The Oxygen Plants Release Does Not Come from CO₂—Tracing How Water Becomes O₂

Plants take in carbon dioxide (CO₂) and release oxygen (O₂).

Placed side by side, those two facts invite a very natural assumption: perhaps the O₂ released by plants is simply made from the oxygen atoms in the CO₂ they absorbed.

It is not.

The oxygen atoms in the O₂ released by oxygenic photosynthesis come from water (H₂O), not from CO₂.

That is not a new discovery. Isotope experiments settled the question in 1941.

The harder question came next.

Where do the oxygen atoms from water enter the photosynthetic machinery? How are two of them brought together, and at what point do they form the O–O bond of molecular oxygen?

More than eight decades later, that reaction is still not completely resolved at the atomic level. A study by Bhowmick and colleagues, published in Nature Communications on August 22, 2026, examined the S₃ state of photosystem II (PSII) at approximately 1.9 Å resolution and provided renewed structural support for a long-debated sixth oxygen ligand, OX or O6.

This article follows the question from the historical discovery that photosynthetic O₂ comes from water to the Mn₄CaO₅ cluster, the Kok–Joliot cycle, XFEL crystallography, the O5–O6 distance, and the current debate over when the O–O bond actually forms.

Why “plants take in CO₂ and release O₂” can be misleading

Photosynthesis is often summarized with an overall equation such as:

6 CO₂ + 6 H₂O → C₆H₁₂O₆ + 6 O₂

That equation is useful as a net accounting of photosynthesis, but it does not tell us which atoms end up where.

If we isolate the oxygen-evolving reaction of PSII, the chemistry is better written as:

2 H₂O → O₂ + 4 H⁺ + 4 e⁻

In other words, the reaction that generates the O₂ we breathe removes four electrons and four protons from two water molecules and ultimately joins the two remaining oxygen atoms into O₂.

CO₂ is used primarily on the carbon-fixation side of photosynthesis. Oxygen evolution and carbon fixation are coupled within the overall process, but plants are not simply stripping oxygen atoms from CO₂ and releasing them as O₂.

1931: sulfur bacteria pointed to water as the source of oxygen

One of the key conceptual advances came from microbiologist Cornelis B. van Niel.

In 1931, van Niel compared different forms of photosynthesis. Some photosynthetic bacteria use compounds such as hydrogen sulfide (H₂S), rather than water, as electron donors and produce sulfur rather than O₂.

From these comparisons, van Niel developed a more general view of photosynthesis. In plants, he proposed, O₂ should not be viewed as a fragment released by breaking down CO₂. Instead, it should arise from oxidation of the electron donor—H₂O.

Today this sounds textbook-like. At the time, it was a mechanistic hypothesis.

1937: isolated chloroplasts could evolve oxygen without fixing CO₂

Robert Hill provided another crucial piece of evidence in 1937.

Isolated chloroplasts could evolve O₂ in the light when suitable artificial electron acceptors were present, even when carbon fixation was separated from the reaction.

This became known as the Hill reaction.

The result helped split a once-black-box view of photosynthesis into separable processes: light-driven electron transfer and oxygen evolution on one side, and CO₂ fixation on the other.

1941: following oxygen atoms with ¹⁸O

The decisive experiment used isotope tracing.

In 1941, Samuel Ruben, Merle Randall, Martin Kamen and James Logan Hyde used the heavy oxygen isotope ¹⁸O as a tracer in photosynthesis.

When the water was labeled with ¹⁸O, the isotope appeared in the evolved O₂.

This established experimentally that the oxygen atoms released as O₂ during photosynthesis originate from water.

One question had been answered.

But the harder question remained:

Where and when do two water-derived oxygen atoms form the O–O bond that produces O₂?

That is the question to which the 2026 study directly contributes.

The oxygen-making catalyst sits in photosystem II

In oxygenic photosynthesis, water oxidation takes place in photosystem II (PSII), a large membrane protein complex embedded in the thylakoid membrane.

Photoinduced charge separation in the PSII reaction center generates the strongly oxidizing P680⁺. Through the redox-active tyrosine TyrZ in the D1 protein and associated proton-coupled electron-transfer chemistry, oxidizing equivalents are transferred to the catalyst that ultimately extracts electrons from water.

That catalyst is the oxygen-evolving complex (OEC).

At its core is the Mn₄CaO₅ cluster, consisting of four manganese ions, one calcium ion and bridging oxygen ligands, together with bound waters and a surrounding hydrogen-bond network.

In 2011, Umena and colleagues reported a 1.9 Å crystal structure of PSII that resolved the geometry of the Mn₄CaO₅ center, its ligands and nearby waters in remarkable detail.

But knowing a static structure is not the same as knowing how the atoms move during catalysis.

Producing O₂ requires a four-electron oxidation. PSII therefore does not convert water to O₂ in one photochemical event. It stores oxidizing equivalents stepwise and releases O₂ only after the catalytic center has advanced through several intermediates.

The four-step rhythm that led to the Kok–Joliot cycle

In 1969, Pierre Joliot and colleagues showed that when dark-adapted photosynthetic material was exposed to a sequence of short saturating flashes, the amount of oxygen released after each flash oscillated with a period of four.

Bessel Kok and colleagues formalized this behavior in 1970 with the S-state model.

The OEC advances through:

S₀ → S₁ → S₂ → S₃ → S₄ → S₀

S₀ through S₃ are relatively long-lived catalytic states. S₄ is usually treated as a short-lived state associated with O–O bond formation and oxygen release. Dark-adapted PSII accumulates mainly in S₁.

The S-state number should not be interpreted as a simple manganese oxidation number. Each S state describes the redox state of the OEC as a whole, involving metal valence, ligand oxidation, proton transfer and structural rearrangement.

For the present story, the key point is that S₃ is the last metastable state before O₂ evolution.

Does a sixth oxygen enter during S₂→S₃?

The Mn₄CaO₅ cluster contains bridging oxygen sites conventionally labeled O1 through O5.

During the S₂→S₃ transition, structural studies have suggested that an additional oxygen ligand appears near the cluster. This site is commonly called OX or O6.

Beginning with XFEL studies reported in 2017 and later work, electron density consistent with O6 was observed in the S₃ state. In the 2018 structures reported by Kern and colleagues, O6/OX was positioned between Mn1 and Ca, approximately 2.1 Å from O5.

But the interpretation remained controversial.

Different XFEL structures reported somewhat different O5–O6 distances. The chemical identity of O6—whether best described at a given stage as H₂O, OH⁻, an oxo species, an oxyl-like species or something more electronically delocalized—has remained debated. So has the question of whether an O–O bond is already present in S₃.

More recently, reanalyses challenged whether a distinct sixth oxygen had to be modeled at all, proposing that some of the density might instead reflect alternative conformations of O5.

In May 2026, Li, Suga and Shen reanalyzed earlier XFEL data using omit-map strategies and again supported a partially occupied O6 site in S₃.

That context matters: the August 2026 Bhowmick study did not discover O6 for the first time.

Its importance lies in bringing an independent structural strategy to the long-running question of whether S₃ truly contains a distinct O6 ligand.

The 2026 strategy: locate Mn first, then interpret the oxygen

Bhowmick and colleagues obtained high-resolution structural data for the S₁, S₂ and S₃ states at approximately 1.9 Å.

They also collected anomalous diffraction data at two X-ray energies, 9.5 keV and 7 keV.

That methodological choice is important in a metal-containing active site such as the OEC. Mechanistic interpretation depends critically on knowing where the heavy Mn atoms sit before assigning weaker density to nearby oxygen atoms.

The authors therefore used anomalous scattering to constrain the manganese positions and then interpreted the oxygen sites using the high-resolution maps.

In the S₃ state, OX/O6 appeared as a distinct peak without requiring positional restraints and was modeled as a ligand to both Mn1 and Ca.

This strongly supports a model in which an additional oxygen ligand is introduced during the S₂→S₃ transition.

The phrase “a water enters,” however, needs to be read carefully.

Crystallography directly gives evidence for an oxygen position and its electron density. It does not, by itself, fully establish when that atom entered as H₂O, when it lost protons, or exactly what protonation and electronic state it occupies in S₃.

For that reason, it is more precise to describe O6 as an additional oxygen ligand that is interpreted as water-derived.

O5 and O6 are about 2.1 Å apart—too far for a completed O–O bond

One of the most consequential measurements in the 2026 structure is the separation between O5 and O6, approximately 2.1 Å.

A conventional peroxide-like O–O single bond is typically around 1.4–1.5 Å. The O–O bond in molecular oxygen is shorter still.

A separation of about 2.1 Å is therefore much longer than a fully formed conventional covalent O–O bond.

Bhowmick and colleagues conclude that the structure does not support a strong, mature O–O bond already being present in the metastable S₃ state.

This point also requires restraint in interpretation.

The distance alone cannot uniquely determine the protonation state, radical character or partial electronic coupling of O6. Nor does the structure prove that O5 and O6 are the two substrate oxygen atoms that ultimately leave as O₂.

The mechanistic conclusion is therefore conditional:

if O5 and O6 are the oxygen atoms that form dioxygen, the bond is more likely to form after metastable S₃, during the final oxidation associated with the S₃→S₀ transition, rather than being fully formed in S₃ itself.

The actual O₂-forming moment lies beyond metastable S₃

Time-resolved XFEL work published in 2023 provides useful temporal context.

After the third flash, as the system advances from S₃ toward oxygen release, the OX site introduced during S₂→S₃ begins to disappear or relocate at roughly 700 µs. Structural changes associated with the onset of O₂ evolution appear later, around 1,200 µs.

The 2026 picture—O6 is present in S₃, but a strong O–O bond is not yet complete—fits naturally with that time sequence.

A useful mechanistic picture is therefore emerging: by S₃, the OEC has positioned oxygen atoms in a catalytically poised configuration, and the final oxidation then drives O–O bond formation and O₂ release.

The title of this article says that we are “tracing the moment water becomes O₂.” That should not be read as meaning that the 2026 paper directly photographed the instant of bond formation in a single structure.

More precisely, it constrains the atomic geometry immediately before oxygen formation and, when combined with time-resolved work, narrows the trajectory into the O₂-forming step.

Open-cubane, closed-cubane and the S₂→S₃ transition

At a more technical level, the S₂ state itself has been discussed in terms of different structural and electronic forms, commonly described as open-cubane and closed-cubane configurations, with corresponding low-spin and high-spin behavior in spectroscopic and computational models.

The S₂→S₃ transition involves oxidation of the catalytic system, proton transfer, water delivery and rearrangement of the Mn coordination sphere. Mechanisms have been proposed in which water sites such as W3 participate in supplying the ligand that becomes O6.

But the 2026 structure does not uniquely establish which S₂ conformer must be traversed or the exact sequence of proton-transfer events by which O6 is incorporated.

What it strengthens is a more bounded statement: the final S₃ structure contains a distinct additional oxygen ligand between Mn1 and Ca.

Why XFEL matters: the X-rays can damage the very catalyst being studied

The OEC is unusually challenging for structural biology because Mn₄CaO₅ is a highly redox-active metal cluster.

Conventional X-ray exposure can reduce Mn and alter metal–metal distances or ligand positions, potentially perturbing the active site whose chemistry researchers are trying to understand.

This is one reason X-ray free-electron lasers (XFELs) have transformed the field.

Very short, intense X-ray pulses can record diffraction before large-scale structural destruction—the “diffraction before destruction” principle. Visible-light pump pulses can first advance PSII through the Kok cycle, followed by an X-ray probe pulse, allowing room-temperature snapshots of catalytic intermediates.

XFEL does not make every interpretive problem disappear. S-state excitation efficiency, mixed populations, occupancy, data processing, map sharpening, refinement strategy and ultrafast radiation-induced electronic or atomic responses still matter when structures are being compared at sub-ångström scales.

That is why the 2026 study’s use of two-energy anomalous diffraction to locate Mn before evaluating O6 is more than a technical detail. It directly addresses one of the sources of ambiguity in this active site.

The mechanism of “water to O₂” is still incomplete

By 1941, researchers knew that the oxygen atoms in photosynthetic O₂ came from water.

By 2011, the Mn₄CaO₅ catalyst could be visualized at near-atomic resolution.

XFEL studies in the 2010s and 2020s began to reveal structural intermediates along the Kok cycle.

And in 2026, the structural case for O6 in S₃ became stronger again.

Yet major mechanistic questions remain:

• Are O5 and O6 truly the two substrate oxygen atoms released as O₂?
• What are the protonation and electronic states of O6 during the individual steps of S₂→S₃→S₄?
• What is the precise structure and electronic state of the transient S₄ intermediate?
• Does O–O bond formation proceed by oxo–oxyl coupling, nucleophilic attack, a peroxide pathway, another oxygen pair, or state-dependent alternatives?
• How are water delivery, proton release, electron transfer and Mn-cluster rearrangement synchronized?

The 2026 paper does not “solve photosynthetic water splitting.”

Its contribution is more precise and, scientifically, more useful: it eliminates some structural possibilities and narrows the atomic configurations that a viable reaction mechanism must explain.

An 85-year-old question has become a question about the position of one atom

The history of photosynthesis research is also a history of increasing resolution.

In the 1930s, the question was whether the O₂ released by plants came from CO₂ or water.

In 1941, isotope tracing answered: water.

In 1969–1970, the four-flash rhythm of oxygen evolution was organized into the S-state cycle.

In 2011, the Mn₄CaO₅ catalytic center could be inspected at 1.9 Å resolution.

Today the debate has moved to questions such as whether one oxygen atom called O6 is truly present in S₃, how many ångströms separate it from O5, and whether an O–O bond already exists at that point.

“Plants make oxygen from water” is an easy sentence to write.

Inside that sentence is an atomic reaction that scientists have not completely solved after more than 85 years of work.

The 2026 study makes that black box slightly smaller.

References

Bhowmick A, Zhang M, Simon PS, Makita H, et al. An additional water is introduced into the manganese cluster during the formation of the S3 state of photosystem II. Nature Communications. 2026. DOI: 10.1038/s41467-026-76805-9

Ruben S, Randall M, Kamen M, Hyde JL. Heavy Oxygen (O18) as a Tracer in the Study of Photosynthesis. Journal of the American Chemical Society. 1941;63:877–879. DOI: 10.1021/ja01848a512

Umena Y, Kawakami K, Shen JR, Kamiya N. Crystal structure of oxygen-evolving photosystem II at a resolution of 1.9 Å. Nature. 2011;473:55–60. DOI: 10.1038/nature09913

Kern J, Chatterjee R, Young ID, et al. Structures of the intermediates of Kok’s photosynthetic water oxidation clock. Nature. 2018;563:421–425. DOI: 10.1038/s41586-018-0681-2

Bhowmick A, Hussein R, Bogacz I, et al. Structural evidence for intermediates during O2 formation in photosystem II. Nature. 2023;617:629–636. DOI: 10.1038/s41586-023-06038-z

Li H, Suga M, Shen JR. Comment on Photosystem II: light-dependent oscillation of ligand composition at its active site: existence of O6 in S3-state photosystem II revealed by omit maps. Acta Crystallographica Section D. 2026;82:574–586. DOI: 10.1107/S2059798326003621