Dictyosteliumpurpureum 4-NI-20 (wild-type) was used as a heterotrophic host cell and grown with Klebsiellaaerogenes on 3LP (0.3% lactose, 0.3% Bacto-peptone in 1.5% Bacto-agar). To create autotrophic cells from host Dp cells, a small number of Dp spores (5 μl in 20 mM phosphate buffer, pH 7.0) were placed at the center of inorganic agar on which Rhodobactersphaeroides had been uniformly spread. The red photosynthetic bacterial suspension of R. sphaeroides was purchased from the EM Laboratory (Shizuoka, Japan). The green photosynthetic bacterium, Synechocystis spp. PCC 6803, was kindly supplied in a culture solution by Dr. E. Suzuki (Akita Prefectural University). A cell suspension (10 ml) of R. sphaeroides was centrifuged for 8 min at 8000 g and 4˚C to precipitate, and the resulting pellet was suspended in ca. 250 μl of sterilized water. A cell suspension (10 ml) ofSynechocystis was centrifuged for 5 min at 5000 g and 4˚C to precipitate, and the resulting pellet was processed as for R. sphaeroides. This was followed by uniform plating of the dense cell suspensions on 1.5% inorganic agar (Bacto-agar dissolved in distilled water, BG-11 medium [GIBCO], or 20 mM Na/K-phosphate buffer, pH 7.0) in plastic Petri dishes (9 cm in diameter) and semi-drying by evaporation. Subsequently, an aliquot (5 μl) of Dp-cell suspension was placed at the center of the agar plate and incubated at 22˚C under about 12-h light (ca. 1450 lux on the agar surface) and about 12-h dark conditions, unless otherwise noted. During culture, the Dp cells engulfed the surrounding bacteria by phagocytosis and grew by spreading outward while expanding the forefront of growth. Photographs of cell morphology were taken under a stereo-, phase-contrast-, confocal or fluorescence microscopes.

In order to detect the existence of cell walls, cells were washed twice suspended in 20 mM phosphate buffer (pH 7.0), and placed on a clean glass slide. Subsequently, one drop (100 µL) of Calcofluor White Staining Solution was added to the sample. After 20 min of incubation in a moisture chamber at room temperature, the slide was covered with a clean coverslip, and observed under a fluorescence microscope using UV light.

Various types of cells and cell masses were fixed in 1% solution of OsO₄ dissolved in the standard salt solution (BSS) for 20 min at room temperature, as described before [20]. In another fixation, cells and cell masses were first fixed in 2% solution of glutaraldehyde dissolved in 20 mM PIPES buffer (pH 7.0) for one hr at room temperature and post-fixed in 1% solution of OsO₄ as described above after careful washing in 20 mM PIPES buffer (pH 7.0). After dehydration of the fixed samples in a series of ethanol, specimens were embedded in Epoxy resin. Ultrathin sections (60 - 80 nm thick) were cut with a Reichert Ultracut S. The sections were stained with lead citrate and were observed with a Hitachi H-500.

Cell masses grown on 1.5% non-nutrient agar were scraped off from the agar surface by a bacterial spreader, and suspended in 1 ml of 20 mM Na/K-phosphate buffer, pH 7.0, at a density of ca. 1 × 10⁶ cells/ml. Then, 5 µl of 1 µM Mitotracker stock solution in DMSO was added to 100 μl of the cell suspension, and vortexed vigorously. The cell suspension was placed on a cover-slip. After incubation of 30 min in a wet-chamber at 22˚C, the sample was washed three times with 20 mM Na/K-phosphate buffer, pH 7.0, dipped in ice-cold absolute methanol, and fixed for 10 min. Subsequently, 100 µl of DAPI staining solution (10 µg/ml DAPI in 20 mM Na/K-phosphate buffer, pH 7.0) was added, and incubated for further 3 hrs at 22˚C. The cover-slip was washed three times with the Na/K-phosphate buffer, pH 7.0, and observed under a fluorescent microscope (Zeiss, Switzerland, Model Axioscope A1).

Living Dp-green cell masses and host D. purpureum cells (as control cells) were separately placed on glass base dishes (Iwaki) and observed under a confocal microscope (Fluoview FV 1000, Olympus, using an excitation wavelength (440 nm) and an emission wavelength (620 nm) to detect chlorophyll fluorescence. Differential interference contrast (DIC) images of the respective cells were photographed under a DIC microscope.

Vegetative cells of Dp and D. discoideum Ax-2 were harvested, placed into 20 mM Na/K-phosphate buffer, pH 7.0, and washed by three centrifugations. The Dp-green cell masses incubated on 1.5% inorganic agar were also collected and washed in the buffer. Genomic DNA was isolated from the washed cells with the DNeasy Tissue Kit (Qiagen, Germany). PCR was performed by following either the protocol used for the amplification of partial sequences of 5.8S rRNA of Dictyostelids [20] or that of the Dd-MRP4 gene whole sequence in D. discoideum [21]. The DNA sequences of primers used for Dictyosteliumdiscoideum mitochondrial protein S4 (Dd-mrp4; MtS4) and 5.8S rRNA (Dic_5.8S_rRNA) are as follows:

Dic_5.8S_rRNA_F: GAGGAAGGAGAAGTCGTAACAAGGTATC

Dic_5.8S_rRNA_R: GCTTACTGATATGCTTAAGTTCAGCGGG

MtS4_FH3: TAGGATCCATGAGACAACGAAAAAATGTGACAAAATTT

MtS4_RB1: CGAAGCTTTTATCTTAGTCTTTTATATTTCTTTAATAAAG

PCR cycle conditions are as follows: denaturing step at 94˚C for 5 min, followed by 30 cycles of 94˚C for 1 min, 55˚C for 1 min and 68˚C for 1 min.

Dp-green cells were grown on 1.5% 1/5 SM agar plates, feeding with K. aerogenes. About 1 × 10⁷ cells were collected and washed in 20 mM Na/K-phosphate buffer, pH 7.0. Genomic DNA was isolated from washed cells with DNeasy Tissue Kit (Qiagen, Germany). Isolated genomic DNA was analyzed by MiSeq Microbiological Genome Draft Analysis Service provided by FASMAC, Japan. Retrieved genomic scaffold data were analyzed with Mole-Blast (NCBI, NIH, USA).

When wild-type Dp amoeboid cells were incubated with K. aerogenes as a food source on semi-nutrient (3LP) agar plates, they proliferated rapidly and phagocytosed the bacteria (Figure 1(A)). Upon exhaustion of the bacteria, the starving cells differentiated to acquire aggregation-competence and constructed multicellular structures. The cell mass migrated to form a cellular stalk at the tip (Figure 1(B)) and, finally, differentiated into a fruiting body with a mass of spores at the top.

As the first step to create autotrophic cells from host Dp cells, a small number of Dp spores were incubated with R. sphaeroides (see Materials and Methods section for detail). Henceforth, the first inoculated

cells are referred to as the 1st transferred cells (1st Dp cells) for convenience. During incubation, the 1st Dp spores germinated, actively engulfed as amoeboid cells the surrounding bacteria by phagocytosis and multiplied and spread outward while expanding the forefront of growth, which reached the outermost region of the agar plate (diameter, 9 cm) within 5 days of incubation at 22˚C. Starving cells located at and near the inoculation spot exhibited normal morphogenesis to form fruiting bodies. However, the cells located further away failed to aggregate and formed sheet-like cell clusters, as shown in Figure 1(C), consisting of somewhat rounded cells.

The surface of each rounded cell was stained with calcofluor (Figure 1(D)), which that binds to cellulose and/or chitin in the cell wall, and was considered a microcyst, which has not been reported before in D. purpureum. After 7 days of incubation, the outermost cells were transferred in droplets to the center of inorganic agar on which red photosynthetic R. sphaeroides had been uniformly spread, and incubated for another 7 days as 2nd Dp cells. The cells at the outermost region should have been those that had divided the most during the course of the incubation, while phagocytosing the bacteria. As was expected, the 2nd Dp cells grew and divided while actively phagocytosing the bacteria, and most did not exert normal morphogenesis after starvation but formed clusters of microcysts instead of autotrophic cells. The 3rd Dp cells were found to lose their morphogenetic ability after starvation, although their vegetative growth and proliferation in the presence of bacteria were almost normal. Despite this situation, we tried to obtain autotrophic cells by repeated transfer of the Dp cells. Unfortunately, however, no signs of autotrophy were recognized, even in the 21st Dp cells incubated for ca. 5 months.

In the second attempt, we forced the 21st Dp cells to ingest another species of photosynthetic bacteria, the cyanobacterium Synechocystisspp. PCC 6803. When the 21st Dp cells were transferred to the center of inorganic agar (1.5% agar containing BG-11 medium or 20 mM Na/K-phosphate buffer, pH 7.0) on which Synechocystis had been uniformly spread, they grew slowly while feeding on the bacteria, and the forefront of growth reached the outermost region of the 9-cm agar plate after 7 - 8 days of incubation at 22˚C. The outermost cells were again spot-inoculated onto Synechocystis-spread agar, as in the case of R. sphaeroides, and incubated for 7 days. Although autotrophic cells did not appear in the 2nd transferred population, notable results were obtained by chance when the 3rd transferred cells were incubated and abandoned for 53 days: a considerable number of dark-brown cell masses formed near or at the center of the agar plate, as shown in Figure 2(A). From their morphological and behavioral characteristics (Figure 2(B)), they seemed to be formed by autotrophic growth, and this was proven in the next stage of the study.

Green-colored autotrophic cell masses had not been obtained by the repeated culture of host Dp cells with Synechocystis alone for about 8 months. Instead, Dp cells cultured with Synechocystis rapidly lost their morphogenetic ability to form fruiting bodies after starvation and became microcysts, as in the case of the culture with R. sphaeroides.

To confirm that the formation of the dark-brown cell masses described above was due to autotrophic growth through photosynthesis, a part of the mass was aseptically transferred to inorganic agar (typically, 1.5% agar dissolved in BG-11 medium or 10 mM Na/K-phosphate buffer, pH 7.0), using a sterilized toothpick or platinum loop and were incubated without a food source like bacteria. This replanting of cells was performed aseptically under a stereomicroscope on a clean bench. As a result, a considerable number of dark-brown regions with a unique distribution pattern were formed within 10 days of incubation (Figure 2(C)). That is, in addition to a stripe pattern with an underlining pattern of small dots (Figure 2(C), left side), a number of globular dark-brown regions formed as small islands (Figure 2(C), right side). At the earliest developmental stage, many amoeboid cells were produced around the spherical dark-brown mass, as shown in Figure 2(D). Free-living amoeboid cells, some of which were in the process of cell division, were observed on the exterior of the cell masses. When observed in detail, the central dark-brown mass and the surrounding dark-brown islands were found to be connected to each other by a thin multinucleate plasmodium that spread widely and radially as a very thin layer from the center. The multinucleate plasmodium was found to be critical for autotrophic growth, as described later. Importantly, a significant number of amoeboid cells were constructed in a budding-manner from the thin multinucleate plasmodium. Incidentally, we found that the first dark-brown cell masses could live for a long time, as they retained the ability to form the autotrophic organism, even when left unattended for at least 3 years at 22˚C.

During successive planting of the dark-brown cell masses without a food source, some of the cell masses spontaneously transformed into green-colored masses (referred to as Dp-green cell masses for convenience), as shown in Figure 3(A). Many amoeboid cells were produced from the masses (Figure 3(B)). The Dp-green cell mass also formed a multinucleate plasmodium that spread radially outward from the center of the Dp-green mass, as the case in the dark-brown cell mass. A significant number of amoeboid cells were constructed from the tip (Figure 3(C), Figure 3(D), Figure S2(B)) or the intermediate region of a green multinucleate plasmodium (Figure 3(B), Figure S1), possibly via the self-organization of Dp-derived nuclei, mitochondria, red pigments (probably derived from Rhodobacter), and green pigments (possibly derived from Synechocystis) that were present as granular or lamella structures (Figure S1). A thin reticulated multinucleate plasmodium formed at the outermost peripheral region, and apparently cell-like protrusions sometimes formed in a process similar to budding (Figure 3(C), Figure S2(B)). As a result, many amoeboid cells were produced and migrated over the substratum from the outermost area of the plasmodium (Figure 3(D), Figure S2(B)). Importantly, we found that the outermost region of the green plasmodium, apparently without amoeboid cells, had the ability to generate Dp-green cell masses when transferred to a new inorganic agar plate and cultivated without a food source. In the nuclear regions of Dp-green cells, red and green pigments derived from RhodobacterandSynechocystis,respectively, were often observed, as shown in Figure S1.

Cells that could not become green cell masses gradually become light-brown cell masses in which most of Synechocystis-derived green granules were missing. At the earliest developmental stage, the globular cell mass formed a cluster of protrusions with the appearance of small plant roots. Subsequently, a thin multinucleate plasmodium was formed, and some amoeboid cells were produced and released from the plasmodium, as shown in Figure S2. Interestingly, formation of the light-brown cell masses was formed even in the dark, and its property was stably retained in the process of further subculture. The highest growth rate of the green plasmodium was about three times higher than that of light-brown plasmodium: the rate of radial horizontal spread of the green plasmodium on agar was ca.15 μm/hr, while that of the light-brown plasmodium was ca. 5 μm/hr.

When a small number of Dp-green or Dp-light-brown cell masses were transferred in droplets to the center of inorganic agar, on which R. sphaeroides orSynechocystis had been uniformly spread, and incubated for another 7 days, both of the cells failed to ingest temporally the microbesby phagocytosis for about two days and exerted autotrophic growth during the lag period. This was followed by heterotrophic growth, phagocytosing actively the external bacteria. When agar plates were spread with Klebsiellaaerogenes, both the Dp-brown and Dp-green masses lost their ability to phagocytose for about 2 days of incubation: they exhibited autotrophic growth during this period and then began to proliferate, engulfing the K. aerogenes. This seems to indicate that once autotrophic traits were acquired, the cells did not need to feed on bacteria for their growth temporally. Importantly, however, the Dp-green masses that had proliferated in a heterotrophic pattern while engulfing Synechocystis after about 4 days of lag time eventually formed a sheet of microcyst-like cells after about 15 days of incubation. When a small number of the microcyst-like cells were transferred to a new inorganic medium and incubated without food, they grew autotrophically again. Incidentally, the lag-time of the first 4 days was temporally allowed only shortly after transfer from autotroph to heterotroph. Actually, the lag period time was no longer allowed when the heterotrophically grown Dp-green masses were transferred to a new plate on which K. aerogenes had been spread. Taken together, these results indicate that the Dp-green masses could grow heterotrophically, as well as autotrophically, and so retained their basic genetic features required for autotrophic growth.

In the central region of a green multinucleate plasmodium, Synechocystis-derived bodies were predominantly present, and other structures such as Rhodobacter-derived bodies and Dp-derived mitochondria were also observed (Figure S3 and Figure 4 shows fine structures of the middle region of a green multinucleate plasmodium, in which a variety of organelle-like structures, includingSynechocystis-derived bodies,Rhodobacter-derived bodies, and a cluster of structurally modified mitochondria were mainly visible. Surprisingly, however, it was found that the cell membrane of the multinucleate plasmodium and amoeboid cells was not retained by double fixation with glutaraldehyde and OsO₄ as well as by single fixation with OsO₄. Probably, the membrane structures of the created autotrophic plasmodium and amoeboid cells derived from it might be very fragile, thus resulting in failure of preservation of cell membrane systems during fixation and subsequent dehydration. The absence of small ribosome-like granules might be because these structures might be leaked from the plasmodium and amoeboid cells after the cell membrane was greatly damaged. Nevertheless, electron-dense membranes as observed in structurally modified Synechocystis, and structurally modified mitochondria were observed in the cytoplasm of the green plasmodium and amoeboid cells (Figure 4, Figure S4. The cell membrane structure was also scarcely noticed in the light brown plasmodium and amoeboid cells derived from it. Instead, a cluster of unique mitochondria was often observed in the light-brown plasmodium and amoeboid cells as well as in the green plasmodium and amoeboid cells. Interestingly, most of the modified mitochondria were found to be covered

with electron-dense membranes (Figure S4). The mitochondria seemed to correspond to those of host D. purpureum cells. In the case of host Dp-amoeboid cells, at least two kinds of mitochondria with different electron densities were observed, as shown in Figure 5. Such heterogeneity is rarely found in the population of D. discoideum-mitochondria [22]. In the cytoplasm of Dp-amoeboid cells shown in Figure 5, a considerable number of cavities without boundary membranes as observed in Golgi complexes and small vesicles were recognized. In addition, there is a limited part of cell membrane where the unit membrane structure is not so clear. These structural features are not found at least in D, discoideum cells [22]. Accordingly, it is possible that such vulnerabilities of the membranes in host Dp-cells may be passed on to the green plasmodium and amoeboid cells derived from it. The red and/or green pigments, as shown by the phase-contrast micrograph of Figure S1, seemed to correspond to the unique inclusion bodies present in the nuclear region of the plasmodium-derived amoeboid cells (see the EM image of Figure S4).

The green plasmodium-derived amoeboid cells were stained with a mitochondria-specific fluorescent marker, MitoTracker Green, and the DNA-specific dye, DAPI, to examine the distribution of mitochondria and DNA-containing organelles, including cell nuclei, respectively. In a differential interference contrast micrograph of an amoeboid cell shown in Figure 6(A), an almost circular and non-granular area was

observed. From its shape, this area is regarded as the nuclear region. As expected from the electron microscopic observations, however, the cell nuclear regions were not stained with DAPI (Figure 6(B)), possibly because of leakage of DNA from the nucleus during fixation. On the other hand, the number of stained mitochondria was found to be much higher than that in parentalD. purpureum cells (Figure 6(C), Figure S6). Figure 6(D) shows the merged image of DAPI-staining and MitoTracker Green-staining. The original green-colored granular stains were artificially converted to red in Figure 6(C) in order to merge with the DAPI-staining in Figure 6(B). Numerous pink granules with mitochondria with DNA were observed, in addition to and a small number of blue granules containing probably Synechocystis- and/or Rhodobacter-derived DNAs (Figure 6(D)). In (D), three distinct yellow granules were observed near the nuclear region, beside the pink and blue granules. Since yellow is a merged color of green and red, there are the possibility that a small number of green-colored granules will be present in the amoeboid cells. For comparison, the staining patterns of host D. purpureum cells with DAPI and MitoTracker-Green are shown in Figure S5.

To know if the green plasmodium and its derived amoeboid cells have fluorescence with chlorophyll, they were observed under a confocal microscope, using excitation wavelength of 440 nm and emission wavelength of 620 nm (Figure 7(A), Figure 7(B)). As a result, it was found that the plasmodium contains a significant number of granules with chlorophyll fluorescence (Figure 7(B), Figure 7(C)), while that many of the amoeboid cells have no chlorophyll fluorescence-positive granules in the cytoplasm with an exception that a small number of cells have one or two red granules with chlorophyll fluorescence (Figure 7(C), arrow). As expected, host Dp-amoeboid cells that had been heterotrophically grown, feeding on K. aerogenes, had no chlorophyll fluorescence-positive granules in the cytoplasm (Figure 7(D), Figure 7(E)). Figure 8 shows excitation wavelength dependency of chlorophyll fluorescence in the area of a green plasmodium and amoeboid cells derived from it. From the fluorescence spectrum, it is evident that the maximum value (relative fluorescence emitted by 440 nm of wavelength) is around 680 nm (Figure 8), which is

a typical spectrum of chlorophyll fluorescence. According to continuous observation of the green plasmodium and amoeboid cells derived from it under a DIC-microscope, the amoeboid cells were found to be located on the underside of the plasmodium and perform an active amoeboid movement on the surface of the plasmodium. Here it is important to note that the amoeboid cells never take up the constituents of the green plasmodium by phagocytosis. Therefore, it is almost unlikely that the amoeboid cells may ingest the green plasmodium (a kind of biofilm) as a food source to support heterotrophic growth.

The genomic DNAs of parental D. purpureum cells,D. discoideum cells, and green multinucleate plasmodium containing amoeboid cells were isolated and amplified by PCR using 5.8S rRNA (ribosomal RNA (a non-coding RNA component of the large subunit of the eukaryotic ribosome) primers commonly used for phylogenetic studies of Dictyostelium. A single amplicon of a partial 5.8S rRNA sequence was amplified from the parental D. purpureum (Figure S6, Lane 5) andD. discoideum (Figure S6, Lane 4) genomic templates. However, several amplicons were observed with the genomic template of Dp-green cell masses (Figure S6, Lane 6). In addition, the DNA sequences similar to the D. purpureum genome were not found in the strongest band of the Lane 6. When the genomic template of parental D. purpureum cells was amplified with PCR using primers for Dd-MRP4 (mitochondrial ribosomal protein S4 of D. discoideum), a single, weak band was visible (Figure S6, Lane 2), whereas the genomic template of parental D. discoideum cells produced a single, strong band (Figure S6, Lane 1). In contrast, when green plasmodia containing amoeboid cells were used for PCR with the Dd-MRP4 primers, no bands were observed (Figure S6, Lane 3), indicating that the sequences comparable to Dd-MRP4 were either lost or partially mutated in the green plasmodia containing amoeboid cells and also in host D. purpureum cells.

For some of the amplifications, two amplicons were produced, each of which was identified as being DNA sequences derived from Rhodobacter and Synechocystis, suggesting that the two amplicons might be the result of the miss-priming of primers to the template (Figure S6. Lane 6). Parental D. purpureum 5.8S rRNA was not identified in these amplicons. Green plasmodium containing amoeboid cells were also analyzed with MiSeq genome wide sequencing. Curiously, assembled data were not identical with parental Dp cells’ genome, and seemed to contain many kinds of bacterial genome fragments. Although some of them were identified as Synechocystis- and Rhodobacter-genomes, most genome fragments were found to be similar to those of gram-negative aerobic bacteria such as Chitinophaga and Brevundimonas (data not shown). Unexpectedly, in the above MiSeq genome wide sequencing analysis some fragments were found that slightly resembled the gene sequence of free-living Acanthamoeba. Possible interpretation of these strange data will be discussed in the next section.