The chemicals N,N′-dicyclohexylcarbodiimide, 4-(dimethylamino)pyridine, potassium hydride (30% in mineral oil), methyl iodide, periodic acid, chromium(VI) oxide, Zn(NO₃)₂·6H₂O (98%, Fluka), Al(NO₃)₃·9H₂O (98.5%, Riedel de-Haën), 1 M NaOH (Fluka), Na₂S₂O₄ (Panreac), phosphate buffered saline (PBS) tablet, and 4-(2-hydroxyethyl)piperazine-1-ethanesulfonic acid (HEPES, 99.5%) were obtained from commercial sources (Sigma-Aldrich, unless otherwise indicated) and used as received. Lyophilized horse heart myoglobin and PBS solution were acquired from Sigma-Aldrich. THF (Aldrich) was dried over 4 Å molecular sieves. A nitrate-form Zn-Al LDH (denoted LDH-NO₃) with the composition Zn₄Al₂(OH)₁₂(NO₃)₂·2.5H₂O was prepared by using the standard method of coprecipitation of the Zn²⁺ and Al³⁺ hydroxides (initial Zn²⁺/Al³⁺ molar ratio in solution = 2) in the presence of nitrate ions at a constant pH of 7.5–8 under nitrogen, followed by aging of the gel at 80°C for 20 h (Gomes et al., 2013; Costa et al., 2017).

FT-IR spectra were collected using KBr pellets and a Mattson-7000 infrared spectrophotometer. Solution ¹H NMR spectra were recorded either on a Bruker Avance III spectrometer operating at 400.13 MHz (for irradiated DMSO-d₆ solutions) or (for all other measurements) a Bruker-AMX spectrometer. Solution ¹³C{¹H} NMR spectra were recorded on a Bruker-AMX spectrometer with an operating frequency of 101 MHz. Solid-state ¹³C{¹H} cross-polarization (CP) magic-angle spinning (MAS) NMR spectra were recorded using a wide-bore Bruker Avance III 400 spectrometer (9.4 T) at 100.62 MHz with 3.7 µs ¹H 90° pulses, 3.5 ms contact time, spinning rates of 12 kHz, and 5 s recycle delays.

Microanalyses for C, H, and N were carried out with a Truspec Micro CHNS 630-200-200 elemental analyzer. Powder X-ray diffraction (PXRD) data were collected at ambient temperature on a Philips Analytical Empyrean diffractometer equipped with a PIXcel 1D detector, with automatic data acquisition (X’Pert Data Collector software version 4.2) using monochromatized Cu-K_α radiation (λ = 1.54178 Å). Intensity data were collected by the step-counting method (step 0.02°), in continuous mode, in the 2θ range 3–70°. Scanning electron microscopy (SEM) images were obtained on a Hitachi SU-70 microscope at 15 kV. Samples were prepared by deposition on aluminium sample holders followed by carbon coating using an Emitech K 950 carbon evaporator. Thermogravimetric analysis (TGA) was performed using a Hitachi STA300 system at a heating rate of 5°C min⁻¹ under air.

Baseline-corrected absorption spectra for the Mb assays were measured from 200 to 800 nm at a scanning rate of 600 nm/min on a Cary 5000 UV-Vis-NIR spectrometer. Alternatively, a Shimadzu 2600 was used to obtain the UV-Vis absorption spectra, and a Horiba-Jobin-Yvon Spex Fluorolog 3-2.2. spectrophotometer, corrected for the instrumental response of the system, was used to record the fluorescence spectra.

Structural models and representations were generated using CrystalMaker software (CrystalMaker Software, 2017).

Photolysis experiments were carried out using a 158.5 W medium pressure Peschl Ultraviolet mercury arc lamp (catalog No. 50043) equipped with a circulating water cooled pyrex jacket. Supplementary Figure S1 shows a typical setup for a myoglobin (Mb) UV-Vis assay in which the cuvette containing the test solution is placed at a distance of 5 cm from the lamp. A similar arrangement was used to irradiate solid samples or DMSO-d₆ solutions of 3 in an NMR tube.

The stability of LDH-KDA3 in different aqueous media was explored through the following experiments: 1) A sample of LDH-KDA3 (27 mg) was irradiated in the solid-state for 12 h. The solid (iLDH) was then added to a solution of Na₂CO₃ (125 mg, 1.18 mmol) in deionized water (7 ml), and the suspension was stirred overnight at rt. The resultant solid (designated iLDH^DI, where DI stands for deintercalated) was recovered by filtration, washed with deionized water (2 × 5 ml), and vacuum-dried at rt. A second solid designated as iKDA^DI was recovered from the filtrate by evaporation of the solution to dryness. This procedure was also performed for non-irradiated LDH-KDA3, giving the solids LDH^DI and KDA^DI. 2) In two parallel experiments, LDH-KDA3 (20 mg) was incubated in 0.01 M PBS or 0.01 M HEPES buffer solutions at rt for 5 h. The resultant solids, designated as LDH^PBS and LDH^HEPES, were recovered by filtration, washed with deionized water (2 × 10 ml), and vacuum-dried at rt.

The photoinduced release of CO from the synthesized compounds was assessed by using the Mb assay, in which the conversion of deoxymyoglobin (deoxy-Mb) to carbonmonoxy-myoglobin (MbCO) can be spectrophotometrically measured (Motterlini et al., 2002).

The heme group functions as the active site of Mb, originating two π → π* electronic transitions: one, very intense, at about 400 nm (Soret or B band), and a second at 500–600 nm (the Q bands). The wavelengths corresponding to these wavelength maxima are governed by the oxidation, spin, and coordination states of the heme iron. As a result, it is possible to distinguish between different forms of Mb by the respective peak positions and relative optical density values of the absorption spectra.

All experiments were performed by using a stock solution of Mb, which was freshly prepared by dissolving the protein in degassed (O₂ free) 0.1 M PBS (pH = 7.4). An aliquot of this solution was taken and bubbled with gaseous nitrogen (99.99% pure), to which freshly prepared 1 M Na₂S₂O₄ (300 μl) was added to promote conversion of met-myoglobin into deoxy-Mb. While the solution was being bubbled, a solution (2 ml) of the ketone was added and the solution made up to 3 ml with 0.1 M PBS. After 15 min of bubbling, this solution was stored in quartz cuvettes with a magnetic stirring bar, with an optical path of 1 cm, and sealed with a Teflon stopper and parafilm to prevent any escape or entry of gas. An absorbance spectrum was immediately acquired after preparation of the samples. These were subsequently irradiated at room temperature and absorption spectra of the irradiated solutions were recorded periodically.

Blank solutions of deoxy-Mb with the same concentration for each experiment were prepared simultaneously and their absorption spectra were measured; these blank solutions were immediately saturated with CO gas at 10 bar for at least 1 h to promote a complete conversion of deoxy-Mb to MbCO using an autoclave reactor.

The CO released was quantified by following the absorbance of samples at 540 nm as reported previously (Atkin et al., 2011), using blank solutions and four isosbestic points between deoxy-Mb and MbCO in Q Bands to normalize data.

All theoretical calculations were of the DFT type, carried out using version R3 of GAMESS-US (Schmidt et al., 1993). A range-corrected LC-BPBE ( ω = 0.20 au⁻¹) functional, as implemented in GAMESS-US (Schmidt et al., 1993), was used in both ground- and excited-state calculations. TDDFT calculations, with similar functionals, were used to probe the excited-state potential energy surface (PES). A solvent was included using the polarizable continuum model with the solvation model density to add corrections for cavitation, dispersion, and solvent structure. In TDDFT calculation of FC (Franck-Condon) excitations the dielectric constant of the solvent was split into a “bulk” component and a fast component, which is essentially the square of the refractive index. In “adiabatic” conditions only the static dielectric constant is used. A 6-31G** basis set was used in either DFT or TDDFT calculations. The results obtained with the LC-BPBE(20) functional are essentially unscaled raw data from calculations; for the S₀→S_n transitions, a small correction, which results in the subtraction of 0.05 eV, to account for the difference between zero point and the first vibronic level, was considered. For the resulting optimized geometries time dependent DFT calculations (using the same functional and basis set as those in the previous calculations) were performed to predict the vertical electronic excitation energies.

A Zn-Al LDH intercalated by the deprotonated form of the diketoacid 3 was prepared by a direct coprecipitation method. The PXRD pattern of the resultant solid, designated as LDH-KDA3, is typical of LDHs containing organic guests (Figure 1C). Five equally spaced basal (00l) reflections are observed at 2θ angles below 30°. The sharpness of the basal reflections indicates that the layer structure is well-ordered, i.e., the interlayer spacing is highly regular. On the other hand, the presence of several weak, broad and asymmetric non-basal reflections at 2θ angles above 30° indicates disorder in the layer stacking, such as a turbostratic distortion or an intergrowth of the rhombohedral and hexagonal polytypes (Bellotto et al., 1996; Evans and Slade, 2006). Since the type of polytype cannot be determined with any degree of certainty, the basal reflections have been indexed as a one-layer polytype (001, 002, 003, etc.). The interlayer spacing (c ₀) can be calculated from averaging the positions of the five 00l harmonics: c ₀ = (d ₀₀₁ + 2d ₀₀₂ + 3d ₀₀₃ + 4d ₀₀₄ + 5d ₀₀₅)/5 = 18.0 Å. This is almost exactly double the value of c ₀ (8.89 Å) for the reference material LDH-NO₃ (Figure 1B).

Subtracting the hydroxide layer thickness of 4.8 Å from the interlayer spacing of 18.0 Å for LDH-KDA3 gives a gallery height of 13.2 Å. To the best of our knowledge, no crystal structure has been reported for the ketodiacid 3. A structure has, however, been deposited with the Cambridge Structural Database (Groom et al., 2016) for the synthetic precursor to 3, i.e., 2,4-dimethyl-2,4-di-p-tolylpentan-3-one (2) (CSD Refcode WIVHOZ). A realistic model for the molecular structure of the deprotonated form of 3 was created by extracting a molecule of 2 from the crystal structure and replacing the p-tolyl methyl groups by carboxylate groups with a fixed C‒O bond length of 1.25 Å and a <OCO angle of 125°. Figure 2 shows a schematic representation of a guest orientation that is calculated to give the experimentally observed gallery height of 13.2 Å. In this model, the guest molecules are markedly inclined from the normal to the hydroxide layers; the molecular axis defined by a vector joining the two carboxylate carbon atoms is at an angle of ca. 52° to the hydroxide surface of the layers. Apart from the fact that this arrangement gives a gallery height of 13.2 Å, there are other features that support this model: 1) hydrogen-bonding interactions between the carboxylate oxygen atoms and the layer hydroxyl groups are fortified, while positioning the hydrophobic dicumyl ketone section of the molecule towards the center of the interlayer region; 2) the inclined orientation may allow an efficient guest packing mode in which molecules can interlock as represented in Figure 2. The last argument is reinforced by the fact that a strikingly similar arrangement is present in the crystal packing of compound 2, when viewed down the crystallographic b axis, as shown in Figure 2B.

The FT-IR spectra of the free ketodiacid 3 and the intercalated material LDH-KDA3 are shown in Figure 3. The spectrum of 3 is dominated by the very strong absorption at 1,686 cm⁻¹ assigned to overlapping ν(C=O) bands of the ketone and carboxylic acid groups. LDH-KDA3 displays this band at the same frequency, albeit with reduced relative intensity, and it is therefore assigned to ν(C=O) of the ketone group. The deprotonation of the carboxylic acid groups of the guest molecules in LDH-KDA3 is confirmed by the appearance of two strong bands attributable to ν _sym(CO₂) (1,396 cm⁻¹) and ν _asym(CO₂) (1,532 cm⁻¹) vibrations, together with the absence of a ν(C‒O) band (present at 1,285 cm⁻¹ for 3). A weak, sharp band at 1,385 cm⁻¹ that overlaps with the ν _sym(CO₂) band may be due to an asymmetric ν ₃ stretching mode of nitrate ions (present via cointercalation with 3 and/or formation of a secondary LDH-NO₃ phase). Bands at 1,586 and 1,609 cm⁻¹ for the intercalated LDH are assigned to aromatic ring stretching vibrations. Below 700 cm⁻¹, the three main bands observed at 426, 559 and 621 cm⁻¹ are attributed to the characteristic Zn/Al-OH lattice translation modes of Zn-Al LDHs (Kloprogge et al., 2004). Similar lattice mode bands are observed for the reference nitrate-form material LDH-NO₃ (Figure 3A). The Raman spectrum of LDH-KDA3 is fully consistent with the FT-IR spectrum, showing a very strong band at 1,606 cm⁻¹ (ν _ring) and a strong band at 1,399 cm⁻¹ [(ν _sym(CO₂)] (Figure 4).

The solid-state ¹³C{¹H} CP MAS NMR spectrum of LDH-KDA3 is shown in Figure 5 alongside the solution spectrum of the ketodiacid 3 in DMSO-d₆. The LDH displays eight distinct resonances that have a clear counterpart in the solution spectrum. Thus, the signals for the aliphatic (20–60 ppm), aromatic (120–150 ppm) and carbonyl (170–220 ppm) carbon atoms of LDH-KDA3 (and 3, given in parentheses) are attributed as follows: (δ, ppm) CH₃ at 27.4 (27.5), C(CH₃)₂ at 53.4 (53.0), phenylene-CH at 126.5, 129.4, 133.6, and 147.2 (126.0, 129.0, 129.5, 148.9), ‒CO₂ ⁻ at 174.1 (167.1), and C=O at 215.1 (211.7). Hence, as expected, the most significant change is the downfield shift of the carboxylate resonance owing to deprotonation.

TGA revealed that the organic guest anion in LDH-KDA3 starts to decompose at a significantly higher temperature than that determined for the free ketodiacid 3 (Figure 6). The intercalated LDH shows three main weight loss steps between ambient temperature and 500 °C corresponding to removal of physisorbed and cointercalated water molecules (8.9% mass loss up to 150°C), dehydroxylation of the hydroxide layers (6.9% loss in the range 150–210°C), and decomposition of the organic guest (31.2% loss in the range 330–500°C). The enhanced thermal stability of the intercalated ketodicarboxylate anions is underscored by the shift of the decomposition onset from about 230°C for 3 to 330°C for LDH-KDA3, as well as the shift of the maximum of the differential thermogravimetric (DTG) curve from 335°C for 3 to 470°C for LDH-KDA3. Comparable increases in thermal stability of organic carboxylates (vs. the initial free acid or sodium salt) have been reported for other systems, including Zn-Al LDHs containing intercalated benzenecarboxylate derivative anions, e.g., positive shifts of ca. 150°C for 5,5′-methylenedisalicylic acid (Cui et al., 2010), 130°C for aurintricarboxylic acid (Zhu et al., 2011), and 50°C for benzene carboxylate and 4-hydroxy-benzene carboxylate (Fleutot et al., 2012). These improvements in thermal stability are attributed to the formation of an ordered supramolecular structure with significant interactions (including electrostatic interaction between opposite charges, hydrogen bonding and van der Waals interactions) between the organic guest and the host layer.

The phase purity of the material LDH-KDA3 was verified by SEM and EDS (Figure 7). The morphology of the intercalated LDH consisted of large aggregates of irregular sheet-like nanoparticles. EDS analyses indicated a uniform Zn/Al atomic ratio of 2.15 ± 0.05, which is quite close to the initial value of 2 in solution. The slight increase may have been caused by some leaching of aluminium during excessive washing with water. No evidence was found for secondary phases containing Zn (e.g., ZnO) and/or Al (e.g., Al₂O₃), in agreement with the PXRD data.

From CHN, EDS, and TGA analyses, together with the other characterization data, the formula Zn_4.3Al₂(OH)_12.6(C₂₁H₂₀O₅)_0.92(NO₃)_0.16(H₂O)₅ is proposed for the material LDH-KDA3. As discussed above, the presence of a small amount of nitrate ions was indicated by FT-IR spectroscopy, and the nitrate content in the formula is consistent with microanalyses for N. Since neither the IR spectrum nor the solid-state ¹³C{¹H} CP MAS NMR spectrum of LDH-KDA3 show peaks characteristic of carbonate ions, any interference from such species is deemed to be insignificant, and hence the presence of CO₃ ²⁻ (in addition to NO₃ ⁻) is not contemplated in the proposed formula.

To assess the stability of LDH-KDA3 in aqueous media, the material was incubated in two different biological buffers, 0.01 M PBS and 0.01 M HEPES, for 5 h at room temperature. No significant alterations in the IR spectrum or PXRD pattern of LDH-KDA3 were observed after these treatments (Supplementary Figures S2, S3), indicating that no structural changes took place, such as deintercalation of ketodicarboxylate anions by ion-exchange with phosphate ions in the PBS buffer solution.