RfA1, A2, and B1 possess different numbers and localization of two types of canonical reflectin motifs (see Figure 1A), the regular reflectin motifs {RMs [M/FD(X)5MD(X)5MDX3/4]}, and a N-terminal reflectin motif {RM_N [MEPMSRM-(T/S)MDF(H/Q)GR (Y/L) (I/M)DS (M/Q) (G/D)R (I/M)VDP (R/G)]} (see Figure 1B). Specifically, RfC is the shortest reflectin, containing a GMXX motif and RM*. The GMXX motif is a unique region of increased overall hydrophobicity composed of a four-amino acid repeat, and “X” represents less conserved locations within the repeat. Asterisk marked RM* of RfC contains substantial deviations in sequence not observed in any other reflectin motifs (Crookes et al., 2004) (see Figure 1B). The disorder tendencies of reflectins were calculated by PONDR (Xue et al., 2010), DISOclust (Ward et al., 2004; McGuffin, 2008), and ESpritz (Walsh et al., 2012). The possibility of forming protein droplets via liquid-liquid phase separation is also predicted by FuzDrop (Hardenberg et al., 2020). These calculations indicate a high disorder (see Supplementary Figure S1) and LLPS tendency (see Supplementary Figure S2) of four reflectin proteins. Reflectin proteins are introduced into the HEK-293T cells to verify these computational calculations.

After transfection and culturing for 12 h or longer, all RfA1, RfA2, RfB1, and RfC are phased out from the cytoplasm or nucleoplasm, which is consistent with computational predictions. However, their difference is more significant. For RfA1, protein condensates are exclusively enriched in the cytoplasm. In a sharp contrast, the perfect spherical RfC droplets are only located in the nucleoplasm. While the distribution of RfA2 and RfB1 seems to stay in a middle state, spherical protein droplets mainly exist in the nucleus, with the minority staying in the cytoplasm (see Figures 1C,E).

According to the report by Guan et al., reflectin genes in cephalopods come from a transposon in symbiotic Vibrio fischeri (Guan et al., 2017). The repetitions of conserved motifs in reflectin sequences are the results of transposon self-replication and translocation. Thus, it is speculated that diverse reflectin proteins with different motifs stay at different evolution stages or represent different evolution directions.

From this point, RfC with only a GMXX motif and RM* may be the earliest appeared ancestor, while RfA1 with a RM_N and five RMs is the further evolved offspring protein. Based on this consideration, truncated variants were constructed in the work by cutting off the RMs from the RfA1 one by one, to make their composition comparable to shorter natural reflectins. The dynamic intracellular performance of these RfA1 variants were then investigated.

Six pairs of primers were designed to clone RfA1 truncations (see Supplementary Table S1). The PCR products responsible for coding six peptides, including RM_Nto5, RM_Nto4, RM_Nto3, RM_Nto2, RM_Nto1, and RM_N (schematics shown in Figure 2A), were subsequently ligated to vector pEGFP-C1. After transfection, the expressions of GFP-tagged RfA1 truncations were validated by an anti-GFP Western blotting (Figure 2B).

Cells were harvested and RfA1-binding protein was extracted by an immunoprecipitation kit (Sangon) 24 h after transfection. GFP monoclonal antibody was used to validate the expression of GFP-tandem RfA1 truncations, while GADPH monoclonal antibody was employed to adjust the sample-loading quantity. In Figure 2B, the intensities of 36kD GADPH are almost the same among different lanes, implying that the sample loadings are normalized. The numbers labeled above bands are molecular weights calculated by ExPASy ProtParam software, consistent with the sums of GFP and RM_NtoX (X = 0, 1, 2, 3, 4, and 5). Moreover, for longer variants RM_Nto5, RM_Nto4, and RM_Nto3, there are significant smears with very large molecular weights. Meanwhile, the lanes of shorter variants are quite clean. It indicates that RfA1 truncations with more motifs tend to tightly interact with other cellular proteins, which is discussed in detail in the next section.

On the other hand, a confocal image system is employed to investigate the intracellular localization of GFP-tagged RfA1 truncations (see Figure 2C). The nuclei and membrane are stained with DAPI (blue) and DiD (red), and RfA1-derived peptides are indicated by tandem EGFP (green). For cells transfected with pEGFP-C1-RM_N and pEGFP-C1-RM_Nto1, though the peptide condensation seems absent, a remarkable enrichment of the green fluorescence signal in the nuclei areas appears. Amazingly, this nucleus-enrichment is executed more thoroughly by RM_Nto2, along with the recurrence of proteinaceous droplets. With increase in conserved motifs, RM_Nto3 are found to be distributed in both cyto- and nucleo-plasm. For longer derivatives with more motifs, condensates of RM_Nto4 and RM_Nto5 are blocked out from nuclei and located exclusively in cytoplasmic localization, extremely resembling their template RfA1.

Briefly, RM_Nto2 is most similar to RfC among all variants, which forms condensates exclusively in the nucleoplasm. More significantly, RM_Nto2 is also a definite boundary: truncations shorter than RM_Nto2 tend to enrich in the nuclei but cannot form droplets; derivatives longer than RM_Nto2 escape from the nuclei and accumulate and condensate into droplets in the cytoplasm.

Note smears in PAGE lanes of RM_Nto3, RM_Nto4, and RM_Nto5 in Figure 2B. Those are not sample contaminations or non-specific labeling. Subsequent CoIP-MS results reveal that various cellular components tightly interact with RfA1 variants, which assists their phase separation and diverse cellular localization.

A monoclonal antibody of GFP was prepared and a CoIP-MS approach was used to identify protein-protein interactions, thus exploring the mechanism underlying phase separation and distinguished distribution of RfA1 and RM_Nto2.

Aforementioned, there must be a massive interaction between RfA1 variants and host cell proteins, resulting in smears in Figure 2B. These smears are not meaningless miscellaneous proteins, but hybrid aggregates recruited by RfA1 derivatives. Therefore, valuable information would be lost if only one single band was collected and analyzed. However, the collection of multiple bands as a mixed sample will inevitably bring in disturbing data. Based on these considerations, cells that express RM_N are selected as a control group instead of GFP-expressed cells since RM_N presents no significant phase separation or selective distribution behavior. As shown in Figure 3A, the IgG and IP bands of RM_N, RM_Nto2, and RfA1 are collected and sent for mass-spectrum analysis. After that, proteins specifically interacting with RM_N, RM_Nto2,and RfA1 are screened with a simple mathematical logic: Unique (RM_N) = IP(RM_N)-IP(RM_N)∩IgG (RM_N); Unique (RM_Nto2) = IP(RM_Nto2) -IP(RM_Nto2)∩IgG (RM_Nto2); Unique (RfA1) = IP(RfA1)-IP(RfA1)∩IgG (RfA1) (see Figures 3B,C).

To further identify proteins which are responsible for intracellular localization or the formation of condensates, Unique (RM_N), Unique (RM_Nto2), and Unique (RfA1) are divided into subsets according to the venn diagram in Figure 4A. According to the venn diagram, overlaid sets Set4/6/7 are eliminated as backgrounds. Within Set1 are the proteins combined into whole-length RfA1 and contribute to the phase separation and cytoplasmic localization. Besides, proteins in Set3 promote the importation of RM_Nto2 into nuclei and its phase out from the nucleoplasm. Co-targets of RM_Nto2 and RfA1 are sorted into Set 5.

Gene Ontology (GO) annotation is then performed to investigate biological processes associated with interacting proteins of RfA1 and RM_Nto2. Figure 4C shows the items of interest with an adjusted p-value<0.05 (see Supplementary Table S2 for gene lists and Supplementary Figure S3 for PPI networks).

Taking Set1-i for example, since RfA1 binds to proteins like HNRNPA1 (Matthew Michael et al., 1995; Roy et al., 2014), SRSF1 (Hautbergue et al., 2017), and RAN (Connor et al., 2003; Güttler and Görlich, 2011) and participates in processes including “protein export from nucleus”, RfA1 condensates are exclusively distributed in the cytoplasm. Oppositely, RANBP2 (Zhang et al., 2010) and NUPs (Cautain et al., 2015) in Set3-v facilitate the importation of RN_Nto2 into nuclei.

One remarkable common point among Set1-ii, Set5-iii, and Set3-iv is the involvement of “ribonucleoprotein complex” or “ribonucleotide metabolic”. The interaction with ribonucleoprotein complexes brings in a massive interplay with proteins and mRNA (Sfakianos et al., 2016; Mukherjee et al., 2019), leading to almost inevitable phase separation. This is consistent with the results in Figure 2C. However, the involvement of the ribonucleoprotein complex makes the situation more elusive. It is equivocal whether RfA1/RN_Nto2 molecules gather together and are phased out from the surroundings on their own, or they are drawn into the formation processes of ribonucleoprotein complex bodies.

Moreover, PPI information also demonstrates that RfA1 extensively interacts with actin and actin-binding proteins (ABPs) (Supplementary Table S3 for gene lists and Supplementary Figure S4 for PPI networks). These proteins are in charge of various biological processes, including actin polymerization, nucleate assembly of new filaments, and promotion of elongation (Uribe and Jay, 2009; Pollard, 2016). Considering the critical role of actin filaments in forming protrusive structures such as lamellipodia and ruffles (Mejillano et al., 2004; Chhabra and Higgs, 2007; Weiner et al., 2007; Innocenti, 2018), the results presented here highly suggest RfA1’s potential contribution in the establishment and regulation of Bragg lamellae via its collaboration with the actin system. RfA1 can anchor at the inside of the Bragg lamella by cooperating with actin and ABPs, or even participate in the organization of the actin network, which regulates cell morphogenesis.

Actin should be indispensable for either the establishment or regulation of delicate Bragg reflectors in iridocytes. However, so far, except for the Co-IP results presented in this work, no evidence has been issued about the interaction between reflectin proteins and the cytoskeleton. As a preliminary start, we stained the HEK-293T cells with ActinRed 555 (Thermo) to trace the actin networks and verified the potential relationship between RfA1 and the cytoskeleton.

One notable trouble is the ubiquity of the actin cytoskeleton. It is challenging to judge the co-localization of RfA1 and actin (see Supplement Movie. 1mmc1). However, after viewing thousands of cells, an interesting scenario emerges. For plenty of transfected cells, there is a bright green sphere in the center or on the central axis, which are condensates of RfA1 (see Figure 5A for GFP channel, Supplementary Figure S5). These central localized protein condensates are not found in RfA2, RfB1, or RfC transfected cells. Around these spherical condensates, the red signal of stained actin is also significantly stronger than in other areas (see Figure 5A for ActinRed channel). The fluorescence statistics of selected areas in Figure 5B show an almost synchronized fluctuation of ActinRed and GFP signals, indicating their co-localization (see Supplementary Table S4 for raw data).

Except for the fluorescent observation of the actin system, RNA-seq results also strongly suggested the tight relationship between RfA1 and the cytoskeleton. Cells transfected with pEGFP-C1-RfA1 and no-load pEGFP-C1 (set as control) are harvested for sequencing. After Metascape screening (Zhou et al., 2019), 461 differentially expressed unigenes are identified in RfA1 expressed cells (Supplementary Table S5). GO analysis is then performed to annotate differentially expressed genes (DEGs). Figure 6A shows the selected items of interest. Genes evolved in biological processes such as “regulation of microtubule cytoskeleton organization,” “microtubule polymerization or depolymerization,” and “microtubule nucleation” respond to RfA1 expression. Correspondingly, these microtubule-relevant genes are also sorted into cellular component items, including “spindle microtubule,” “spindle pole centrosome,” “mitotic spindle,” and “spindle” (Wittmann et al., 2001; Foley and Kapoor, 2013), which indicates the tight relationship between RfA1 and microtubules.

The analysis based on RNA-Seq is further confirmed by confocal observation. After staining the fixed cells with Tubulin-Tracker Red (Beyotime), RfA1 condensates are significantly enriched at MTOC, located around the centrosome (see Figure 6B). For cells staying in late mitosis, RfA1 condensates are no longer located in the central areas but seem to chase the movements of daughter centrosomes (see Figure 6C and Supplementary Movies S3, S4 for Z-axis scanning). This significant enrichment of RfA1 at MOTC can be severely abolished by nocodazole treatment (data not shown). More directly, except for fluorescent colocalization observation, RfA1 is also detected to bind to tubulin subunits TBB2B, TBB4A, and TBB4B (Supplementary Table S6) by CoIP, suggesting physical interaction between RfA1 and microtubules.

The nucleotide sequence of D. (Loligo) pealeii reflectin A1 (RfA1) (Genbank: ACZ57764.1) and D. (Loligo) Opalescens reflectin C (Genbank: AIN36559.1) were optimized for human cell expression, and then synthesized and sequencing-identified by Sangon Biotech^® (Shanghai, China). Primers (F-GAATTCTAT GAA​TAG​ATA​TTT​GAA​TAG​ACA; R-GGATCCATACATATGATAATCATAATA ATTT) were designed to introduce EcoR I and BamH I cutting sites, so the modified RfA1 CDS can be constructed into pEGFP-C1 via a standard restriction-enzyme cloning process. As for the truncated RfA1 derivatives, six pairs of primers were coupled. For example, if RM_N-F and RM₃-R primers were selected, then a nucleotide sequence responsible for the coding of RM_N-RL₁-RM₁-RL₂-RM₂-RL₃-RM₃ (simplified as RM_Nto3 in the work) would be obtained after PCR. Meanwhile, 5′ GCA​TGG​ACG​AGC​TGT​ACA​AG 3′ and 5′ TTATGATCAG- TTATCTAGAT 3’ were added to F-primers and R-primers during the synthesis of the primer, respectively, which enables the sequences to be ligated to pEGFP-C1 by Ready-to-Use Seamless Cloning Kit from Sangon Biotech^® (Shanghai, China).

The HEK-293T cells (ATCC^®, CRL-3216^TM) were cultured on plastic dishes in Dulbecco’s Modified Eagle Medium (DMEM, Gibco^TM) supplemented with 10% fetal bovine serum (FBS, Gibco^TM) at 37°C under 5% CO₂. One day before transfection, cells were seeded at ∼33% of the confluent density for the glass bottom dishes from Cellvis (California, United States) and grown for another 24 h. Then a transfection mixture containing Lipofectamine 3,000 (Thermo Scientific) and recombinant vectors were added to the medium and incubated for ∼16–∼24 h. For CCK-8 test, 1 × 10⁴ cells were seeded into each hole of 96-well plates 1 day before transfection, then transfected with recombinant vectors and incubated for another 24 h. After that, 10 μl CCK-8 solution was added to the wells for a chromogenic reaction of ∼2–∼4 h. OD₄₅₀ was detected by Multiskan FC (Thermo Scientific).

Transfected HEK-293T cells grown in Cellvis plastic dishes were first fixed with 4% paraformaldehyde at room temperature for 30 min, then stained with DiD or ActinRed (diluted in 0.5% Triton X-100 PBS) for ∼30 min after PBS rinses. After washing off the fluorescent dye with PBS, fixed cells were embedded in DAPI-Fluoromount (Beyotime, Shanghai, China) and characterized with a Leica TCS SP8 imaging system in fluorescence imaging mode. The resulting images were analyzed with ImageJ (Java 1.8.0_172/1.52b) (Schindelin et al., 2012).

Disorder prediction of RfA1 and derivatives was conducted with PONDER^® (http://www.pondr.com/), DISOclust (http://www.reading.ac.uk/bioinf-/DISOclust/) and ESpritz (http://protein.bio.unipd.it/espritz/). The probability of liquid-liquid phase separation was calculated by FuzPred (http://protdyn-fuzpred.org/).

Total proteins of cells were extracted by using the immunoprecipitation kit (Sangon). Protein samples were separated by SDS-PAGE and transferred to polyvinylidene fluoride (PVDF) filter membranes (Millipore, United States) for immune-hybridization. After 1 h of blocking in PBST (phosphate-buffered saline containing 0.05% Tween-20 and 2.5% BSA), the membranes were incubated with one of the following primary antibodies at corresponding concentration: anti-GFP mouse monoclonal antibody (Sangon), and anti-GAPDH mouse monoclonal antibody (Sangon). Subsequently, band visualization was performed by electro-chemiluminescence (ECL) and detected by the Digit imaging system (Thermo, Japan).

For enzymolysis in protein gels (the gel strip, protein solution, and protein freeze-dried powder), a clean blade was used to dig out the strip of interest and cut it into 0.5–0.7 mm cubes. After decolorizing with test stain/silver stain decolorizing solution, the rubber block was washed with 500 μl acetonitrile solution three times until the rubber particles become completely white. Add 500 μl of 10 mM DTT to a water bath at 56°C for 30 min. Use low-speed centrifugation for a reduction reaction. Then add 500 μl decolorizing solution and mix at room temperature for 5–10 min Wash the gel spots once, and centrifuge at low speed to discard the supernatant After quickly adding 500 μl 55 mM IAM, place it in a dark room at room temperature for 30 min. Centrifuge at a low speed for an alkylation reaction. Then add 500 μl of decolorizing solution and mix at room temperature for 5–10 min After washing the gel spots once, centrifuge at low speed, and discard the supernatant; then add 500 μl of acetonitrile until the colloidal particles become completely white, and vacuum dry for 5 min. Add 0.01 μg/μl trypsin according to the gel volume. After an ice bath for 30 min, add an appropriate amount of 25 mM NH₄HCO₃ into pH8.0 enzymatic hydrolysis buffer, and enzymatically hydrolyze overnight at 37°C. After enzymolysis, add 300 μl extract with sonicate for 10 min. Centrifuge at a low speed, to collect the supernatant. After repeating the process twice, combine the obtained extracts and perform a vacuum dry.

Dissolve the sample with 10–20 μl 0.2% TFA, and centrifuge at 10,000 rpm for 20 min. Then wet the Ziptip 15 times with a wetting solution and equilibrate with a balance solution 10 times. Finally, inhale the sample solution for 10 cycles and blow 8 times with the rinse liquid. Note: To avoid cross-contamination, aliquot 100 μl flushing solution per tube and use a separate tube for each sample. Add 50 μl eluent to a clean EP tube, and pipette it repeatedly to elute the peptide. Finally, drain the sample.

The peptide samples were diluted to 1 μg/μl on the machine buffet. Set the sample volume to 5 μl and collect the scan mode for 60 min. Scan the peptides with a mass-to-charge ratio of 350–1200 in the sample. The mass spectrometry data was collected using the Triple TOF 5600 + LC/MS system (AB SCIEX, United States). The peptide samples were dissolved in 2% acetonitrile/0.1%formic acid and analyzed using the Triple TOF 5600 plus mass spectrometer coupled with the Eksigent nano LC system (AB SCIEX, United States). The peptide solution was added to the C18 capture column (3 μM, 350 μM × 0.5 mm, AB Sciex, United States) and the C18 analytical column (3 μM, 75 μM × 150) was applied with a 60 min time gradient at a flow rate of 300 nl/min mm(Welch Materials, Inc.) for gradient elution. The two mobile phases were buffer A (2% acetonitrile/0.1%formic acid/98% H₂O) and buffer B (98% acetonitrile/0.1% formic acid/2% H2O). For IDA (Information Dependent Acquisition), the MS spectrum was scanned with an anion accumulation time of 250 ms, and the MS spectrum of 30 precursor ions was acquired with an anion accumulation time of 50 ms. Collect MS1 spectrum in the range of 350–1200 m/z, and collect MS2 spectrum in the range of 100–1500 m/z. Set the precursor ion dynamic exclusion time to 15 s.

Genes identified from LC-MS and RNA-Seq were converted to ENTREZ ID and an R package, clusterProfiler, was used to perform gene ontology annotation (Yu et al., 2012). The mass spectrometry proteomics data have been deposited to the Proteome Xchange Consortium via the PRIDE (Perez-Riverol et al., 2022) partner repository with the dataset identifier PXD031350. RNA-Seq data were uploaded to the NCBI GEO database, numbered as GSE186861.

All the data needed to evaluate the conclusions are present in the paper and/or the supplementary information. All other relevant data are available from the authors upon reasonable request.