Targeted, high-resolution sensing of volatile organic compounds by covalent nanopore detection

1 Volatile organic compounds are choice analytes in a variety of contexts. For 2 example, humans release over 4000 volatile organic compounds, many of which are 3 diagnostic of life-threatening medical conditions. A combination of a large number of 4 potential analytes requires the application of costly, cumbersome technology. Here, 5 we show that covalent nanopore sensing can be used for the targeted detection of a 6 reduced set of analytes in a mixture: in this case aldehydes, which constitute ~5% of 7 human volatiles. Further, nanopore engineering permits high-resolution detection, 8 which allows closely related aldehydes including isomers to be distinguished. 9 Differential sensing of other chemical classes, such as alcohols, is demonstrated by 10 leveraging their enzymatic conversion to aldehydes. Our approach is compatible with 11 the use of cheap, portable, user-friendly diagnostic devices applicable to a wide 12 variety of objectives, including pollutant monitoring, food and beverage testing and 13 the quality control of pharmaceuticals, as well as disease diagnostics. 14 15 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 18, 2024. ; https://doi.org/10.1101/2024.10.15.618413doi: bioRxiv preprint 3

1 Aldehydes are ubiquitously produced through chemical processes across various 2 biological, environmental, and industrial contexts. These include lipid peroxidation in 3 the human body, 1 incomplete combustion during wood burning, 2 and Strecker 4 degradation during beer storage. 3 Single-molecule detection of diverse aldehydes 5 against background chemicals therefore offers wide applications. 6 For example, small-molecule aldehydes are found in human breath or bodily fluids, 7 alongside a spectrum of volatile organic compounds (VOCs) generated through 8 metabolic processes 1 or by the gut microbiome. 4 Metabolic alterations in specific 9 conditions, such as cancers, 5 gastrointestinal diseases,6 respiratory disorders,7 and 10 viral infections,8 lead to distinct VOC profiles, which offer a non-invasive window into 11 the body's internal biochemical processes. The current gold standard for detection of 12 small molecules is liquid or gas chromatography-mass spectrometry (LC/GC-MS), 9 13 which generates a near-complete profile of all collected VOCs, but requires 14 centralized labs that use expensive equipment and sophisticated analysis packages. 15 Here, we focus on the development of a single-molecule, real-time detection 16 technology for rapid ratiometric profiling of aldehydes. The targeted detection of 17 aldehydes, comprising about 170 species among over 4000 VOCs of human origin,10 18 presents an appealing approach to reduce the complexity of small-molecule 19 fingerprints, while retaining significant diagnostic value. Aldehyde mixtures in breath 20 or bodily fluids commonly contain linear or branched, unsaturated or saturated 21 carbon chains, containing from 1 to 17 carbon atoms, as well as aromatic species 22 such as benzaldehyde and its derivatives. 10 A ratiometric fingerprint of aldehydes 23 holds promise for disease detection and health monitoring, provided that tests can 24 be made rapid, simple and inexpensive. For example, the ethanal:pentanal:heptanal 25 ratio in breath shifts from 112:1:1.5 in healthy individuals to 24:1.7:1 in patients with 26 lung cancer11, and similarly, the hexanal to heptanal ratio in urine changes from 1.3:1 27 to 2.5:1.12 For COVID infection, the relative abundance of octanal and benzaldehyde 28 is a hallmark of recent infection.8 29 We have established single-molecule covalent sensing mediated by protein 30 nanopores to detect analytes based on chemical reactivity. 13 The concept was first 31 demonstrated with engineered α -hemolysin ( αHL) pores14-17 and later validated by 32 others using alternative nanopores 18-20. Analyte molecules form covalent bonds 33 reversibly or irreversibly with a sensing group on the internal surface of a pore, 34 thereby generating characteristic changes in the ionic current flowing through the 35 pore under a transmembrane potential. When examining mixed analytes, the current 36 signatures (e.g., the extent of current blockade and the noise during blockades) 37 identify analytes, while the frequency of reversible events reveals the concentration 38 of individual analytes.21 39 One general challenge is to identify sensing chemistry with suitable kinetics to 40 ensure rapid and quantitative detection of analytes; the ideal lifetimes of covalent 41 analyte-pore adducts and the intervals between sensing events are tens to hundreds 42 of milliseconds, which can be reliably measured by electrical recording. This 43 requirement has so far limited the practical application of covalent sensing. Another 44 challenge is the rational engineering of nanopore sensors to resolve structures with 45 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 18, 2024. ; https://doi.org/10.1101/2024.10.15.618413doi: bioRxiv preprint 4 minor variations, such as chain isomers that differ by the position of a methyl group. 1 To date, this has largely relied on trial and error. A more systematic investigation is 2 essential to guide the future development of covalent sensors. 3 Here, we apply such a systematic approach, to exploit hemithioacetal chemistry for 4 the covalent sensing of aldehydes within a thiol-containing αHL nanopore, achieving 5 rapid, high-resolution analysis by the formation of short-lived adducts that are 6 identified by a machine learning algorithm. The diastereomeric adducts produce 7 current signatures with differences that are accentuated by rational engineering of 8 nanopores, affording information that discriminates between closely similar 9 molecular isomers. We extend the scope of our approach to alcohols by selectively 10 converting them to aldehydes. Hence, we have developed a versatile means for the 11 targeted detection of a subset of analytes present within a complex mixture of 12 molecules. 13 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 18, 2024. ; https://doi.org/10.1101/2024.10.15.618413doi: bioRxiv preprint 5

and discussion 1 Covalent detection of aldehydes through reversible hemithioacetal formation 2 Stochastic sensing of aldehyde analytes was achieved using reversible thiol-3 aldehyde chemistry. To eliminate background reactions, notably imine formation and 4 metal chelation, four mutations were introduced in wild-type α HL (i.e., AG = K8A-5 M113G-K131G-K147G) (Supplementary Fig. 1). 22 A single-cysteine mutation was 6 then introduced at position 115 (i.e., AG-T115C). Heteroheptameric nanopores 7 containing one cysteine-bearing subunit, (AG)6(AG-T115C), were prepared and used 8 for covalent detection (Fig. 1a, see Methods). (AG) 6(AG-T115C) carried a single-9 channel current (IP) of −129 ± 3 pA at −50 mV (N = 80 pores) (recording buffer: 2 M 10 KCl, 200 mM PIPES and 20 μ M EDTA at pH 6.8). 11 The introduction of an aldehyde in the trans compartment (Fig. 1a) produced 12 reversible current blockades, which were attributed to the formation of hemithioacetal 13 adducts (Fig. 1b,c and Supplementary Figs. 2-10). Residual currents (I res) after 14 hemithioacetal formation are given as percentages of the open pore current (I res% = 15 Ires/IP × 100 %) (Supplementary Table 1). For example, for propanal, I res% = 98.7 ± 16 0.1 % (N = 3 pores, >300 events). Aldehydes exist in both hydrated and non-17 hydrated forms in aqueous solution. 23 The rates of hemithioacetal formation ( von) 18 were expressed in terms of the total aldehyde concentration and were consistent 19 with bimolecular kinetics (i.e., for propanal, von = k on[propanal]tot, where k on is the 20 apparent rate constant of adduct formation). The rates of hemithioacetal dissociation 21 (voff) were independent of propanal concentration, consistent with a unimolecular 22 step (i.e., voff = koff, where koff is the rate constant of adduct dissociation) (Fig. 1d and 23 Supplementary Figs. 2-10). Both the association and dissociation reactions were pH-24 dependent and single-channel recordings were performed at pH 6.8. At this pH value, 25 frequent events occurred at µM-mM analyte concentrations (e.g., ~500 events within 26 10 min for 3 mM butanal), and the lifetimes of the hemithioacetal adducts were long 27 to allow accurate aldehyde identification from Ires% values (e.g., ~130 ms for butanal). 28 In total, 10 different aldehydes were characterized with the (AG) 6(AG-T115C) 29 nanopore (Fig. 2a and Supplementary Table 1), ranging from straight-chain to 30 branched-chain to aromatic aldehydes. We demonstrated single CH 2 resolution in 31 distinguishing straight-chain aldehydes (Fig. 2b and Supplementary Table 1). From 32 ethanal to octanal, each additional CH 2 group reduced the I res% value by ~0.7 %. 33 Interestingly, our approach was able to distinguish between diastereomeric 34 hemithioacetal adducts with opposite chirality at the Ca position for both heptanal 35 and octanal (∆ Ires% = 0.29 ± 0.03 % for heptanal (N = 3 pores, >100 events for each 36 diastereomer); 0.39 ± 0.05 % for octanal (N = 2 pores, >40 events for each 37 diastereomer)). We arbitrarily assigned adducts with larger I res% as diastereomers A 38 and those with smaller I res% as diastereomers B (i.e., I res%,A > I res%,B). For shorter 39 straight-chain aldehydes, diastereomeric adducts were not clearly separated by 40 (AG)6(AG-T115C) with ∆ Ires% 300 events) and 96.1 ± 0.1 % for hexanal (N 43 = 3 pores, >400 events); I res% = 96.6 ± 0.1 % for phenylacetaldehyde (N = 4 pores, 44 >400 events) and 95.7 ± 0.1 % and 95.4 ± 0.1 % for heptanal (N = 3 pores, >100 45 events for each diastereomer). We speculate that the open-chain aldehydes extend 46 away from the protein wall, creating larger steric blockades than the corresponding 47 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 18, 2024. ; https://doi.org/10.1101/2024.10.15.618413doi: bioRxiv preprint 6 cyclic aromatic aldehydes. Branched-chain aldehydes and their straight-chain 1 isomers produced similar I res% values, i.e., I res% = 98.1 ± 0.1 % for 2-methylpropanal 2 (N = 5 pores, >500 events) and 97.9 ± 0.1 % for butanal (N = 4 pores, >400 events), 3 with ∆ Ires% ~0.2 % for 2-methylpropanal and butanal simultaneously recorded with 4 the same channel (N = 1 pore, >600 events). 5 Simultaneous detection of aldehydes facilitated ratiometric profiling 6 To enable quantitative aldehyde detection, kinetic analysis of the formation and 7 dissociation of hemithioacetal adducts was carried out. The solubility of each 8 aldehyde in the recording solution was measured by NMR in the presence of internal 9 standards (5 mM calcium formate and 5 mM maleic acid). Aldehydes formed 10 hydrates to varying extents in aqueous solutions, and the hydrated forms were 11 unreactive towards thiols and thus not detected by single-channel recording (Fig. 1b). 12 We determined the hydration equilibrium constants (Khyd) by 1H NMR under electrical 13 recording conditions (Supplementary Table 2 and Supplementary Fig. 11) and 14 derived the corrected rate constants of adduct formation ( kon/i3 ) by using the 15 equations: 16 17 von = kon[aldehyde]tot = kon/i3 [aldehyde]ald 18 19 Khyd = [aldehyde]hyd/[aldehyde]ald = [aldehyde]tot/[aldehyde]ald – 1 20 21 kon/i3 = kon × (1 + Khyd) 22 23 von, rate of hemithioacetal formation; kon, apparent rate constant of hemithioacetal 24 formation; kon/i3 , corrected rate constant of hemithioacetal formation; Khyd, hydration 25 equilibrium constant; [aldehyde]tot, total concentration of aldehyde; [aldehyde] hyd, 26 concentration of aldehyde in the hydrate form; [aldehyde] ald, concentration of free 27 aldehyde. 28 For the bimolecular formation of hemithioacetal, kon/i3 remained at ~0.5-0.6 mM -1s-1 29 for ethanal, propanal, butanal, and pentanal but dropped to ~0.3 mM -1s-1 for hexanal 30 and heptanal. While the energy difference is small, these observations could reflect 31 the steric hindrance for the thiolate to approach the carbonyl group along the Bürgi-32 Dunitz angle, caused by the conformationally labile alkyl chains.24 As rates of adduct 33 dissociation were independent of aldehyde concentration, no correction for hydration 34 was required. For the unimolecular dissociation of hemithioacetals formed with 35 straight-chain aldehydes, koff remained within an order of magnitude, gradually 36 decreasing as the chain length increases from 9.7 s-1 for ethanal to ~5 s-1 for hexanal, 37 heptanal and octanal (Fig. 2c). This could be attributed to the positive inductive effect 38 of the increasing alkyl chain length. 39 Simultaneous detection of multiple aldehydes with a single nanopore was 40 demonstrated with a mixture of 7 straight-chain aldehydes (i.e., ethanal to octanal) 41 (Fig. 3a). The consistent current signatures for individual aldehydes and the clear 42 separation between them enabled automated assignment of events by using 43 machine learning (Fig. 3b). Almost 1000 individual events per aldehyde were 44 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 18, 2024. ; https://doi.org/10.1101/2024.10.15.618413doi: bioRxiv preprint 7 collected from separate traces of each analyte (Supplementary Fig. 12). A stratified 1 random split was performed, leaving 30% of the events as the test set. Three 2 features were extracted to characterize each event: I res%, event duration, and the 3 root-mean-squared noise of the event (See Supplementary Section 5). The random 4 forest model achieved the highest accuracy of 98% on both the training and test 5 sets, which was calculated as the fraction of correctly classified events in the dataset 6 (Fig. 3b and Supplementary 13). Further, we demonstrated ratiometric determination 7 of aldehydes in mixtures with pentanal and butanal, for which the detection 8 frequencies reflected their varied concentration ratios (Fig. 3c). For disease 9 diagnosis and product quality control applications, ratiometric profiles are expected 10 to be the most informative readout. 11 Rational nanopore engineering for diastereomer and structural isomer 12 resolution 13 Challenged by the similar current signatures observed with aldehyde structural 14 isomers (e.g., butanal and 2-methylpropanal) with the (AG) 6(AG-T115C) nanopore, 15 rational nanopore engineering was undertaken. We hypothesised that current level 16 resolution of diastereomeric hemithioacetal adducts would provide an additional 17 layer of information to aid structural isomer resolution. To this end, we designed two 18 additional nanopores in which the reactive cysteine residue was positioned within a 19 narrower, and hence more sensitive, region of the nanopore β barrel (Fig. 4a). In the 20 (MK)6(MK-T115C) nanopore (MK = WT-K8A-K131G-K147G), methionine residues 21 were reintroduced at position 113 to reduce the internal diameter near the sensing 22 site. In the (AG)6(AG-G137C) nanopore, the cysteine residue was moved to position 23 137, which was in close proximity to the narrowest region of the nanopore bearing an 24 AG background (Supplementary Fig. 14). Three asparagine-to-alanine mutations 25 around the sensing site were further introduced in to reduce sterics and promote 26 diastereomeric interactions of hemithioacetal adducts with the local protein 27 environment, leading to the (AG)6(AG-G137C-Ala3) nanopore (AG-G137C-Ala = AG-28 G137C-N139A-N121A-N123A) (Fig. 4a). 29 The I res% of 7 different straight-chain aldehydes (i.e., ethanal to octanal) were 30 characterized with the (MK)6(MK-T115C) and (AG)6(AG-G137C) nanopores (Fig. 4b-31 c and Supplementary Tables 4-5). Improved diastereomer discrimination was 32 observed in both nanopores: diastereomeric hemithioacetal adducts from ethanal to 33 octanal were easily separable in both nanopores: 0.3 % < ∆ Ires% < 1.0 %, in the 34 (MK)6(MK-T115C) nanopore (Supplementary Table 4) and 0.4 % < ∆ Ires% < 1.6 %, in 35 the (AG) 6(AG-G137C) nanopore (Supplementary Table 5). Good current level 36 separation was achieved across all diastereomeric adducts in the (MK) 6(MK-T115C) 37 nanoreactor, whereas some overlap was seen in the (AG) 6(AG-G137C) nanoreactor 38 (e.g., I res% = 95.8 ± 0.1 % for diastereomer A of pentanal and 95.9 ± 0.1 % for 39 diastereomer A of hexanal). 40 Current level resolution of diastereomeric hemithioacetal adducts subsequently 41 allowed for structural isomer resolution. In the (AG) 6(AG-G137C) nanopore, butanal 42 and 2-methylpropanal could be distinguished based on diastereomers B (i.e., Δ Ires% 43 of 0.25 ± 0.02 % for diastereomers B of butanal and 2-methylpropanal, N = 3 pores, 44 >35 events for each aldehyde) (Supplementary Fig. 15). In contrast, within the 45 (AG)6(AG-G137C-Ala3) nanopore, diastereomer pairs were better resolved for both 46 butanal or 2-methylpropanal; distinct separation between diastereomers A enabled 47 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 18, 2024. ; https://doi.org/10.1101/2024.10.15.618413doi: bioRxiv preprint 8 clear resolution of these chain isomers (i.e., Δ Ires% of 0.42 ± 0.03 % for 1 diastereomers A of butanal and 2-methylpropanal, N = 4 pores, >60 events for each 2 aldehyde). In the (MK) 6(MK-T115C) nanopore, butanal and 2-methylpropanal could 3 be distinguished using diastereomer A of 2-methylpropanal and diastereomer B for 4 butanal (i.e., Δ Ires% of 0.43 ± 0.03 % for diastereomer B of butanal and diastereomer 5 A of 2-methylpropanal, N = 2 pores, >50 events for each aldehyde). As a proof of 6 concept, simultaneous detection of propanal, butanal, 2-methylpropanal and 7 pentanal was demonstrated in the (MK) 6(MK-T115C) nanopore (Fig. 4d). As the 8 current level blockades for pentanal and propanal do not overlap with those of 9 butanal and 2-methylpropanal, all four aldehydes could be differentiated from I res% 10 alone. 11 Enzyme-assisted differential sensing of alcohols and aldehydes 12 Nanopore covalent sensing is inherently selective for a targeted class of analytes. 13 For example, in a mixture of alcohols (1-pentanol, 1-hexanol and 1-heptanol) and 14 aldehydes (propanal and butanal), only the aldehydes were picked up by the 15 (AG)6(AG-T115C) nanopore sensor (Fig. 5a). After treatment with an engineered 16 alcohol oxidase25, additional aldehyde species—pentanal, hexanal, and heptanal—17 were identified (Fig. 5b), indirectly revealing the presence of alcohols in the original 18 sample. This simple functional group conversion step, coupled to nanopore sensing, 19 therefore allowed for differential single-molecule detection of alcohols and aldehydes 20 in a mixture—a powerful strategy worth further development. 21 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 18, 2024. ; https://doi.org/10.1101/2024.10.15.618413doi: bioRxiv preprint 9

1 The present work exploited thiol-aldehyde chemistry, a previously underexplored 2 dynamic covalent chemistry, within protein nanopores to target the sensing of 3 aldehydes, a key class of chemicals ubiquitous in daily life. We achieved single-4 molecule identification of 10 straight-chain, branched-chain, and aromatic aldehydes, 5 which represent potential biomarkers for lung cancer,5 Crohn’s disease,6 SARS-CoV-6 2,8 as well as atmospheric pollutants, 2 and beverage impurities. 3 Our approach 7 resolved straight-chain aldehydes that differed by single CH 2 groups, members of 8 hemithioacetal diastereomeric pairs, and structural isomers—challenges often faced 9 by conventional small-molecule detection methods. In particular, we systematically 10 engineered the protein environment surrounding the covalent sensing site, achieving 11 improved separation for diastereomeric adducts by promoting interactions with the 12 pore interior. Simple engineering of the sensing region thus holds promise for 13 improving small-molecule covalent detection with a given nanopore scaffold. 14 As a proof of concept, we established ratiometric profiles of mixed aldehydes at mM 15 concentrations with >500 events collected in 10 min using a single nanopore. Given 16 that common biological samples contain aldehydes at µM concentrations26,27, routine 17

can be applied to capture and concentrate analytes.28 To analyze biological 18 samples directly, the detection frequency could be further increased by >500,000-19 fold using nanopore mutants engineered with more than one covalent sensing sites 20 (e.g., (AG-T115C) 7), higher recoding pH values, elevated temperatures, and 21 nanopore sensing devices containing arrays of pores (e.g. the PromethION device 22 contains up to 2675 nanopores per flow cell, and 48 flow cells per machine). Based 23 on our findings, we envision an accessible nanopore platform for rapid aldehyde 24 detection, paving the way for applications in disease diagnosis, environmental 25 monitoring, and the quality control of food and beverages. 26 Moving forward, the detection of other biologically relevant chemical classes can be 27 explored by capitalizing the aldehyde-sensing system here. As central metabolites, 28 aldehydes are generated by a wide range of enzymes from various functional groups, 29 including carboxylic acids, primary alcohols, and primary amines 29. Many of these 30 convertible chemical classes are also potential disease biomarkers. For example, 1-31 pentanol, detected as 1-pentanal in this work, has been found in the exhaled breath 32 by lung cancer patients but not healthy individuals 5. While we have demonstrated 33 alcohol sensing after enzymatic conversion to aldehydes, detecting other chemical 34 classes requires careful identification of suitable enzymes, particularly with respect to 35 substrate scope and catalytic efficiency. In the long term, by leveraging thiol-36 aldehyde sensing chemistry, we envision a versatile sensing workflow that employs 37 a suite of aldehyde-converting reagents, which will produce informative single-38 molecule profiles of various chemical classes for nanopore diagnostics. 39 40 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 18, 2024. ; https://doi.org/10.1101/2024.10.15.618413doi: bioRxiv preprint 10

1 Preparation of nanopore sensors 2 Nanopore heteroheptamers were prepared according to the procedure below, which 3 is a modification of a method previously reported. 30 αHL monomers were prepared 4 with an E. coli in vitro transcription and translation (IVTT) system ( E. coli T7 S30 5 Extract System for Circular DNA, Cat #L1130, Promega). A standard reaction 6 comprised: DNA plasmid mixture (<4 µg, plasmids encoding cysteine-free and 7 cysteine-containing subunits were in an 8:1 ratio), amino acid mixture without 8 methionine (5 µL, as supplied in the kit), S30 premix without amino acids (20 µL, as 9 supplied in the kit), [ 35S]methionine (2 µL, 1200 Ci/mmol, 15 mCi/mL, MP 10 Biomedicals), T7 S30 extract for circular DNA (15 µL, as supplied in the kit), and 11 rabbit red blood cell membranes (2 µL, ~ 1 mg protein/mL). The reaction mixture was 12 incubated at 37°C for 2 h. 13 αHL heptamers containing different numbers of mutant subunits were separated in 14 the gel based on their electrophoretic mobilities which were determined by the 15 number of octa-aspartate (D8) tails present (i.e., each cysteine-bearing mutant 16 subunit contained a D8 tail). Hence, the top band corresponded to homoheptamers 17 bearing no octa-aspartate tail (i.e., AG- αHL)7, the second band corresponded to 18 heteroheptamers bearing a single octa-aspartate tail (i.e., (mutant-D8) 1(AG-αHL)6) 19 and so on, with consecutive bands having an aspartate-tail-free subunit replaced 20 with a mutant subunit bearing an octa-aspartate tail. In this work, the desired protein 21 pore containing a single cysteine residue was extracted from the second band from 22 the top. 23 Single-channel electrical recordings 24 Single-channel recordings were carried out in a planar bilayer apparatus as 25 previously described. 31 A single αHL pore was allowed to insert into the bilayer. 26 Aldehyde substrates were introduced from the trans compartment. Experiments were 27 conducted using recording buffer containing 2 M KCl, 200 mM PIPES and 20 μ M 28 EDTA titrated to pH 6.8. Aldehyde solutions were prepared with recording buffer and 29 titrated to pH 6.8. Single-channel recordings were conducted with a coverslip placed 30 atop the recording chamber. 31 Ionic currents were recorded by using a patch clamp amplifier (Axopatch 200B, Axon 32 Instruments), and filtered with a low-pass Bessel filter (80 dB/decade) with a corner 33 frequency of 10 kHz. Signals were digitized with a Digidata 1320A digitizer 34 (Molecular Devices) at an acquisition frequency of 50 kHz. The current traces were 35 processed with Clampfit 10.7 (Molecular Devices). Current traces were idealized by 36 using Clampfit 10.7 (Molecular Devices). The idealized data were analyzed with QuB 37 2.0 software (www.qub.buffalo.edu). 32 Dwell time analysis and rate constant 38 determinations were performed by using the maximum interval likelihood (MIL) 39 algorithm of QuB.33 40 41 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 18, 2024. ; https://doi.org/10.1101/2024.10.15.618413doi: bioRxiv preprint 11 1 Fig. 1: Single-molecule covalent sensing of aldehydes. a, The engineered α-hemolysin 2 (αHL) nanopore, (AG) 6(AG-T115C), bears a single cysteine at position 115 on one of the 3 seven subunits (boxed). b, An aldehyde molecule was detected upon reaction with the 4 cysteine thiol to form a hemithioacetal adduct. Aldehydes are hydrated to varying extents in 5 aqueous solutions. Concentrations of free aldehy de were calculated by using the hydration 6 constant Khyd determined by 1H NMR. c, Single-channel recordings at −50 mV (trans) with 7 0.0, 3.0, 5.8, 8.5 or 11 mM propanal (trans) in 2 M KCl, 200 mM PIPES and 20 μ M EDTA at 8 pH 6.8. Signals were low-pass filtered at 10 kHz and sampled at 50 /i2 kHz. Traces were 9 further filtered at 100 Hz for display. Current levels correspond to the hemithioacetal adducts 10 formed (level 2, orange) and the unoccupied nanopore (level 1, blue). d, Rates of adduct 11 formation (v on) and dissociation ( voff), recorded with individual pores, plotted against total 12 propanal concentration showing bimolecular kinetics for the forward association step (level 1 13 to 2), and unimolecular kinetics for the reverse dissociation step (level 2 to 1). 14 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 18, 2024. ; https://doi.org/10.1101/2024.10.15.618413doi: bioRxiv preprint 12 1 Fig. 2: Thiol-aldehyde chemistry within the (AG) 6(AG-T115C) nanopore. a, Aldehydes 2 reacted with the cysteine at position 115 to generate hemithioacetal adducts reversibly. 3 Diastereomeric adducts were resolved for heptanal and octanal. Diastereomer A is arbitrarily 4 assigned with a higher I res% than diastereomer B. Recording conditions: −50 mV (trans) with 5 2 M KCl, 200 mM PIPES and 20 μ M EDTA at pH 6.8. Aldehyde concentrations (trans): 5.8 6 mM ethanal, 5.8 mM propanal, 6.1 mM 2-methylpropanal, 5.8 mM butanal, 5.7 mM 7 benzaldehyde, 5.5 mM pentanal, 5.5 mM phenylacetaldehyde, 2.7 mM hexanal, 1.5 mM 8 heptanal, 1.3 mM octanal. Signals were low-pass filtered at 10 kHz and sampled at 50 /i2 kHz. 9 Traces were further filtered at 100 Hz for display. b, I res% values for the hemithioacetal 10 adducts. c, Rate constants for adduct formation corrected for hydration ( kon/i2 ) and 11 dissociation ( koff). Errors are standard deviations across 3 different nanopores. Corrected 12 rate constants for adduct formation were obtained from apparent rate constants ( kon) by 13 using the equation: k on/i2 = kon × (1 + K hyd). Rates of adduct formation for octanal were not 14 reported due to the poor solubility of octanal in recording solutions, precluding accurate 15 determinations. 16 17 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 18, 2024. ; https://doi.org/10.1101/2024.10.15.618413doi: bioRxiv preprint 13 1 Fig. 3: Single-molecule profiling of aldehyde mixtures. a, Simultaneous detection of 7 2 straight-chain aldehydes at the single-molecule level. Total aldehyde concentrations (trans): 3 0.7 mM ethanal, 0.7 mM propanal, 0.7 mM butanal, 0.5 mM pentanal, 0.6 mM hexanal, 0.5 4 mM heptanal, and 0.3 mM octanal. b, Performance of the Random Forest model on the test 5 set is shown as a confusion matrix. The accuracy on the test set is 0.98, calculated as the 6 ratio of correctly classified datapoints over all datapoints in the dataset. c, Single-channel 7 recordings of pentanal and butanal at various ratios of total concentrations ([aldehyde] tot) as 8 noted on the panel. Recording conditions: − 50 mV (trans) with 2 M KCl, 200 mM PIPES and 9 20 μ M EDTA at pH 6.8. Signals were low-pass filtered at 10 kHz and sampled at 50 /i2 kHz. 10 Traces were further filtered at 100 Hz for display. 11 12 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 18, 2024. ; https://doi.org/10.1101/2024.10.15.618413doi: bioRxiv preprint 14 1 2 Fig. 4: Diastereomer discrimination in engineered nanopores facilitated structural 3 isomer resolution. a, Rational engineering of nanopor es to enhance diastereomer 4 discrimination through two parallel strategies: Left: narrowing the pore diameter around the 5 sensing group (yellow); Right: moving the sensing group to the narrowest internal site. 6 Mutations are highlighted relative to the WT background in all subunits (blue) or in the 7 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 18, 2024. ; https://doi.org/10.1101/2024.10.15.618413doi: bioRxiv preprint 15 cysteine-bearing subunit (red). b-c, Within the (MK) 6(MK-T115C) nanopore, diastereomeric 1 adducts were resolved for all straight-chain aldehydes tested. Diastereomer A was arbitrarily 2 assigned with a higher I res% than diastereomer B. d, Within the (AG) 6(AG-T115C) nanopore, 3 diastereomeric adducts were resolved for only heptanal and octanal. e, Within the (AG)6(AG-4 G137C) nanopore, diastereomeric adducts were resolved for all straight-chain aldehydes 5 tested. f, A pair of chain isomers, butanal and 2- methylpropanal, were distinguished by 6 diastereomers B in the (AG) 6(AG-G137C) nanopore or by diastereomers A in the (AG) 6(AG-7 G137C-Ala3) nanopore. g, Propanal, butanal, 2-methylpropanal, and pentanal were 8 detected within a single (MK) 6(MK-T115C) pore. Butanal and 2-methylpropanal were 9 distinguished by the diastereomer A of 2-methylpropanal and the diastereomer B of butanal. 10 Recording conditions: −50 mV (trans) with 2 M KCl, 200 mM PIPES and 20 μ M EDTA at pH 11 6.8. Signals were low-pass filtered at 10 kHz and sampled at 50 /i2 kHz. Traces were further 12 filtered at 50 Hz for display. 13 14 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 18, 2024. ; https://doi.org/10.1101/2024.10.15.618413doi: bioRxiv preprint 16 1 Fig. 5: Enzyme-facilitated indirect alcohol detection. The (AG) 6(AG-T115C) nanopore 2 selectively detected aldehydes in an aldehyde-alcohol mixture (Left). After treatment with an 3 engineered alcohol oxidase (AcCO6), additi onal aldehydes species—pentanal, hexanal, 4 heptanal—were identified (Right), which indi rectly confirmed the presence of the 5 corresponding alcohols in the original sample. Recording conditions: −50 mV (trans) with 2 6 M KCl, 200 mM PIPES and 20 μ M EDTA at pH 6.8. Signals were low-pass filtered at 10 kHz 7 and sampled at 50/i2 kHz. Traces were further filtered at 100 Hz for display. 8 9 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 18, 2024. ; https://doi.org/10.1101/2024.10.15.618413doi: bioRxiv preprint 17

1 1. Sutaria, S. R., Gori, S. S., Morris, J. D., Xie, Z., Fu, X.-A. & Nantz, M. H. Lipid 2 Peroxidation Produces a Diverse Mixture of Saturated and Unsaturated Aldehydes in 3 Exhaled Breath That Can Serve as Biomarkers of Lung Cancer—A Review. Metabolites 12, 4 561 (2022). DOI: 10.3390/metabo12060561. 5 2. Schauer, J. J., Kleeman, M. J., Cass, G. R. & Simoneit, B. R. T. Measurement of 6 Emissions from Air Pollution Sources. 3. C 1-C29 Organic Compounds from Fireplace 7 Combustion of Wood. Environ. Sci. Technol . 35, 1716-1728 (2001). DOI: 8 10.1021/es001331e. 9 3. Moreira, M. T. G., Pereira, P. R., Aquino, A., Conte-Junior, C. A. & Paschoalin, V. M. F. 10 Aldehyde Accumulation in Aged Alcoholic Beer: Addressing Acetaldehyde Impacts on Upper 11 Aerodigestive Tract Cancer Risks. Int. J. Mol. Sci . 23, 14147 (2022). DOI: 12 10.3390/ijms232214147. 13 4. Smolinska, A., Tedjo, D. I., Blanchet, L., Bodelier, A., Pierik, M. J., Masclee, A. A. M., 14 Dallinga, J., Savelkoul, P. H. M., Jonkers, D. M. A. E., Penders, J. & van Schooten, F. J. 15 Volatile Metabolites in Breath Strongly Correl ate with Gut Microbiome in CD Patients. Anal. 16 Chim. Acta 1025, 1-11 (2018). DOI: 10.1016/j.aca.2018.03.046. 17 5. Rudnicka, J., Kowalkowski, T. & Buszewski, B. Searching for Selected VOCs in Human 18 Breath Samples as Potential Markers of Lung Cancer. Lung Cancer 135, 123-129 (2019). 19 DOI: 10.1016/j.lungcan.2019.02.012. 20 6. Hicks, L. C., Huang, J., Kumar, S., Powles, S. T., Orchard, T. R., Hanna, G. B. & Williams, 21 H. R. T. Analysis of Exhaled Breath Volatile Organic Compounds in Inflammatory Bowel 22 Disease: A Pilot Study. J. Crohn's Colitis 9, 731-737 (2015). DOI: 10.1093/ecco-jcc/jjv102. 23 7. Ibrahim, W., Wilde, M. J., Cordell, R. L., Richardson, M., Salman, D., Free, R. C., Zhao, 24 B., Singapuri, A., Hargadon, B., Gaillard, E. A., Suzuki, T., Ng, L. L., Coats, T., Thomas, P., 25 Monks, P. S., Brightling, C. E., Greening, N. J. & Siddiqui, S. Visualization of Exhaled Breath 26 Metabolites Reveals Distinct Diagnostic Signatures for Acute Cardiorespiratory 27 Breathlessness. Sci. Transl. Med. 14, eabl5849 (2022). DOI: 10.1126/scitranslmed.abl5849. 28 8. Ibrahim, W., Cordell, R. L., Wilde, M. J., Richardson, M., Carr, L., Devi Dasi, A. S., 29 Hargadon, B., Free, R. C., Monks, P. S., Brightling, C. E., Greening, N. J. & Siddiqui, S. 30 Diagnosis of COVID-19 by Exhaled Breath Analysis Using Gas Chromatography–Mass 31 Spectrometry. ERJ Open Res. 7, 00139-2021 (2021). DOI: 10.1183/23120541.00139-2021. 32 9. Su, R.; Yang, T., Zhang, X., Li, N., Zhai, X. & Chen, H. Mass Spectrometry for Breath 33 Analysis. TrAC Trends Anal. Chem. 158, 116823 (2023). DOI: 10.1016/j.trac.2022.116823. 34 10. Drabiń ska, N., Flynn, C., Ratcliffe, N., Belluomo, I., Myridakis, A., Gould, O. & Costello, 35 B. D. L. A Literature Survey of All Volatiles from Healthy Human Breath and Bodily Fluids: 36 The Human Volatilome. J. Breath Res. 15, 034001 (2021). DOI: 10.1088/1752-7163/abf1d0. 37 11. Fuchs, P., Loeseken, C., Schubert, J. K. & Miekisch, W. Breath Gas Aldehydes as 38 Biomarkers of Lung Cancer. Int. J. Cancer 126, 2663-2670 (2010). DOI: 10.1002/ijc.24970. 39 12. Guadagni, R., Miraglia, N., Simonelli, A., Silvestre, A., Lamberti, M., Feola, D., 40 Acampora, A. & Sannolo, N. Solid-Phase Microextraction–Gas Chromatography–Mass 41 Spectrometry Method Validation for the Dete rmination of Endogenous Substances: Urinary 42 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 18, 2024. ; https://doi.org/10.1101/2024.10.15.618413doi: bioRxiv preprint 18 Hexanal and Heptanal as Lung Tumor Biomarkers. Anal. Chim. Acta 701, 29-36 (2011). 1 DOI: 10.1016/j.aca.2011.05.035. 2 13. Bayley, H., Luchian, T., Shin, S.-H. & Steffensen, M. B. “Single-Molecule Covalent 3 Chemistry in a Protein Nanoreactor” in Single Molecules and Nanotechnology (Springer, 4 2008). 5 14. Wu, H.-C. & Bayley, H. Single-Molecule Detection of Nitrogen Mustards by Covalent 6 Reaction within a Protein Nanopore. J. Am. Chem. Soc . 130, 6813-6819 (2008). DOI: 7 10.1021/ja8004607. 8 15. Boersma, A. J. & Bayley, H. Continuous Stochastic Detection of Amino Acid 9 Enantiomers with a Protein Nanopore. Angew. Chem. Int. Ed. 51, 9606-9609 (2012). DOI: 10 10.1002/ange.201205687. 11 16. Boersma, A. J., Brain, K. L. & Bayley, H. Real-Time Stochastic Detection of Multiple 12 Neurotransmitters with a Protein Nanopore. ACS Nano 6, 5304–5308 (2012). DOI: 13 10.1021/nn301125y. 14 17. Ramsay, W. J. & Bayley, H. Single-Molecule Determination of the Isomers of D-Glucose 15 and D-Fructose That Bind to Boronic Acids. Angew. Chem. Int. Ed. 57, 2841-2845 (2018). 16 DOI: 10.1002/anie.201712740. 17 18. Liu, Y., Zhang, S., Wang, Y., Wang, L., Cao, Z., Sun, W., Fan, P., Chen, H.-Y. & Huang, 18 S. Nanopore Identification of Alditol Epimers and Their Application in Rapid Analysis of 19 Alditol-Containing Drinks and Healthcare Products. JACS 144, 13717-13728 (2022). DOI: 20 10.1021/jacs.2c04595. 21 19. Fan, P., Cao, Z., Zhang, S., Wang, Y., Xiao, Y., Jia, W., Zhang, P. & Huang, 22 S. Nanopore Analysis of Cis-diols in Fruits. Nat. Commun. 15, 1969 (2024). DOI: 23 10.1038/s41467-024-46303-x. 24 20. Wang, K., Zhang, S., Zhou, X., Yang, X., Li, X., Wang, Y., Fan, P., Xiao, Y., Sun, W., 25 Zhang, P., Li, W. & Huang, S . Unambiguous Discrimination of All 20 Proteinogenic Amino 26 Acids and Their Modifications by Nanopore. Nat. Methods 21, 92–101 (2024). DOI: 27 10.1038/s41592-023-02021-8. 28 21. Zhang, M., Tang, C., Wang, Z., Chen, S., Zhang, D., Li, K., Sun, K., Zhao, C., Wang, Y., 29 Xu, M., Dai, L., Lu, G., Shi, H., Ren, H., Chen, L. & Geng, J . Real-Time Detection of 20 30 Amino Acids and Discrimination of Pathologically Relevant Peptides with Functionalized 31 Nanopore. Nat. Methods 21, 609–618 (2024). DOI: 10.1038/s41592-024-02208-7. 32 22. Choi, L. S., Mach, T. & Bayley, H. Rates and Stoichiometries of Metal Ion Probes of 33 Cysteine Residues within Ion Channels. Biophys. J . 105, 356-364 (2013). DOI: 34 10.1016/j.bpj.2013.04.046. 35 23. Hilal, S. H., Bornander, L. L. & Carreira, L. A. Hydration Equilibrium Constants of 36 Aldehydes, Ketones and Quinazolines. QSAR Comb. Sci. 24, 631-638 (2005). DOI: 37 10.1002/qsar.200430913. 38 24. Daw, G., Regan, A. C., Watt, C. I. F. & Wood, E. Steric Effects and Mechanism in the 39 Formation of Hemi-acetals from Aliphatic Aldehydes. J. Phys. Org. Chem. 26, 1048-1057 40 (2013). DOI: 10.1002/poc.3138. 41 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 18, 2024. ; https://doi.org/10.1101/2024.10.15.618413doi: bioRxiv preprint 19 25. Heath, R. S., Birmingham, W. R., Thompson, M. P., Taglieber, A., Daviet, L. & Turner, N. 1 J. An Engineered Alcohol Oxidase for the Oxidation of Primary Alcohols. ChemBioChem 20, 2 276-281 (2019). DOI: 10.1002/cbic.201800556. 3 26. Silva, L. K., Espenship, M. F., Newman, C. A., Zhang, L., Zhu, W., Blount, B. C. & De 4 Jesús, V. R. Assessment of Serum Concentrations of 12 Aldehydes in the U.S. Population 5 from the 2013-2014 National Health and Nutrition Examination Survey. Environ. Sci. 6 Technol. 55, 5076-5083 (2021). DOI: 10.1021/acs.est.0c07294. 7 27. Yoshioka, N., Kurata, K., Takahashi, T., Ariizumi, M., Mori, T., Fujisawa, H., Kameyama, 8 N. & Okuyama, Y. Body Odour Aldehyde Reduction by Acetic Acid Bacterial Extract 9 including Enzymes: Alcohol Dehydrogenase and Aldehyde Dehydrogenase. Int. J. Cosmet. 10 Sci. 40, 425-428 (2018). DOI: 10.1111/ics.12473. 11 28. Hamidi, S. Recent Advances in Solid-P hase Extraction as a Platform for Sample 12 Preparation in Biomarker Assay. Crit. Rev. Anal. Chem. 53, 199-210 (2023). DOI: 13 10.1080/10408347.2021.1947771. 14 29. Schober L., Dobiašová H., Jurkaš V., Parmeggiani F., Rudroff F., Winkler M. Enzymatic 15 Reactions Towards Aldehydes: An overview. Flavour Fragr. J. 38, 221-242 (2023). 16 DOI:10.1002/ffj.3739 17 30. Braha, O., Walker, B., Cheley, S., Kasianowicz, J. J., Song, L., Gouaux, J.E. & Bayley, 18 H. Designed Protein Pores as Components for Biosensors. Chem. Biol. 4, 497-505 (1997). 19 DOI: 10.1016/s1074-5521(97)90321-5. 20 31. Montal, M. & Mueller, P. Formation of Bimolecular Membranes from Lipid Monolayers 21 and a Study of their Electrical Properties. Proc. Natl. Acad. Sci. USA 69, 3561-3566 (1972). 22 DOI: 10.1073/pnas.69.12.3561. 23 32. Nicolai, C. & Sachs, F. Solving Ion Channel Kinetics with the QuB Software. Biophys. 24 Rev. Lett. 08, 191–211 (2013). DOI: 10.1142/S1793048013300053. 25 33. Qin, F., Auerbach, A. & Sachs, F. Estimating Single-Channel Kinetic Parameters from 26 Idealized Patch-Clamp Data Containing Missed Events. Biophys. J. 70, 264–280 (1996). 27 DOI: 10.1016/S0006-3495(96)79568-1. 28 29 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 18, 2024. ; https://doi.org/10.1101/2024.10.15.618413doi: bioRxiv preprint 20

1 This research was supported by the Bill & Melinda Gates Foundation, a European 2 Research Council Advanced Grant (SYNTISU), a European Research Council 3 Starting Grant (NANOPRO). We thank Steven Barry at the Physical and Theoretical 4 Chemistry Laboratory, University of Oxford for fabricating the recording chamber. 5 Author information 6 Contributions 7 H.B. and Y.Q. conceived and supervised the project. L.E.M. and Z.H.L. prepared the 8 proteins and conducted the single-channel recording experiments. Z.B. prepared the 9 script for machine learning and performed all molecular dynamics simulations. Y.Y. 10 prepared the AcCO6 proteins. L.E.M., Z.H.L., Z.B. and Y.Q. performed the data 11 analysis. L.E.M., Z.H.L., Y.Y., Z.B., H.B. and Y.Q. wrote the manuscript. 12 Corresponding authors 13 Correspondence to Hagan Bayley or Yujia Qing 14 Ethics declarations 15 Competing interests 16 H.B. is the founder of, a consultant for and a shareholder of Oxford Nanopore 17 Technologies, a company engaged in the development of nanopore sensing and 18 sequencing technologies. L.E.M., Z.H.L., H.B. and Y.Q. have filed patents describing 19 the engineered nanopores and their applications in small-molecule covalent sensing. 20 Z.B. and Y.Y. are listed as contributors to the IP. 21 22 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 18, 2024. ; https://doi.org/10.1101/2024.10.15.618413doi: bioRxiv preprint