, a hypothetical protein with partial similarity to TonB (BlastP), but no further information is provided in this study ( Figure 3 ).

We investigated whether Cyt c _6BC and Cyt c _6D can interact with the subunit Cyt f of Cyt b ₆ f complex. One of the key factors for this interaction is the redox potential of the Cyts. As we discussed before, the redox potential of Cyt c _6BC (+199 mV) and Cyt c _6D (+300 mV) is lower than that of Cyt c ₆ (+335 mV), which at first glance indicates that this interaction is unlikely to happen (Molina-Heredia et al., 1998; Reyes-Sosa et al., 2011; Torrado et al., 2017). However, in our previous study (Torrado et al., 2017), we analysed the redox potential under the physiological condition of thylakoidal lumen of pH 4, instead of the standard pH 7. We found that the redox potential of Cyt c _6D (+343 mV) became isopotential with Cyt c ₆ (+340 mV) under that condition. Still, Cyt c _6BC (+230 mV) was found to be far from the redox potential of the other Cyts. However, we could not measure the empirical redox potential of soluble Cyt f in vitro at pH 4, as this protein was denatured and degraded at that pH. We explain this phenomenon because physiologically, Cyt f should be embedded within Cyt b ₆ f membrane complex, but in the soluble form heterologously expressed, the integrity of this protein cannot withstand the acidic pH. In summary, with the redox potential found under those conditions, Cyt c _6D would be able to oxidise Cyt f while Cyt c _6BC will be unable to do it. However, the redox potential is not a sufficient condition to demonstrate that the interaction can happen between the proteins, since it only establishes the direction of the electron transfer in case they would interact. To evaluate if the proteins can interact in vitro, we performed Isothermal Titration Calorimetry (ITC) experiments and protein-protein docking analysis.

For ITC experiments we prepared purified samples of each soluble Cyts in an oxidised state, and samples with Cyt f completely reduced to facilitate the interaction, at 25 °C. In all the cases, we observed interaction between the proteins, with a 1:1 binding stoichiometry ( Figure 4A ). The analysis of the binding isotherms, according to a ligand-binding model with a single binding site, allowed us to determine the thermodynamic interaction parameters ( Figure 4B ). Both, K_A and K_D for the interaction of Cyt c _6BC with Cyt f are similar to those obtained with Cyt c ₆. However, K_A and K_D for the interaction of Cyt c _6D with Cyt f are 6-fold higher and lower, respectively, than those obtained with both Cyt c ₆ and Cyt c _6BC. These results show that Cyt c _6D has a higher affinity for Cyt f than Cyt c ₆. Looking at the thermodynamic parameters, although the ΔG values of the interaction of the three Cyts are somewhat similar, the values ​​of ΔH and –TΔS indicate that the three interactions are quite different. In the case of Cyt c ₆, whose isoelectric point (pI) is 9.0 (Molina-Heredia et al., 1998), both ΔH and –TΔS are negative, indicating that the interaction with Cyt f might be driven by attractive electrostatic interactions and by hydrophobic interactions. This agrees with the fact that Cyt c ₆ presents electrostatic and hydrophobic regions interacting with both PS I and Cyt f (Molina-Heredia et al., 1999; Crowley et al., 2002; Albarrán et al., 2005). In the case of Cyt c _6BC, ΔH and –TΔS are large and of opposite sign, cancelling partially each other. These values ​​indicate that, although an electrostatic repulsion may occur, the interaction is exergonic and might be driven primarily by hydrophobic interactions. This is in agreement with the data previously reported (Reyes-Sosa et al., 2011) that described the conserved hydrophobic interaction surface with Cyt f in Cyt c _6BC; but not the surface of electrostatic interaction, being positive in Cyt c ₆, but negative in Cyt c _6BC, even though it has a pI of 8.0, very close to that of Cyt c ₆. Likewise, in the case of Cyt c_{6 D}, whose pI is 5.2 (Torrado et al., 2017), the interaction with Cyt f is driven by hydrophobic forces, not presenting attractive electrostatic interactions. As in the case of Cyt c _6BC, Cyt c _6D preserves the hydrophobic surface for the interaction with Cyt f; however, the region for the electrostatic interaction that is present in Cyt c ₆ results practically neutral in Cyt c _6D (Torrado et al., 2017).

Besides the fact that the interaction can happen between Cyt c ₆ paralogous proteins and Cyt f, one of the key factors for a redox interaction involving cytochromes lays in the orientation of their heme groups (Cruz-Gallardo et al., 2012). The interaction may be happening but the heme groups might be so distant that the electron transfer would not be possible. To assess such orientation along the interaction, we performed a structural modelling by molecular docking of the soluble Cyts with Cyt f ( Figure 5 ). As a control, we used Cyt c ₆, considering that its interaction and orientation with Cyt f has been previously reported (Díaz-Moreno et al., 2005). The molecular docking was conducted using the Cluspro V.2.0 server, selecting the equilibrated models with minimum energy. According to these models, Cyt c _6BC and Cyt c _6D ( Figures 5B, C ) will interact with Cyt f through the same hydrophobic area as Cyt c ₆ ( Figure 5A ). Further, their heme groups will be oriented opposed to the Cyt f heme group, as it happens as well with Cyt c ₆. In conclusion, Cyt c _6BC and Cyt c _6D could interact with Cyt f in the same areas, with the same orientation as Cyt c ₆, and with a distance between the redox centres similar to that of Cyt c ₆, which would allow an electron transfer between them.

One of the key questions yet to be addressed is the role of these Cyt c ₆-like proteins in respiration. As we described before, Cyt c ₆ and Pc are the main interaction partners with terminal oxidases in respiration. However, little information has been provided about the interaction of Cyt c ₆-like proteins in respiration (Torrado et al., 2016). Nostoc sp. PCC 7119 is an heterocyst forming filamentous cyanobacterium that can express two different set of aa3-type copper oxidases. One of these oxidases is expressed in vegetative cells and is the main respiratory oxidase, also called COX1 (Valladares et al., 2014). In heterocyst, which are differentiated cells under nitrogen-deficient conditions, COX2, the alternative aa3-type oxidase, is expressed. Thus, under nitrogen-deficient conditions where vegetative cells and heterocysts cohabit in the filament, COX1 is expressed only in vegetative cells and COX2 is expressed only in heterocysts. As we described before, Cyt c _6D gene is only found in filamentous cyanobacteria, including heterocyst-forming cyanobacteria. Hence, our approach comprised the study of both terminal oxidases, with particular interest on the heterocyst-specific COX2 (Valladares et al., 2014). To address this question, we performed O₂ uptake experiments with membranes enriched in terminal oxidases in the dark ( Figure 6 ), using an oxygen electrode (Oxygraph, Hansatech). First, we evaluated the reactivity with the main oxidase that is expressed only in vegetative cells, COX1. In this condition, membranes enriched in COX1 were isolated and assessed against the three Cyts. As expected, Cyt c ₆ reacts with these membranes producing a high rate of oxygen consumption. However, neither Cyt c _6BC or Cyt c _6D produced any significant increase in oxygen consumption, indicating no reaction with COX1. Next, we purified heterocysts, and from these preparations, we isolated membranes enriched in COX2. In this condition, we found that Cyt c ₆ behaved as described previously, reacting with COX2 and producing an elevated oxygen consumption (Torrado et al., 2019). However, Cyt c _6BC was not able to react with these membranes enriched in COX2, observing only an unspecific oxygen consumption, at the level of a non-native Cyt c. Surprisingly, Cyt c _6D was able to react with membranes enriched in COX2 with a similar rate of oxygen consumption as Cyt c ₆. Thus, contrary to Cyt c ₆, Cyt c _6D reacts specifically with COX2.

These results could lead us to think that Cyt c _6D may transport electrons from Cyt b ₆ f complex to the COX2, specific to heterocysts. However, previous in vivo GFP report of the Cyt c _6D promoter activity experiments showed that in the presence of combined nitrogen, Cyt c _6D is homogeneously expressed in all the cells of the filament, and that under nitrogen starvation, the expression of Cyt c _6D was partially repressed in vegetative cells, but such repression was practically complete in heterocysts (Torrado et al., 2017).