Photosystem I structural fusion with IsiA enhances light-harvesting capabilities of filamentous cyanobacterium during iron deficiency

SUMMARY

During iron stress, cyanobacteria produce iron-stress-inducible protein A (IsiA), which associates with photosystem I (PSI) to support cell growth. In most cyanobacteria, the stress induces formation of large IsiA₁₈-PSI₃ complexes. However, some species, including Anabaena sp. PCC 7120, studied in this work, form an unusual, smaller PSI-monomer-based IsiA-PSI supercomplex containing a unique fusion of IsiA with the PSI subunit PsaL, IsiA-PsaL. In this study, genetically tagged IsiA-PsaL was used to isolate several IsiA-PsaL-containing supercomplexes composed of multiple IsiA homologs. A combination of static and time-resolved optical spectroscopies was used to evaluate the energetic properties of gel-embedded supercomplexes, including the excitation energy transfer efficiency from IsiA-PsaL to PSI. These analyses demonstrated that excitation energy transfer in the IsiA-PsaL complex is highly efficient. Enhanced efficiency may be a key factor underlying the downsizing of the IsiA-PSI assembly, which could compensate for the smaller size of the complex.

Graphical Abstract

In brief

Biswas et al. describe the spectroscopic properties of gel-embedded atypical IsiA-photosystem I complexes from Anabaena sp. PCC 7120. These unique complexes contain an IsiA homolog with two domains, IsiA and PsaL, resulting in an enhanced excitation energy transfer to photosystem I under iron deficiency.

INTRODUCTION

Photosynthesis is the only naturally evolved process capable of harvesting solar energy at a planetary scale to drive chemical production. In oxygenic phototrophs, light energy is captured by light-harvesting pigment-protein antenna systems and funneled to reaction center (RC) complexes, photosystem I (PSI), and PSII, where it is converted into chemical potential.¹ The efficiency, robustness, and adaptability of this process are governed not only by the RCs themselves but also by the dynamic organization of their associated antenna proteins.² Photosynthetic electron transport is critically dependent on iron, and PSI contains at least twelve iron atoms.^3–7 Iron availability thus represents an important environmental constraint on photosynthetic energy conversion. Consequently, iron limitation disproportionately impacts PSI structure, function, and associated light-harvesting capacity, reshaping the photosynthetic apparatus.⁸

Iron is the fourth most abundant element in the Earth’s crust by mass, yet its bioavailability is severely limited in oxic, neutral-to-alkaline environments, where ferrous iron is rapidly oxidized to insoluble ferric oxides.^9–12 Such conditions dominate modern aquatic ecosystems, imposing chronic iron stress on free-living oxygenic phototrophs.¹¹ Cyanobacteria, which contribute substantially to global primary productivity, respond to iron limitation with pronounced remodeling of their photosynthetic machinery, including reduced PSI content and degradation of phycobilisome antennas.¹³ To compensate, many cyanobacteria synthesize a chlorophyll (Chl) a-binding protein, IsiA, which associates specifically with PSI.^12,14 IsiA functions as a stress-responsive antenna, partially restoring light capture capacity while simultaneously enhancing non-photochemical quenching to mitigate oxidative damage.^15–17 The long-term evolutionary retention of IsiA in cyanobacteria suggests that iron scarcity has been a persistent selective pressure throughout much of Earth’s history.¹⁸ Indeed, under iron starvation, cyanobacteria can accumulate IsiA to extraordinarily high levels,^17,19 so that IsiA represents one of the most abundant Chl-binding proteins in contemporary aquatic environments.

Structural studies have revealed remarkable diversity in IsiA-containing assemblies.^20,21 In the model cyanobacterium Synechocystis sp. PCC 6803 (hereafter Synechocystis 6803), IsiA most commonly forms an 18-subunit ring surrounding trimeric PSI,²² creating a modular supercomplex that enhances excitation energy transfer (EET) under iron stress. IsiA can also assemble into PSI-independent rings, including multiple concentric layers, which have been proposed to function in photoprotection or Chl a storage.²⁰ More recently, cryo-electron microscopy partially resolved an unusual IsiA-PSI architecture in the filamentous cyanobacterium Anabaena sp. PCC 7120 (hereafter Anabaena 7120), in which IsiA subunits are asymmetrically positioned on one side of a monomeric PSI.²³ Notably, one IsiA homolog in Anabaena 7120 is a fusion protein containing both IsiA and the PSI subunit PsaL, suggesting a fundamentally altered mode of antenna-RC integration. This architecture implies a stronger association between IsiA and PSI, with EET occurring over shorter, more homogeneous distances between them.

Adding to this complexity, Anabaena 7120 encodes four IsiA homologs arranged in a contiguous gene cluster (Figure 1). All four have been proposed to participate in the asymmetric IsiA-PSI supercomplex, yet their individual contributions, assembly pathways, and functional roles remain unresolved. Among these homologs, IsiA2 (also referred to as “IsiA-PsaL” in the main text and figures) is particularly distinctive: it contains a fused PsaL domain, replaces the canonical PsaL subunit of the PSI in the IsiA-PSI holocomplex, and represents the largest IsiA variant in the cluster. How such structural divergence influences excitation energy flow, photoprotection, and supercomplex stability under iron stress remains a central unanswered question.

Figure 1. Gene arrangement and nomenclature of IsiA protein-coding genes in Anabaena 7120.

Conventional gene nomenclature is shown in italics, and the corresponding protein nomenclature used in this manuscript is shown below the ORF (green arrow) in bold.

Here, we address this gap by selectively tagging IsiA-PsaL in Anabaena 7120 using CRISPR-Cpf1-mediated genome editing, enabling purification of native IsiA2-containing assemblies under iron-starved conditions. This approach, widely used in unicellular cyanobacteria^24–27 but previously unexplored in Anabaena 7120, allowed us to isolate several IsiA2-containing supercomplexes. Through integrated spectroscopic and proteomic analyses, we resolve the functional heterogeneity of IsiA2-containing complexes and elucidate how distinct IsiA assemblies modulate EET and quenching. Our results provide mechanistic insight into how photosynthetic systems reorganize under micronutrient limitation and reveal design principles for dynamically reconfigurable light-harvesting architectures—principles that may inform future efforts in sustainable energy conversion and bio-inspired photonic materials.

RESULTS

IsiA homologs in Anabaena 7120 are significantly different from each other

To establish the similarity and evolutionary relationship between the IsiA homologs from Anabaena 7120 and their occurrence in nature, we performed a BLAST search and phylogenetic analysis of the identified proteins. Figure 2 presents the analysis, which features protein sequence alignment, gene neighborhood, and phylogeny (Figures 2A–2C). It reveals the unusual IsiA-PsaL protein to be widespread across the cyanobacteria phylum. The protein sequence alignment comparing the IsiA protein from Synechocystis 6803 (IsiA-6803) with the IsiA proteins from Anabaena 7120 suggests that the histidine (His) residues providing ligands for Chl a binding are mostly conserved (black and red arrows, Figure 2A). In addition, IsiA1 and -2 have unique His residues that do not align with the His residues of IsiA-6803 (blue arrow, Figure 2A). In Synechocystis 6803, the isiA gene is located upstream of the isiB gene, which encodes the flavodoxin protein. Flavodoxin is known to replace the usual ferredoxin under iron-stress conditions. However, the gene context of the isiA-encoding gene cluster in Anabaena 7120 does not show the flavodoxin-encoding gene downstream of the isiA genes (Figure 2B). Moreover, the number of amino acids found in the IsiA proteins from different cyanobacteria shows significant variation, ranging from 283 to 924. BLAST search against IsiA2 (Figure 2B) identified 31 proteins (Table S1) among the six representative species. The identified list of proteins included the CP43 proteins from the six representative species. The CP43 and the IsiA protein sequences from Synechocystis 6803 were later added to the list for comparison. The compressed version of the phylogenetic tree (Figure 2C) shows that different IsiA homologs in Anabaena 7120 cluster into separate clades (see Figure S1 for the expanded version of the phylogenetic tree). Notably, IsiA3 clusters in the IsiA-6803 clade, suggesting functional similarity between the two proteins. The sharing of the same branch node between IsiA3 and IsiA4 suggests a common ancestor protein shared by the two homologs. The branch length (substitutions per site) suggests a faster evolutionary rate among the IsiA2–IsiA4 proteins. The most distal branches are observed for the clades containing IsiA/IsiA3 and IsiA1, indicating a significant divergence from the CP43 proteins. The results suggest modifications at the amino acid level to improve an organism’s adaptability under iron stress.

Figure 2. Sequence alignment, gene context, and phylogenetic relationship of IsiA homologs.

(A) Multiple polypeptide sequence alignment of IsiA-6803 and IsiA1–IsiA4 from Anabaena 7120. For IsiA2, only the region corresponding to the IsiA domain is shown. Arrows indicate specific His and Cys residues. Red arrows indicate a His residue present in all IsiA proteins considered for alignment. Blue arrows indicate a His residue in either IsiA1 or IsiA2 but not in IsiA-6803 or IsiA3. Black arrows indicate His residues in IsiA-6803 but absent in IsiA1 or IsiA2. Pink arrow indicates the Cys residue known to quench excitation energy in IsiA-6803. The olive arrow indicates the Cys residue of IsiA1 and IsiA2.

(B) Gene neighborhood of the IsiA2-coding gene in Anabaena 7120. The specific gene arrangement is also seen in six representative cyanobacteria species.

(C) Compressed version of the rooted phylogenetic tree showing the relationship between the IsiA homologs and the CP43 protein (see Figure S1; Table S1 for details).

SYNE, Synechocystis 6803; ANAB, Anabaena 7120; FD, flavodoxin; cpcG, phycobilisome linker protein. In (B), protein IDs corresponding to the genes are shown above the gene regions. In (C), the scale represents the branch length distance; the number at the nodes is the frequency. Grouping frequency greater than 60 is shown at the nodes.

PAGE and mass spectrometry show multiple IsiA2-containing supercomplexes with different IsiA protein composition

To aid subsequent analysis and protein purification, a 6×His tag was added to the C terminus of the IsiA-PsaL protein in Anabaena 7120. Using an anti-6×His antibody, the IsiA-PsaL protein was positively identified in the purified fractions. The apparent molecular weight of the IsiA-PsaL protein from Anabaena 7120 was estimated to be between 37 and 25 kDa (Figure S2A). The 6×His-tagged IsiA2-containing complexes were purified by affinity chromatography from iron-starved (approximately 30 days in no-iron YBG11 medium²⁸; 77 K fluorescence emission profile at the time of harvesting cells shown in Figure S2B) Anabaena 7120. Mass spectrometry (MS) analysis confirmed the presence of all four IsiA homologs (see Table S2 for unique peptides from IsiA homologs identified by MS). The purified fraction was further subjected to first-dimension separation by CN-PAGE (clear native-PAGE), yielding multiple green bands (Figures 3A and S3A). After separation of the green bands in the denaturing second dimension (Figure 3A), we identified three green bands that contained IsiA-PsaL. MS analysis of the three bands suggested varying amounts of the IsiA homologs and PSI proteins (Figure 3B). Based on the known structure of the IsiA-PsaL-PSI complex (PDB: 7Y3F) and the MS analysis from this work, band B1 is most probably the holocomplex, while bands B2 and B3 could be the intermediate IsiA2-containing complexes, though artifact complexes induced during sample preparation cannot be excluded. The protein abundance profile of different IsiAs in the B1 band suggests that in the IsiA2-containing holocomplex, IsiA3 is most abundant while IsiA4 is least abundant. Among the bands B1 and B2, a considerable variation is observed between the IsiA3 and IsiA4 proteins. In contrast, band B3, which is made up of two closely associated bands, shows a more substantial presence of the IsiA3 protein. It is possible that band B3 contains lower-order oligomeric species (probably dimers) of IsiA3, IsiA2, and IsiA1, which then assemble with PSI. IsiA1–IsiA3 form an initial complex and later incorporate the monomeric PSI by rearrangement of IsiA3 subunits under long-term iron starvation. The IsiA4 protein could be the last IsiA protein recruited to form the holocomplex. MS analysis also identified several other proteins associated with the bands B1–B3 (Figure S3B). The most abundant protein was All3041, which is a possible thylakoid membrane-bound protein of unknown function.^29,30 In addition, two chaperonin proteins, GroEL1 and GroEL2, were also identified.

Figure 3. Protein analysis of IsiA2-containing complexes.

(A) Representative image showing purified IsiA2-containing complexes separated in the first dimension using CN-PAGE in a 4%–16% polyacrylamide gel (n = 2), followed by separation in the second dimension using SDS-PAGE in a 4%–20% denaturing polyacrylamide gel. Approximately 600 ng Chl a-equivalent protein complex was loaded on the gel. Red arrows indicate the position of IsiA2 and other IsiA proteins.

(B) Heatmap displaying protein abundance values determined by LC-MS/MS analysis of gel bands separated by native PAGE. Native gel electrophoresis resolved multiple bands, with bands B1 and B3, each appearing as two closely positioned bands that were excised separately: B1–1 and B1–2 for band B1 and B3–1 and B3–2 for band B3. Color intensity represents relative abundance on a logarithmic scale (10⁵–10¹¹). UniProt accession numbers are shown on the right.

As shown in Figure 3, CN-PAGE-mediated separation followed by liquid chromatography-tandem MS (LC-MS/MS) analysis demonstrated that the abundance of IsiA3 (All4001/Q8YQ35) is dramatically reduced in B2 (33- and 15-fold lower than B1 and B3, respectively). The data may suggest that B2 could be an assembly intermediate. However, owing to the loose attachment of most IsiA homologs to PSI, as seen in PDB: 6NWA, B2 could also be a degradation product of B1. The question of B2 being an assembly intermediate or a degradation product cannot be ascertained definitively at this stage. However, spectroscopic analysis of the bands B1 and B2 does give us an insight into the spectral contributions from IsiA3 and IsiA4.

Using a previously standardized method for assessing the spectroscopic properties of Chl a-containing native complexes in polyacrylamide gel,³¹ spectroscopic examination of various IsiA2-containing complexes was undertaken. By incorporating advanced spectroscopic methods, such as time-resolved fluorescence (TRF) and transient absorption (TA), along with spectro-temporal analysis of data, we could evaluate pathways for the transfer of electronic excitations in IsiA2-containing complexes and hypothesize whether they provide any tangible benefits in the early stage of photosynthesis, warranting the development of such unique complex protein formations.

Absorption and fluorescence emission reveal spectroscopically different supercomplexes

Results from MS analysis of the protein content of B1–B3 bands, as seen in Figure 3B, highlighting the abundance of specific protein species, will have profound implications for the spectroscopic properties of protein assemblies present in those gel bands (Figure 4). The content of B3 is the simplest of the three and indicates the presence of oligomerized IsiA homologs and essentially no PSI. On the other hand, the most diverse is the protein supercomplex present in B1, indicating a considerable abundance of IsiA1–IsiA4 and PSI. Spectroscopically, the B2 band is likely an intermediate between B1 and B3, with PSI associated mainly with IsiA-PsaL.

Figure 4. Basic spectroscopic properties of pigment-protein complexes embedded in B1–B3 CN-PAGE gel cuts.

(A) Room temperature absorption spectra normalized for better comparability. The inset shows zoomed Chl a Q_y bands and maxima.

(B and C) Room temperature (B) and 77 K (C) fluorescence emission spectra of B1–B3 and PSI obtained after excitation at the Chl a Soret band. All spectra are normalized at the maximum for clarity.

(D) Absorption of bands B1–B3 recorded at 77 K and normalized at the Chl a Soret band. The inset shows the zoomed Chl a Q_y range and absorption of so-called red Chl a absorbing at 708 nm.

(E) 77 K absorption spectra of B1 and B2, normalized at the 708 nm peak exclusively associated with PSI. In the legend, in parentheses, the most probable IsiA-PSI arrangements contributing to the band absorption (deduced from spectral reconstruction) are indicated.

Considering those findings, the bands should have distinct spectroscopic characteristics with differences that should be enhanced at low temperature. Figure 4A displays absorption spectra of B1–B3 bands recorded at room temperature. For better comparison, the spectra were normalized at the maximum of the Chl a Q_y band. The profiles reveal patterns typical of photosynthetic proteins binding Chl a and carotenoids, though a closer inspection of the Q_y range (inset of Figure 4A) clearly shows subtle differences between protein assemblies, which support findings from MS. The most important feature is the hypsochromic shift of the Q_y band maximum from 676 nm for B2, through to 673 nm (B1), and finally to 670 nm for B3. Previous spectroscopic studies on IsiA₁₈-PSI₃ assembly from cyanobacterium Synechocystis 6803 revealed that such a spectral shift is associated with dissociation of PSI from the IsiA-PSI superstructure.^32–36 Therefore, the appearance of a Chl a Q_y band at 670 nm in the B3 spectrum confirms its assignment to PSI-free multi-IsiA assembly. On the other hand, the Chl a Q_y peak at 676 nm in the B2 spectrum indicates that the major contribution to the band originates from PSI. The B1 band is essentially the average between those peak values, indicating simultaneous contribution of PSI and a substantial amount of IsiA.

Those assumptions are further confirmed in fluorescence emission spectra. The emission spectra recorded at room temperature and 77 K are given in Figures 4B and 4C, respectively. For better clarity, all spectra were normalized at the maximum. At room temperature, the major emission contributors are Chl a in the IsiA antennas, and all bands show the same fluorescence maximum at 681 nm. However, notable differences are seen in the long-wavelength range, in which B1 and B2 have substantially higher intensity than B3. That difference, marked with an arrow, corresponds to fluorescence from PSI, and its spectral line shape can be estimated by proper spectral subtraction of B3 from either B1 or B2. The resulting spectrum of PSI fluorescence, marked as “predicted,” agrees well with its experimentally obtained counterparts from other cyanobacterial species available in the literature.³⁷ Even better clarity on PSI content in B1–B3 is seen if fluorescence emission is recorded after lowering the temperature to 77 K. It is known that due to temperature-induced energetic stabilization of so-called red Chl a present in PSI, its emission intensity multiplies and shifts to longer wavelengths. Without exception, it is also seen here, as B1 and B2 show a major band at 728 nm, associated with fluorescence of red Chl a, which almost completely masks a minor band at 685–688 nm, associated with emission mostly from IsiA antennas. On the other hand, B3 reveals an intense single, narrow band at 688 nm with only a small vibronic overtone at 750 nm, a spectrum that does not differ from the typical low-temperature emission of monomeric Chl a,³⁸ validating that B3 is free of PSI. Induction of strong, redshifted fluorescence by low temperature implicates that the absorption spectrum should also indicate the presence of a pool of energetically stabilized Chl a. Indeed, 77 K absorption spectra of B1–B3 shown in Figure 4D clearly resolve the long-wavelength band at 708 nm that is present only in the spectra of B1 and B2. Considering that the 708 nm band is an exclusive marker of PSI, 77 K absorption spectra of B1 and B2 could be normalized at its amplitude (assuming both IsiA-PsaL-PSI assemblies have the same quantity of PSI) and reconstructed to hypothetical intrinsic contributions of IsiA homologs and PSI. Based on that analysis (Figure S4), it was estimated that B1 corresponds to the IsiA_6–7-PSI₁ supercomplex stoichiometry and B2 to the IsiA_1–2-PSI₁ supercomplex stoichiometry (Figure 4E). Taken together, the spectroscopic and MS analyses of Anabaena’s IsiA-PSI holocomplex suggest that it may have a total of seven IsiA subunits surrounding the monomeric PSI, a mix of IsiA1, IsiA3, and IsiA4 along with IsiA-PsaL, in comparison to the total six resolved in the published IsiA-PsaL-PSI structure (PDB: 7Y3F).

TRF indicates efficient IsiA-to-PSI EET

Static spectroscopic methods, such as absorption and fluorescence emission measurements, do not provide crucial information on the energetics and dynamics of early photosynthetic events occurring in those IsiA-PSI super assemblies. Though fluorescence emission spectra give some clues on IsiA-to-PSI EET, e.g., a small IsiA band with respect to a large PSI peak in the 77 K spectra likely indicates efficient IsiA→PSI EET, they do not provide any details, such as temporal characteristics of the EET process. For that purpose, time-resolved techniques such as TRF decay mapping with application of a streak camera, often enriched with femtosecond (fs)-transient absorption (TA), are used. In principle, TRF can only be applied to fluorescing samples, and it is an excellent choice to study EET in IsiA-PSI assemblies, as previously demonstrated in equivalent studies of IsiA₁₈-PSI₃ from Synechocystis 6803 or Thermosynechococcus vulcanus.^32,35,39,40 The advantage of this method is that, as an excitation source, it utilizes an unamplified laser, and low laser intensity warrants annihilation-free excitation conditions, crucial for a correct assessment of dynamics of excitation transfer/decay in multi-pigment assemblies such as IsiA-PSI, which may contain up to 200 Chl a. Results from TRF studies of B1–B3 and accompanied spectro-temporal modeling of the TRF maps are shown in Figure 5. Exemplary raw TRF spectra and kinetic traces are shown in Figure S5.

Figure 5. Time-resolved fluorescence of IsiA and IsiA-PSI supercomplexes from B1–B3 bands.

(A) Collage of pseudo-color fluorescence decay maps of B1–B3 recorded at 77 K after excitation of the Chl a Soret band. To enhance visibility, the color gradient is shown on a log scale.

(B and C) Results of global analysis of fluorescence decay maps, SASs (species-associated spectra), with corresponding lifetimes. The fitting was done according to the kinetic scheme depicted in (C). A minor spectro-kinetic component associated with fluorescence of impurities in B2 is simultaneously excited with others and independently decays and was omitted in the kinetic scheme in (C) for clarity. EET (%), efficiency of excitation energy transfer.

Figure 5A presents TRF maps of B1–B3 recorded at 77 K after excitation at 410 nm (Chl a Soret). The color contours represent fluorescence intensity, which is displayed on a log scale to enhance visibility. Note that B1 and B2 reveal very intense fluorescence bands with a maximum at 728 nm, which represents emission from PSI decaying almost entirely within 5 ns after excitation. On the other hand, the emission from IsiA homologs, localized at ~688 nm, is barely noticeable (the log scale of fluorescence intensity substantially amplifies actual differences). Contrary to B1 and B2, B3 shows long-lived fluorescence emission typically observed for IsiA oligomers at 77 K.³² To fully utilize information “imprinted” into TRF maps, those need to be globally analyzed with an adequate spectro-temporal model mimicking the probable fates of laser-excited Chl a. Outcomes of such analysis are provided in Figure 5B. These were obtained based on the kinetic model depicted in Figure 5C. Briefly, it assumes that 410 nm excitation is not discriminatory; it promotes Chl a in IsiA homologs and in PSI. If there is only an IsiA homolog oligomer present in the sample (B3), excited Chl a will intrinsically recover to its ground state with a lifetime of 5.5 ns. This spectro-kinetic component is represented by olive SAS (species-associated spectrum) in Figure 5B. Otherwise, a 410 nm wavelength simultaneously excites the Chl a in IsiA and “bulk” Chl a in PSI. Both pools of excited pigments are quenched extremely fast (black SAS, Figure 5B) due to IsiA→PSI EET, trapping to the PSI RC, or to PSI-embedded long-wavelength Chl a (red Chl a). Consequently, the actual dynamics of the IsiA→PSI EET route cannot be reliably established due to temporal resolution limitations, though upper-level estimates of 56 (B2) or 71 (B1) ps indicate high efficiency of IsiA→PSI EET (see the formula in Figure 5B) in both B1 and B2. The final SAS (red line) represents fluorescence decay from red Chl a, essentially the same in PSI in both B1 and B2 IsiA-PSI assemblies.

TA reveals ultrafast nature of IsiA-PsaL→PSI EET process

Considering the very short timescale of the IsiA→PSI EET process, whose dynamics cannot be accurately determined via TRF, application of fs-transient absorption, TA, is a good approach to shine some light on this process. The TA setup typically uses an amplified laser as an excitation source, and consequently, the sample could experience laser intensity tens of thousands of times higher compared to the one used in TRF. Such disparity may have a profound consequence on excitation dynamics. The chance of multi-photon excitation within individual IsiA-PSI assemblies and subsequent excitation annihilation increases, particularly within clusters of red Chl a, particularly at 77 K due to preferential energetic trapping of excitation. Since TA and TRF measure the dynamics of the same processes but from two different perspectives, well-determined fluorescence decay dynamics, such as decay of red Chl a, could be used as a benchmark to tune experimental conditions of the TA measurements. Therefore, laser intensity was intended to be minimized to the level that recovery of bleaching of red Chl a absorption will closely match the dynamics of long-lived Chl a observed in TRF. However, it was impossible without dialing down signal quality below the level of usability. Consequently, in TA, the longest-lived Chl a species show somewhat shorter decay lifetimes compared to TRF counterparts, though they are still close enough to warrant comparison between excitation decay dynamics in B1–B3. Moreover, as the IsiA-PSI supercomplex is larger in size in B1 compared to B2, it has a larger absorption cross-section, and consequently, the decay lifetime of red Chl a in the B1 sample also has an increased probability of variation from the 670 ps lifetime seen in TRF, even if the same laser intensities were applied for B1 and B2.

Exemplary raw TA spectra and traces are shown in Figure S6, and an overall summary of TA findings is provided in Figure 6. Figure 6A is for illustrative purposes only and simplistically shows structural assemblies of studied protein supercomplexes associated with each gel band. Figure 6B presents schemes for customized fitting models (target analysis) for TA datasets, developed based on the initial global fitting of the TA data provided in Figure S7. Figure 6C shows a comparison of SADSs (species-associated difference spectra), spectral profiles resulting from target analysis of TA datasets. Those are, for reference, supplemented with the reversed and scaled absorption spectra of each band (shaded spectra). For dynamics and molecular origin of each SADS, according to color matching, refer to corresponding compartments in the kinetic scheme of the excitation relaxation (Figure 6B). The quality of target fits is shown for selected wavelengths in Figure S8. Given that samples contained IsiA oligomers and IsiA-PSI supercomplexes, by evaluating the initial simplified global fitting of TA datasets (Figure S7), it is possible to assign the compartments to Chl a embedded in either IsiA homologs or PSI, which are indicated by different background colors. However, it should be noted that 665 nm excitation, though it is preferential for direct excitation of Chl a in IsiA homologs, is not exclusive, as the absorption spectrum of PSI also spans this spectral range. This fact must be reflected in the fitting models of B1 and B2 TA, and for those it was assumed that laser flash simultaneously excites Chl a in both IsiA and PSI (Figure 6B). As expected, the simplest outcomes are for B3 (IsiA oligomer), for which the excitation at 665 nm populates the Chl a pool initially bleaching at 674 nm, and this band quickly evolves (with a lifetime of 1.16 ps) to a sharper band with a minimum at 684 nm. This process likely represents Chl a→Chl a energy transfer within the Chl a network to the energetically lowest Chl a in an individual IsiA complex. To distinguish it from inter-protein Chl a→Chl a excitation transfer, this step was abbreviated as Eq—the energetic equilibration of excitation within the same pigment-protein antenna. If there is no subsequent acceptor of excitation, the excitation is intrinsically dissipated, and Chl a recovers to the ground state. On the other hand, for B1 and B2, SADSs associated with this pool of Chl a would have faster dynamics due to the quenching effect of EET to PSI-embedded Chl a. Indeed, for B1, the lifetime shortens to 4.8 ps. However, an interesting scenario is seen for B2, for which it looks like those two SADSs associated with IsiAs merge into a single spectro-kinetic component decaying with ~900 fs, indicating that for IsiA-PsaL-PSI assembly, those two processes (excitation equilibration within IsiA and EET to PSI) are extremely fast and essentially cannot be temporally resolved. It could be linked to the structural difference between IsiA-PSI assemblies in B1 and B2. As indicated before, B2 contains maximally only 2 IsiA homologs: it must have IsiA2, fused with PSI through a shared PsaL subunit and, according to MS analysis, IsiA1, which statistically will be only in partial occupancy (Figure S4). As IsiA-PsaL-PSI is essentially a single protein with a fixed pigment network, the majority of excitation will be routed via the IsiA-PsaL-PSI bridge that is likely optimized for the most efficient EET. On the other hand, the IsiA-PSI supercomplex in B1 contains fused IsiA2 and 5–6 other type IsiA antennas, coordinated around the PSI monomer. The random possibility of excitation of Chl a in any of IsiAs implies that the majority of EET will be mediated via those classic IsiAs and due to fluid rather than the optimal distance between them and PSI-embedded acceptors, and the time constant of EET will be notably slower (4.82 ps). Once excitation enters the PSI Chl a network (Figure 6B, bulk Chls), there are a few scenarios for its fate. It could be trapped by the RC special pair or by red forms of Chl a, pigments whose excited states energetically lie below the absorption of the special pair, as indicated in the target analysis models (Figure 6B). Excitation trapped at red Chl a will initially energetically equilibrate, a process that, in the corresponding SADS (blue and magenta), is reflected by a few-nm bathochromic shift of the SADS minimum; subsequently, it will decay via either internal relaxation to the ground state or slow trapping to RC. The final long-lived SADS, whose lifetime, due to temporal limitations of the spectrometer, could not be defined, was assigned to bleaching of the absorption of the PSI special pair Chl a, most likely mixed with the electrochromic response of surrounding pigments.

Figure 6. Target analysis of TA datasets of IsiA oligomer and IsiA-PSI supercomplexes.

(A) Cartoon-like visualization of each studied supercomplex. IsiA, drawn with a dashed line, symbolizes partial occupancy.

(B) Kinetic models used for target analysis of TA datasets of each supercomplex. Note that for the IsiA-PSI supercomplex, it was assumed that 665 nm excitation simultaneously excites Chls a in both IsiA and PSI.

(C) SADSs (species-associated difference spectra) resulting from target analysis according to schemes given in (B). The shaded curves represent the inverted and scaled absorption spectra of each band and are added for reference. Note the color matching between SADSs and compartments in kinetic schemes.

DISCUSSION

Cyanobacteria, as progenitors of both unicellular and multicellular photosynthetic eukaryotes, have served as model organisms to study photosynthesis. However, unlike their eukaryotic cousins, cyanobacteria harbor proteins such as IsiA, which play an essential role in excitation energy management under iron deficiency. The most common IsiA₁₈-PSI₃ pigment-protein complexes have been a focus of researchers since their discovery in the 1980s.⁴¹ In recent years, cyanobacteria with more than one copy of IsiA-encoding genes have been documented. Moreover, the recent structure of IsiA-PSI from Anabaena 7120 has further complicated the stage by introducing a complex that contains a monomeric PsaL-lacking PSI and distinct IsiA homologs. A central question this study explores is how these associations affect the spectroscopic behavior of the IsiA-PSI complex in Anabaena 7120. The focus of this study is the IsiA2-containing complex; however, knowing the heterogeneity of IsiA complexes observed in other species, other types of IsiA-PSI complexes may accumulate depending upon the severity of iron deficiency.^42,43 Having different types of IsiA proteins possibly allows more flexibility in the antenna system to capture light energy that can be fine-tuned to the amount of available iron. Such flexibility, in addition to the transcriptional and post-transcriptional control mechanisms involving regulators such as Fur and NtcA proteins and antisense isrR microRNA,^44,45 could help save resources.

Attempts to determine EET pathways and associated time constants in IsiA-PSI supercomplexes from Anabaena 7120 are scarce, and those available were done only very recently and with a limited scope. For example, the original TRF experiments performed at 77 K by Nagao et al.⁴⁶ focused on thylakoids, not on the IsiA-PSI supercomplex per se, and more specific details were obtained in the subsequent work of the same group on the isolated IsiA₆-PSI₁.²³ From the application of standard global fitting of the TRF dataset, utilizing the sum of simultaneously excited and independently decaying components, it was postulated that the time constant of EET from the bulk of Chl a in the IsiA-PsaL subunit is 55 ps, while remaining classic IsiAs exhibit a time constant of 125 ps.²³ However, those assumptions have not been verified with the application of any alternative method, such as fs-TA, which would allow more selective excitation of Chl a bound in IsiA homologs and more precise tracking of the EET pathway due to much higher temporal resolution of that spectroscopic method. Moreover, if correctly interpreted, such time constants do not confer any advantages for cyanobacteria producing and accumulating atypical IsiA-PSI assemblies compared to more common ring-shaped IsiA₁₈-PSI₃ supercomplexes. It was demonstrated that the time constant of IsiA→PSI EET in IsiA₁₈-PSI₃ isolated from Synechocystis 6803 and Thermosynechococcus vulcanus ranges between 20 and 85 ps,^32,39,40,47 which is faster compared to the proposed time constant of 125 ps in atypical IsiA-PSI supercomplexes from Anabaena 7120.

Application of fs-TA on the IsiA-PSI, as performed in this work, allowed us to induce preferential excitation of IsiA and revealed, particularly for the IsiA-PsaL/PSI assembly (band B2), much better energetic coupling between this unique antenna and the PSI core complex than previously thought. Outcomes of simultaneous spectro-kinetic analysis of TA data indicate that the time constant of IsiA-PsaL→PSI EET is faster and appears on a sub-picosecond timescale, which is substantially faster compared to the 55 ps time constant suggested previously based on TRF studies.²³ However, it should be emphasized that considering the intrinsic lifetime of the excited singlet state of antenna-bound Chl a of a few nanoseconds, additional shortening of the time constant of antenna-to-PSI core EET, achieved in this atypical IsiA-PSI supercomplex, would not translate to a much higher overall EET efficiency of the antenna-to-PSI core EET process because that in IsiA₁₈-PSI₃ is already high, above 90%. On the other hand, the beneficial aspects of reducing antenna sizes are well known and even translate to the level of biomass production.^48,49 Therefore, from this perspective, implementation of an unusual, reduced-in-size IsiA-PSI supercomplex may give some physiological advantages over its betterk-nown, three-times-larger counterpart.

The current study provides insight into the different IsiA2-containing complexes amassing under extreme iron stress. A detailed spectroscopic investigation of the complexes suggests that the association of IsiA-PsaL with PSI gives some advantage via improved EET under iron stress. The results presented herein explored the advantage of covalently linking an antenna to the PSI RC. It should be emphasized that this is probably the only known example of such cooperation between the external antenna and photosystem. It could be reasoned that, under physiological conditions, the faster EET may become irrelevant; however, the resources saved by reducing the antenna size could give the organism an edge in the natural environment. Moreover, reducing the oligomeric state of PSI could make iron available to support other processes. Since the PsaL subunit is responsible for maintaining the oligomeric state of PSI⁵⁰ (a tetramer in Anabaena 7120), it would make sense to replace it with a fusion protein (IsiA2) to alter the oligomeric state. Among the six species with the IsiA-PsaL protein (Figures S1 and 2), five also have the NifH protein. It is possible that a change in the oligomeric state of PSI could be beneficial for nitrogen fixation by providing iron through PSI degradation. The overall organization of the thylakoid membrane under iron stress is also poorly understood. Whether the change in the oligomeric state of PSI provides a space advantage, serves as a strategy to make iron available for nitrogen fixation, or both warrants further targeted studies.

METHODS

Multiple protein sequence alignment, phylogenetic tree construction, and gene neighborhood

The alignment of protein sequences presented in Figure 2A was initially performed in MEGA12 using default parameters. Figure 2A was generated using BoxShade. The amino acid sequences were identified by doing a BLAST search against the IsiA2 protein sequence using NCBI-BLASTp. More than 100 sequences were identified in the BLAST search. The sequences were downloaded from the NCBI database between September 15 and September 30, 2025. Sequences from six representative species (Figure 2B) were selected to illustrate gene context using the Gene Graphics⁵¹ software. Protein IDs were used to download the gene context of the proteins showing high similarity to the IsiA2 protein from Anabaena 7120. The region size was set to 9,000 bp. Other unreviewed protein sequences (amino acid sequences) labeled as either “Chl a-binding protein” or “CP43” from the six representative species were also downloaded to construct the phylogenetic tree. The Mega12 software was used to build the maximum-likelihood phylogenetic tree using the cpREV+G+I+F model (best-performing model), with a bootstrap value set to 500. The number of discrete gamma categories was set to five, and the branch swap filter was set to strong. The gaps/missing data were set to use all sites.

Mutant construction

We employed the CRISPR-Cas12 mutagenesis system⁵² to introduce the 6×His coding sequence at the C-terminal end of the isiA2 open reading frame (ORF) in wild-type Anabaena 7120. Briefly, the backbone plasmid pSL2680 was digested with AarI to introduce the guide RNA, followed by the insertion of the repair template at the KpnI site (see Table S3 for primer details). The plasmid, pSL3680 (Figure S9), was sequenced before being introduced into wild-type Anabaena 7120 using the triparental conjugation method as described previously.⁵³ The E. coli strains, SL3680 (WM6026 E. coli strain with pRL623 and pSL3680) and SL3580 (HB101 E. coli strain with pRL443), were grown overnight and mixed with wild-type Anabaena 7120 (grown at 30°C under 50 μmol photons·m⁻²·s⁻¹ light intensity in BG11 medium) resuspended at an OD_730nm of 1.0. The E. coli cells were washed twice with sterile ultrapure water, while Anabaena 7120 was washed with BG11 medium. All centrifugation was performed at room temperature and at 4,000 × g. The mixture was incubated for 1 h at 30°C under low light in micro-centrifuge tubes (1.7 mL) and plated on an Immobilon-NC membrane (HATF08250, Millipore-Sigma, USA) on BG11-agar plates with 5% (v/v) LB (lysogeny broth). The plates were then incubated overnight at 30°C under continuous low light, after which the membranes were transferred onto BG11-agar plates supplemented with neomycin (40 μg·mL⁻¹). The conjugant colonies were repatched after 14 days onto fresh BG11-agar plates supplemented with neomycin (40 μg·mL⁻¹) weekly. The genomic regions upstream and downstream of the repair template were amplified using primers SeqP1 and SeqP2 (Table S3) and sequenced to confirm the correctness of the insert.

Strain maintenance and growth

The mutant cell line, IsiA2–6×his_7120, was maintained on BG11-agar plates at 30°C under 50 μmol photons·m⁻²·s⁻¹ continuous light. For liquid cultures, cells were scraped from the plates and transferred into 25 mL of low-iron BG11 (2 μM Fe) medium in a 125 mL Erlenmeyer flask. The flasks were continuously shaken on an orbital shaker set at 150 rpm at 30°C under 50 μmol photons·m⁻²·s⁻¹ continuous light supplied by an NS1 light fixture (Valoya, Finland). To grow a large volume of cells under iron deficiency, the low-iron-grown cells were gradually scaled up by transferring to a larger volume of low-iron BG11 medium every week until the total culture volume was 500 mL. After that, the cultures were harvested by centrifugation at 5,000 × g for 10 min, washed twice with YBG11–Fe²⁸ (no iron), and transferred to 1 L of YBG11–Fe medium. The cells were then routinely monitored by low-temperature fluorescence emission (77 K) every 3 days to monitor the IsiA-signature peak at 685 nm. The low-temperature fluorescence emission profile of the cells at the time of harvest for protein purification is shown in Figure S2B. The collected cells were stored at −80°C until further use.

Isolation of IsiA2-containing complexes

The IsiA2-containing protein complexes were isolated using the same method as described earlier for isolating CP43-containing complexes from Synechocystis 6803.⁵⁴ Briefly, Anabaena 7120 cells, grown under iron deficiency and stored at −80°C, were partially thawed and resuspended in resuspension buffer: 50 mM MES monohydrate (Thermo Scientific, USA; Cat No. J18886-A1)-NaOH (pH 6.0), 10 mM MgCl₂.6H₂O (Sigma-Aldrich, USA; Cat No. M0250), 5 mM CaCl₂.2H₂O (Sigma-Aldrich, USA; Cat No. C3881), and 25% w/v glycerol (Fisher Scientific, USA; Cat No. BP229–1). The cells were mixed in a 1:1 (v/v) ratio with Zirconia glass beads (0.1 mm) and broken by a bead beater (10 cycles of 30 s bead beating with a 2 min gap between cycles). The microfuge tubes containing cells and beads were constantly maintained on ice throughout bead beating. The remaining steps in the purification of IsiA2-containing complexes involved total membrane solubilization for 30 min in 0.8% β-DDM (n-dodecyl-β-D-maltopyranoside; Anatrace, USA; Cat No. D310) on ice, followed by centrifugation at 16,000 × g at 0°C and supernatant collection. The supernatant was immediately loaded on Ni-NTA Agarose (Qiagen, USA; Cat No. 30210) columns using the ÄKTA pure chromatography system and eluted with the resuspension buffer containing 0.04% β-DDM and 200 mM L-His (Sigma-Aldrich, USA; Cat No. H8000). The elutes were concentrated using a 50 kDa MWCO filter (Vivaspin2, Sartorius, UK), washed twice, and resuspended in resuspension buffer with 0.04% β-DDM. The resuspended purified fraction was stored at −80°C until further use. All reagents and buffers were made in deionized water.

Protein analysis

The Chl a content of the purified fraction was determined using the method described by McKinney.⁵⁵ The gradient gels used for CN-PAGE, SDS-PAGE, and 2D-PAGE were poured in-house as per the protocol described previously.^31,54 For immunoblot analysis (Figure S2A), the purified fraction was separated on a denaturing gel (refer to Figure S2A for details) and transferred to a PVDF membrane. The membranes were probed with anti-6×His antibody (Invitrogen, USA; Cat No. PA1–983B) at dilutions of 1:3,000. The blots were visualized by probing with an anti-rabbit HRP-conjugated secondary antibody (Sigma-Aldrich, USA; Cat No. A0545) using enhanced chemiluminescence. The purified IsiA2-containing complexes were separated on a 4%–16% in-house poured native gel; the green bands, B1–B3, were excised and analyzed by spectroscopy and MS. The western blots and clear native gels were visualized using either chemiluminescence or the 700 nm channel, respectively, of the LI-COR Odyssey XF imaging system.

In-gel digestion

Protein complexes separated by native gel electrophoresis were subjected to in-gel digestion for subsequent MS analysis. Individual protein bands were excised using a clean scalpel blade and cut into approximately 1 mm³ pieces, which were then transferred to clean 1.5 mL Eppendorf tubes. Gel pieces were destained with 250 μL of destaining buffer: 25% acetonitrile (ACN; Thermo Scientific, USA; Cat No. A956–1) in 25 mM ammonium bicarbonate (ABC; Sigma-Aldrich, USA; Cat No. 09830). The buffer was replaced every 10 min until the gel pieces were completely colorless. Following destaining, gel pieces were reduced with 200 mM DTT (Sigma-Aldrich, USA; Cat No. D9779) in 25 mM ABC buffer for 1 h at room temperature, followed by alkylation with 200 mM iodoacetamide (IAM; Sigma-Aldrich, USA; Cat No. I6125) in 25 mM ABC buffer for 1 h in the dark. The gel pieces were then washed with 200 μL of 25% ACN/25 mM ABC buffer for 5 min, followed by dehydration with 100 μL of 100% ACN for 5 min until the gel pieces turned completely white. After removing ACN, the gel pieces were dried completely using a SpeedVac concentrator. The dried gel pieces were rehydrated with 150 μL of 25 mM ABC buffer containing 1 μg of sequencing-grade modified trypsin (Promega, USA; Cat No. V5111) and incubated at 37°C for 16 h. Following digestion, peptides were extracted by adding 50 μL of 50% ACN/5% trifluoroacetic acid (TFA; Sigma-Aldrich, USA; Cat No. 302031) and sonicating in an ultrasonic water bath (10 s on and 10 s off, repeated 10 times). After centrifugation, the supernatant was collected into a new 1.5 mL Eppendorf tube. A second extraction was performed by adding 150 μL of 50% ACN/5% TFA to the gel pieces, and the supernatant was combined with the first extract. The pooled extracts were dried using a SpeedVac concentrator, reconstituted in 100 μL of 0.1% TFA, and desalted using Pierce C18 Tips (Thermo Scientific, USA; Cat No. 8774). Peptides were eluted in 50 μL of 75% ACN and 0.1% acetic acid (Sigma-Aldrich, USA; Cat No. 695092), lyophilized again, and resuspended in 15 μL of 5% ACN and 0.1% formic acid (Supelco, Millipore-Sigma, USA; Cat No. 1.59013) for LC-MS/MS analysis.

MS analysis

Nanoscale LC separation of the tryptic peptides was performed using a Dionex Ultimate 3000 Rapid Separation system equipped with a 75 μm × 25 cm Acclaim PepMap RSLC C₁₈ column (Thermo Scientific, USA), in combination with a 2 h linear 4%–24% ACN gradient in 0.1% formic acid and a flow rate of 250 nL/min. Eluted peptides were analyzed online by a Q-Exactive Plus spectrometer (Thermo Scientific, USA) in the positive electrospray ionization (ESI) mode. Full MS scans were performed in a mass range of 380–1,600 m/z at a resolution of 70,000, with an automatic gain control (AGC) target of 3 × 10⁶ and a maximum injection time of 200 ms. Data-dependent acquisition (DDA) was applied to fragment the top fifteen most intense peaks using high-energy collision-induced dissociation (HCD) at a normalized collision energy of 28. An intensity threshold of 4 × 10⁴ counts and an isolation window of 3.0 m/z were applied. Precursor ions with unassigned charges or charges of +1 to +8 or >+8 were excluded. MS/MS scans were collected at a resolution of 17,500, with an AGC target of 2 × 10⁵ and a maximum fill time of 300 ms. Dynamic exclusion was performed with a repeat count of 2 and an exclusion duration of 30 s, while the minimum MS ion count for triggering MS/MS was set to 4 × 10³ counts. Raw data and related processed files were deposited to the ProteomeXchange Consortium⁵⁶ via the PRIDE partner repository (PRIDE: PXD072813).

The resulting MS/MS spectra were analyzed using Proteome Discoverer (v.2.5, Thermo Scientific) against the Anabaena 7120 proteome database (https://www.uniprot.org/taxonomy/103690). Peptide assignments were performed using SEQUEST HT with the following parameters: trypsin digestion allowing a maximum of two missed cleavages, a minimum peptide length of six residues, a precursor mass tolerance of 10 ppm, and a fragment mass tolerance of 0.02 Da. Carbamidomethylation of Cys and oxidation of Met were specified as static and dynamic modifications, respectively. False discovery rates of 0.01 (high confidence) and 0.05 (medium confidence) were used to validate peptide spectral matches. Label-free quantification based on MS1 precursor ion intensity was performed in Proteome Discoverer with a minimum Quan value threshold set to 0.0001 for unique peptides; the “3 Top N” peptides were used for area calculation. All protein bands were analyzed in two technical replicates.

Proteomic data processing and protein quantification

MS raw data were processed using Proteome Discoverer 2.5 (Thermo Scientific, USA) for peptide identification and protein abundance calculation. Protein abundances were subsequently normalized against trypsin, which was spiked into each sample at equal amounts during sample preparation as an internal standard. Trypsin abundance showed excellent consistency across samples (coefficient of variation = 3.33%), confirming minimal technical variation and validating the normalization approach. To ensure high-confidence protein identification, stringent filtering criteria were applied: proteins identified by at least four peptide-spectrum matches (PSMs) were selected for downstream analysis and data visualization.

Static spectroscopy

All spectroscopic measurements were done on samples embedded in CN-PAGE gel from which they were cut off. Steady-state absorption spectra were recorded using a UV-visible (UV-vis) 1800 spectrophotometer (Shimadzu, Japan). For cryogenic measurements, gel cuts were placed between two microscopic coverslips with a small addition of glycerol and held in the liquid-nitrogen-vapor-based, temperature-controllable VNF-100 cryostat (Janis Research, USA). Such samples, upon lowering the temperature, froze into transparent glasses. Fluorescence emission of gel cuts at room temperature was obtained using a Shimadzu RF-6000 fluorometer (Shimadzu, Japan). The samples were excited at 440 nm, corresponding to the Soret band of Chl a. The bandwidths of excitation and emission slits were set to 5 nm. To minimize scattering effects, a 610 nm long-pass optical filter was placed at the detector entrance. Expected steady-state fluorescence emission spectra at 77 K were obtained by time integration of TRF emission contours obtained from a streak camera setup (vide infra).

TRF

Fluorescence decay maps were measured with a streak camera system (Hamamatsu, Japan), consisting of a universal C5680 streak camera and spectrograph from Bruker described in detail previously.³¹ It was coupled to the ultrafast-laser system based on an ultrafast Ti:sapphire laser (Mai-Tai, Spectra-Physics, USA), an ultrafast optical parametric oscillator (Inspire100, Spectra-Physics, USA), and a pulse selector (Model 3980, Spectra-Physics, USA), used to lower the frequency of the excitation beam to 8 MHz. The excitation beam was set to 410 nm (exciting B bands of Chl a), focused on a circular spot of <1 mm diameter, depolarized prior to the sample using an achromatic depolarizer (DPU-25, Thorlabs, USA), and set to low photon density of ~10¹⁰ photons/cm² per pulse that warranted annihilation-free excitation migration in large pigment arrays as of those in the IsiA-PSI supercomplex. To avoid light-scattering effects, a 610 nm long-pass filter was placed at the spectrograph entrance.

TA

Transient absorption (TA) experiments were carried out using Helios, a fs-TA spectrometer (Ultrafast Systems, USA) coupled to a fs laser system (Spectra-Physics, USA) based on Solstice, a one-box ultrafast amplifier (Spitfire Pro XP, a Ti:sapphire regenerative amplifier with a pulse stretcher and compressor; Mai-Tai, an fs oscillator, and Empower, a diode-pumped solid-state pulsed green laser). The laser system generates pulses at 800 nm with ~3.5 mJ energy, 1 kHz repetition rate, and ~90 fs duration. The output beam is split into 90%/10%. The 90% split was sent to Topas Prime, an optical parametric amplifier (Light Conversion, Lithuania), to generate pump pulses. The remaining 10% was used to produce probe pulses in the Helios spectrometer. A white light continuum (450–800 nm) probe was generated on a 3 mm sapphire crystal. A complementary metal-oxidesemiconductor (CMOS) linear sensor with 1,024 pixels was used as a detector in the Vis range. To provide an isotropic excitation of the sample and avoid pump-probe polarization effects, the pump beam was depolarized. TA was studied upon excitation at 665 nm, corresponding to the preferential excitation of Chl a bound in IsiA. The energy of the excitation beam (focused to a 1 mm diameter spot) was set to 20 nJ (695 nm), corresponding to 9 × 10¹² photons/cm² per pulse.

Global analysis of time-resolved spectroscopic data

Temporal chirp in the TA datasets was corrected using Surface Xplorer software (Ultrafast Systems, USA) by building a dispersion correction curve from a set of initial times of transient signals obtained from single-wavelength fits of representative kinetics. Both the TRF and TA datasets were analyzed using global analysis. Both TRF and TA datasets were first globally fitted by applying a fitting model that assumes an irreversible direction of excitation decay from fastest to slowest decaying states. According to this model, the transient signal at any time delay and wavelength, ΔA(t,λ) or Fl(t,λ), can be decomposed into a superposition of n^{th $C_i^S\left(t\right)·{EAS}_i\left(\lambda \right)$} products⁵⁷

$$\Delta A,Fl\left(t,\lambda \right)=\sum _{i=1}^n{C}_i^S\left(t\right){EAS}_i\left(\lambda \right),$$ (Equation1)

where C_i(t) is the time-dependent concentration of the i^th EAS (evolution-associated spectrum) defined by a kinetic model of sequential decay:

$$\frac{{dC}_i^S\left(t\right)}{dt}=k_{i-1}{C}_{i-1}^S\left(t\right)-k_i{C}_i^S\left(t\right),i\ne 1,k_{i-1}>k_i,$$ (Equation2)

and the first EAS concentration, C₁(t), is populated by the excitation pulse represented by the instrument response function (IRF) of the spectrometer:

$$\frac{{dC}_1^S\left(t\right)}{dt}=IRF\left(t\right)-k_1{C}_1^S\left(t\right).$$ (Equation3)

To distinguish global fitting results of TRF and TA datasets, the acronym EADS (evolution-associated difference spectrum) was used for TA. Subsequently, selected data were fitted with a customized kinetic model, commonly named as target analysis, which is meant to mimic the true pathway of excitation decay/relaxation/energy transfer processes. For that,

$$\Delta A,Fl\left(t,\lambda \right)=\sum _{i=1}^n{C}_i^t\left(t\right){SAS}_i\left(\lambda \right),$$

where $C_i^t\left(t\right)$ is the time-dependent concentration of the i^th SAS and reflects sometimes complex interactions between various SASs. Depending on the fitting model, at least one $C_i^t\left(t\right)$ is convoluted by IRF. To distinguish target analysis of TRF and TA, for the latter one, the acronym SADS was used. Note that in some simple situations, EA(D)S could be equal to SA(D)S.

Global analysis and target modeling were done using CarpetView software (Light Conversion, Lithuania). The IRF was mimicked by a Gaussian with a full width at half maximum (FWHM) of ~250 fs for TA datasets and 70 ps for TRF datasets. All plots were done in Origin 2026 (Origin Lab, USA).

RESOURCE AVAILABILITY

Lead contact

Requests for further information and resources should be directed to the lead contact, Prof. Himadri B. Pakrasi (pakrasi@wustl.edu), and will be fulfilled upon reasonable request.

Materials availability

This study did not generate new reagents. The plasmid generated in this study is available upon reasonable request, a completed Washington University in St Louis’s materials transfer agreement (MTA), and payment for transport of materials.

Data and code availability

• All relevant data are contained in the manuscript. Requests for details must be directed to the lead contact.

• Proteomics raw data and related processed files have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD072813. These are publicly accessible at https://www.ebi.ac.uk/pride/archive/projects/PXD072813.

• The paper does not report original code.

• Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Supplementary Material

Supplemental information can be found online at https://doi.org/10.1016/j.xcrp.2026.103524.

Highlights.

• Atypical IsiA-PSI complexes purified by affinity tagging IsiA2 in Anabaena 7120

• Different types of native complexes identified and subunit compositions defined

• Spectroscopic properties of gel-embedded native complexes

• Ultrafast antenna-to-photosystem excitation energy transfer at 77 K