Escherichia coli DH5α carrying various plasmids were grown in LB medium, which contained special antibiotics such as 50 μg/ml spectinomycin (Sp^R), 50 μg/ml kanamycin (Km^R), or 25 μg/ml chloramphenicol (Cm^R). Synechocystis cells were grown in the BG11 medium and cultured at light conductions (50 μmol photons m⁻² s⁻¹ and 30°C). Unless otherwise noted, appropriate antibiotics were added to the BG11 medium.

The pMD18-T vector (TaKaRa, Dalian) is used as a backbone to construct cyanobacterial plasmids, which are presented in Supplementary Table S1. Using PCR to amplify the fragments, the fragment and the vector were double-digested by recombinase (NEW ENGLAND BioLabs Beijing, China). Corresponding primers (SupplementaryTable S2) were used to clone up/down-fragments of the Synechocystis genome, and T4 ligase (NEB, Beijing) was used for ligation. The recombinant pBE406 plasmid (containing 600 bp upstream/downstream slr0168 and spectinomycin resistance gene) for gene knockout is shown in Supplementary Figure S2. Similarly, pMD-slr0301-Ω and pMD-slr1176-Ω were constructed. Otherwise, to construct a recombinant ethanol pathway, the synthesized pdc and yqhD genes (Sangon Biotech Co Ltd., Shanghai) coupled with promoters (PpetE or PpsbA2s) and TrbcL terminator were designed in pBE02/pBE03 (see the target genes including other resistance genes in Supplementary Table S1). Plasmid pBE09 was constructed by inserting the PpsbA2s-maeB expression cassette (maeB gene cloned from E. coli) and TrbcL terminator into the pMD-slr1176-Ω vector.

Synechocystis cells collected at the exponential phase (~1 OD₇₃₀) were washed with a fresh BG11 medium three times, and then mixed with plasmids (100 ng DNA to 100 μl cyanobacteria) for 5 h and illuminated incubation at 30°C. The above mixture was streaked on a sterile filter membrane for another 24 h of illuminated incubation on the BG11 solid medium. To select the corrected mutant, the filter membrane was further transferred to a solid BG11 medium with corresponding antibiotics. After 2 weeks, single clones sub-cultured on solid plates were isolated in a liquid BG11 medium for analysis. All the strains referred to in this study are presented in Table 1.

The slr0301 gene is a gene encoding phosphoenolpyruvate synthase (PpsA) in the genome of Synechocystis sp. PCC6803, which catalyzes the conversion of pyruvate to phosphoenolpyruvate. The absence of this gene increases the accumulation of the intermediate pyruvate. The slr1176 gene is a gene encoding glucose-1-phosphate adenylate transferase in the genome of PCC6803, which catalyzes the conversion of glucose-1-phosphate (G1P) to ADP-glucose. It is a major rate-limiting enzyme in glycogen synthesis, and its absence can lead to complete inhibition of glycogen synthesis. The slr0168 is a neutral gene in PCC6803 algae cells, and knocking out this gene has no effect on the growth of algae cells, making it a commonly used expression platform.

For ethanol production, all the mutants were cultured in a fresh BG11 medium with an initial 0.1 OD₇₃₀ and cultivated photoautotrophically in a flask (50 μmol photons m⁻² s⁻¹ without additional CO₂ injection). Notably, the BG11 medium of SYN001 contains 500 nM copper ions to induce the expression of ethanol-producing genes (Ghassemian et al., 1994; Choi and Park, 2016). After centrifugation and filtration, supernatant with ethanol was submitted for high-performance liquid chromatography (HPLC) analysis using an Aminex HPX-87H column (Bio-Rad, United States) (Seo et al., 2017).

Synechocystis wild-type culture and mutants at 0.6 OD₇₃₀ were collected after centrifugation (3,500 × g, 15 min, 4°C). RNA extraction and quantitative reverse transcription PCR (RT-qPCR) analysis were performed according to the previous methods (Gao et al., 2012). The relative transcription levels of targeted genes were estimated using the calculation method of 2^−ΔΔCT, in which a higher ΔCT value means low transcription (Livak and Schmittgen, 2001). The endogenous 16S rRNA was set as a reference gene. All experimental groups were carried out with three biological replicates.

All statistical analyses were performed using GraphPad Prism (version 8.01, United States). The difference in this study was compared by unpaired t-test and statistical significance was set at p < 0.05. ^** represents p < 0.01 and ^*** represents p < 0.001.

Generally, there are two precursors (pyruvate and acetyl-CoA) that are involved in ethanol synthesis (Gao et al., 2012). To determine the optimal ethanol pathway in engineered Synechocystis, the concentrations of those two metabolites were investigated (Supplementary Figure S1). Interestingly, pyruvate linking to the CBB cycle and tricarboxylic acid (TCA) cycle exhibited higher concentration (1.05 μmol/gDW, approximately three times that of acetyl-CoA), indicating that pyruvate was more suitable to serve as the ethanol precursor. Therefore, we attempted to optimize the carbon flow network by selecting a strong pyruvate-acetaldehyde-ethanol pathway.

The synthetic ethanol pathway contained pyruvate decarboxylase (PDC) from Zymomonas mobilis and alcohol dehydrogenase (YqhD) from Escherichia coli (Supplementary Figure S2). Those two enzymes exhibited high activities to convert pyruvate into acetaldehyde and subsequently reduce acetaldehyde into ethanol in other microorganisms (Atsumi et al., 2009; Gao et al., 2012). The slr0168 gene not affecting cell growth or photosynthesis was primarily chosen as an exchange site according to previous reports (Dexter and Fu, 2009; Gao et al., 2012). Vectors containing upstream/downstream slr0168 gene, antibiotic resistance gene (spectinomycin, Sp^R), various promoters, target genes (pdc-yqhD), and terminator TrbcL were constructed and integrated into the Synechocystis genome to obtain stable ethanol-producing recombinants (Supplementary Figure S3). Those two strains (SYN001 and SYN002) showed similar growth rates to wild-type Synechocystis (Figure 1A).

Subsequently, promoter optimization to adjust the transcription level of pdc and yqhD was conducted. The RT-qPCR results showed that strong promoter P_psbA2s in recombinant SYN002 obviously improved pdc-yqhD transcription levels, ~2-fold compared to medium-level promoter P_petE in SYN001 (Figure 1B and Supplementary Figure S4). As expected, both recombinants achieved obvious accumulation of ethanol (rarely detected in wild-type Synechocystis), and SYN002 yielded the highest titer (474 mg/L) at 7 days (Figure 1C). Thus, a basic cyanobacterium with photosynthetic CO₂-to-ethanol ability was obtained.

Blocking carbon loss (phosphoenolpyruvate backflow and glycogen synthesis) was conducted to learn their effects on CO₂-to-ethanol production (Figure 2A). From metabolic network analysis (Supplementary Figures S1, S2) and literature investigation (Angermayr et al., 2014; Dienst et al., 2014), we learned that native Synechocystis could remarkably turn pyruvate back into the upstream module via highly active phosphoenolpyruvate synthase (PpsA, referred to slr0301 gene). To abolish competitive consumption of pyruvate, we constructed a new cassette (pMD-P_psbA2s-pdc-yqhD) to exchange the slr0301 gene on the genome (Figure 2B). The newly obtained recombinant SYN003 (Δslr0301) exhibited 600 mg/L ethanol yield after 7 days of photosynthetic CO₂ conversion (Figure 2C) and approximately 1.3-fold improvement compared to PpsA-existed SYN002. Furthermore, to enhance photosynthetic carbon flux toward the CO₂-to-ethanol pathway, a key gene slr1176 related to glycogen synthesis was knocked out via gene exchange cassette (pMD-slr1176-Ω) (Figure 2B). Impressively, the ethanol yield was further improved via double-knockout recombinant SYN007 (Δslr0301 Δslr1176) and obtained more than 700 mg/L titer (Figure 2C). Pyruvate was deduced as a carbon sink in the Embden–Meyerhof–Parnas pathway according to the previous study of glycogen synthesis abolishment (Van Der Woude et al., 2014). The increased carbon flux of CO₂-to-pyruvate probably supported the pyruvate-utilizing reaction of ethanol accumulation. Interestingly, it slightly inhibited cell growth when blocking glycogen synthesis at the slr1176 site (Figure 2C). We deduced that the shift of excessive carbon from the glycolytic pathway and pentose phosphate pathway to ethanol pathway resulted in carbon deficiency of biomass synthesis (Young et al., 2011).

The engineered strains in cell proliferation stage should use the TCA cycle to support biomass synthesis, however, resulting in carbon atom loss (Zhang and Bryant, 2011). Thus, we designed a simple pathway to modify the TCA cycle by improving glyoxylate flux to reduce the carbon loss between isocitrate and succinate (Supplementary Figure S2). Malate close to the end of the TCA cycle was selected as the key metabolite for carbon recycling through an NADP⁺-dependent malic enzyme from E. coli (Yoshikawa et al., 2015), which not only converted malate into pyruvate but also increased pyruvate and NADPH pool (Figure 3A). The maeB gene was introduced into the slr1176 site of SYN003, establishing the recombinant SYN009 (Figure 3B). Intriguingly, the cell growth of Synechocystis was rescued, and the SYN009 showed a little fast proliferation after 3 days of photoautotrophic growth (Supplementary Figure S6). Under simulated sunlight source, SYN009 used CO₂ as a sole carbon source to produce 1.09 g/L ethanol (Figure 3C and Supplementary Figure S5). Impressively, the time curve showed that before 4 days, SYN009 achieved 248 mg/L/day productivity, the fastest accumulation rate compared to other literature studies (Table 2).

The metabolic flux imbalance between metabolism and synthesis is a big challenge limiting target product yield in microbial cell factories (Oliver et al., 2013). Choosing suitable promoters to overexpress PDC/YqhD has been usually considered to enhance the carbon flux toward ethanol (Dexter and Fu, 2009; Gao et al., 2012), which also exhibited positive results of ethanol production in this study (Supplementary Figure S5). In addition, the overexpression of key enzymes in the CBB cycle was another important strategy to supply sufficient carbon flux in the form of 3-phosphoglycerate (Liang et al., 2018; Roussou et al., 2021). Compared to previous reports (Table 2), the carbon flow optimization strategy in our study consumed the minimum number of days to achieve the highest ethanol productivity of ~248 mg/L/day (Table 1). It indicated that a comprehensive and precise adjustment of carbon flow is promising to improve CO₂-to-ethanol production in cyanobacteria. Notably, this photosynthetic cell factory still faces the challenge of cell density (Supplementary Figure S6), such as only ~0.6 OD₇₃₀ increase after 7 days of cultivation under light and atmospheric CO₂. New strategies, such as co-cultivation and batch culture with high density, can probably yield outstanding ethanol production (Namakoshi et al., 2016; Velmurugan and Incharoensakdi, 2020).