{"id":"3fdb7a49-f89f-41a4-850c-2be4e41d1cd4","arxiv_id":"2502.05563","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Localized NMR shows a monoclonal antibody near an oil-water interface has broader lines, shorter T2, and roughly half the apparent diffusion speed of bulk antibody.","lead":"A concentrated monoclonal antibody solution was scanned with MRI-style NMR in thin slices approaching an oil-water boundary. The slices nearest the boundary showed broader NMR peaks, faster spin-spin relaxation, and slower apparent diffusion, suggesting the antibody interacts with the interface and may cluster there.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The factor-of-two interface ADC drop cannot be produced by the reported nm-to-sub-µm interfacial layer; without a large slow pool or susceptibility control, the aggregation/exchange conclusion is unsupported.","rationale":"The paper is careful in places: it uses a water SSI control to show that magnetic susceptibility contributes to line broadening, and it explicitly acknowledges that the voxel mixes bulk and interfacial regions and that direct particle formation cannot be assessed. Those are real strengths. However, the central mechanistic conclusion—that slower ADC and reduced T2 near the interface reflect mAb aggregation/multilayer formation and possible exchange with bulk—depends on the unstated assumption that the interfacial material contributes enough signal to shift the voxel-averaged ADC by a factor of two. Simple arithmetic from the manuscript's own voxel dimensions and cited interphase thickness shows that a 200–300 nm layer contains far too little mAb mass to do so. This makes the interpretation underdetermined even before considering susceptibility artifacts in T2 and diffusion, which are acknowledged for line broadening but not controlled. The reader's weakest_assumption identified the general risk from susceptibility and concentration gradients and voxel mixing; the present concern sharpens that into a quantitative volume-fraction inconsistency. The appropriate disposition remains CONDITIONAL: the observations are potentially valuable, but the aggregation/exchange claim needs either a direct slow-pool measurement, a spatial map of ADC as a function of offset, or a softened conclusion. Since the reader already requested conditions, the verdict is unchanged.","tokens_in":16292,"tokens_out":12294,"duration_ms":138926,"concrete_test":"Reanalyze the existing interface DW-REPRESS attenuation data (Fig. 6c) with a two-pool model S(b) = f·exp(-bD_slow) + (1-f)·exp(-bD_bulk), fixing D_bulk from the bulk voxel and reporting f and D_slow with confidence intervals. Then compare f with the maximum mAb mass fraction that can reside in a 200–300 nm interphase within the water-side half-voxel (approximately f_max ≈ 0.001 for a 10× local concentration). If the two-pool fit requires f > 0.1 to reproduce the ~50% ADC reduction, the aggregation/exchange interpretation is inconsistent with an interface-localized pool and should be withdrawn or reframed as an unresolved susceptibility or concentration artifact. As a cross-check, acquire ADC/T2 at voxel offsets 0.375, 0.75, and 1.125 mm from the interface; if ADC returns to bulk already at 0.375 mm, the slow signal source is confined to <375 µm and the same fraction argument applies.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section III.B and the Conclusion attribute the ~2× lower ADC and shorter T2 in the interface-centered voxel to mAb association/aggregation and possible exchange with bulk. The load-bearing problem is one of volume fraction. The DW-REPRESS voxel is 1.5×1.5×0.75 mm (Methods; the abstract's 375 µm and the text's 187 µm are inconsistent), so the water-side half-voxel is ~375 µm thick. The interfacial layers/interphase invoked by the authors (Kalonia et al., ~270 nm; text says 200–300 nm) are sub-micron: a 300 nm layer spanning the voxel cross-section is about 0.08% of the water-side voxel volume. Even with a 10× local concentration it contributes <1% of the detected mAb mass. To reduce a two-pool ADC to ~50% of bulk requires roughly half the detected mAb signal to be in a slow pool. No aggregate mass fraction, local concentration measurement, or slow-pool fraction is reported. The manuscript itself says particle formation cannot be assessed by MRI and exchange rate would require isotope labeling. Thus the factor-of-two ADC, and the parallel T2 shift, cannot be assigned to a nm-to-sub-µm interfacial layer; either a much larger slow compartment exists (asserted only as \"interact over long distances\") or the changes are artifacts (susceptibility, concentration, partial-volume). Susceptibility is acknowledged for line broadening but not controlled in T2/D; water ADC being similar is only a partial control because water and mAb have different T2 and D, and water T2 was not measured.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript reports a localized 1H NMR study of a model IgG2a-κ monoclonal antibody near a fluorocarbon-oil/water interface. Using a 21.1-T magnet with REPRESS and diffusion-weighted REPRESS sequences, the authors acquire spectra, T2 relaxation data, and apparent diffusion coefficients (ADC) from a voxel centered at the interface and from bulk solution. They find that the number of chemical resonances and their chemical shifts do not change with proximity to the interface, while line broadening, a lower self-similarity index, differential T2 shortening, and a roughly factor-of-two lower ADC are observed in the interface-centered voxel. The paper interprets these changes as evidence that the mAb associates at the oil-water interface, possibly forming multilayers or small aggregates that exchange with bulk mAb near the interface.","tokens_in":16583,"tokens_out":4677,"duration_ms":48837,"significance":"If the central claim is substantiated, the paper would introduce a useful new in-situ approach for probing mAb behavior at pharmaceutically relevant oil-water interfaces, where existing methods such as neutron reflectivity have contrast limitations for thick interphases. The bulk ADC value is consistent with published mAb diffusion coefficients, the water-signal SSI control is a sensible partial check on susceptibility artifacts, and the dynamic surface-tension data independently confirm interfacial adsorption. The main weakness is quantitative: the reported voxel dimensions and the sub-micrometer interfacial layers invoked by the authors make the factor-of-two ADC change difficult to attribute to the proposed interfacial aggregate pool without additional evidence.","major_comments":[{"comment":"The factor-of-two reduction in the mAb ADC at the interface is presented as evidence of mAb association and aggregate formation, but the volume fraction of the proposed interfacial layer cannot support this effect. The DW-REPRESS voxel is stated as 1.5x1.5x0.75 mm, so the water-side half-voxel is roughly 375 µm thick. The interphase thickness cited from Kalonia et al. is 200-300 nm, which is about 0.1% of that half-voxel; even a tenfold local concentration increase would contribute less than one percent of the detected mAb signal. A two-pool model that reduces the apparent ADC to roughly half the bulk value requires about half of the detected mAb signal to reside in a slow pool, yet no aggregate mass fraction, local concentration profile, or slow-pool fraction is reported. Given that the manuscript itself states that particle formation cannot be assessed by MRI and that exchange-rate quantification would require isotope labeling, the attribution of the ADC drop to aggregates or a thick interphase is not quantitatively supported.","section":"§III.B, Fig. 6(c)"},{"comment":"The voxel size is reported inconsistently: the abstract and Section II.C state 375 µm, the Methods section states 1.5x1.5x0.75 mm, and the text near Fig. 6 and the Conclusion refers to a 187 µm voxel and a 'voxel less than 200 µm'. These numbers differ by up to a factor of four in each dimension, and the discrepancy directly affects the volume-fraction argument and the spatial-resolution claims. The authors should specify the exact voxel dimensions, identify which dimension corresponds to the slice-selection axis, and reconcile the abstract and text.","section":"II.C, III.B, and Conclusion"},{"comment":"No uncertainties or error bars are reported for the diffusion coefficients or T2 values. The factor-of-two ADC difference in Fig. 6(c) and the differential T2 changes in Fig. 8(c) are the central quantitative results, so the paper needs confidence intervals, the number of replicate measurements, and the fitting residuals or goodness-of-fit metrics. Without these, the reader cannot judge whether the apparent differences are statistically significant or whether the mono-exponential fits are actually distinguishable from bi-exponential alternatives.","section":"II.D, Figs. 4(c), 6(c), and 8(c)"},{"comment":"The susceptibility control is incomplete for the D and T2 interpretations. The water SSI analysis shows that susceptibility gradients broaden the water resonance, but the water ADC similarity is not a sufficient control for mAb T2 and ADC because water T2 was not measured and because susceptibility-induced dephasing can attenuate signal and bias T2 and diffusion measurements differently for short-T2 species. To support the claim that the observed T2 shortening and ADC reduction are due to mAb association rather than susceptibility artifacts, the authors should either measure water T2 in the same voxels, use susceptibility-compensated diffusion-weighting schemes, or compare with a susceptibility-matched control interface.","section":"III.B, SSI control and Discussion"},{"comment":"The mono-exponential character of the diffusion attenuation is used to argue that restriction or exchange are unlikely contributors to the slower interfacial ADC. This inference is not valid: a two-site fast-exchange or partial-volume model with a small slow pool can produce a single exponential decay with a reduced apparent diffusion coefficient over the accessible B range. Thus the mono-exponential fit does not exclude the exchange/aggregation interpretation, nor does it rule out a mixture of free and hindered mAb populations.","section":"III.B, Eq. (1)"}],"minor_comments":[{"comment":"The symbol 'um' should be typeset as 'µm' throughout; additionally, the sentence 'The voxel sizes used in these experiments was...' contains a subject-verb agreement error.","section":"Abstract and Section II.C"},{"comment":"The caption reports the diffusion coefficients as D = 6.8 × 10−5 and D = 3.7 × 10−5 mm2/s, while the text compares these to literature values in m2/s; please state both units consistently and define the conversion explicitly.","section":"Fig. 6 caption"},{"comment":"The assignments of the three resonances to specific amino-acid side chains are tentative; consider labeling them as tentative or providing a reference spectrum or additional evidence to support the assignments.","section":"III.B, chemical shift assignments"},{"comment":"The statement that 'the spatial resolution of MRI is on the order 10 µm' appears without a citation and seems inconsistent with the claimed voxel dimensions; please clarify the intended meaning.","section":"III.B, line broadening"}],"recommendation":"major_revision","confidential_remarks":"The paper presents an interesting application of localized NMR to a pharmaceutically relevant problem, and the bulk ADC consistency with literature is a positive sign. However, the central interpretation requires either direct evidence of a large slow-diffusing pool or a more cautious reframing in terms of voxel-averaged apparent mobility changes. The voxel-size inconsistencies and missing uncertainties are fixable but must be addressed before the quantitative claims can be evaluated. I do not see grounds for rejection, but the current manuscript overreaches in attributing the ADC and T2 changes to aggregates and multilayers."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: this is the first spatially resolved 1H NMR study of a mAb at an oil-water interface, and the raw observations are probably real. The bulk diffusion coefficient matches published values, the pendant-drop data confirm adsorption, and the differential T2 is a nice fingerprint. The problem is the leap from 'slower ADC and shorter T2 in a voxel at the interface' to 'mAb aggregates and exchange with bulk.' That leap doesn't survive a volume-fraction check.\n\nWhat's genuinely new: REPRESS and DW-REPRESS localized NMR applied to mAb at an oil-water interface. The spectral line broadening, unchanged chemical shifts, differential T2 across the resonances, and roughly 2x slower ADC are new observations. The paper is careful in places: water SSI control for susceptibility, use of FC-43 to avoid proton background and reduce susceptibility mismatch, and the bulk D is in line with prior mAb values.\n\nThe soft spots, in order of severity. First, the voxel geometry is reported three different ways: 375 µm in the abstract, 1.5x1.5x0.75 mm in Methods, and 187 µm in the text. That's not a typo you can ignore; it changes the interpretation. Second, there are no error bars on D or T2. For a factor-of-two claim you need to know the noise floor. Third, the mechanistic conclusion is quantitatively shaky. The voxel centered at the interface has a water-side half-thickness of ~375 µm (or 187 µm, depending on which line you read). The interfacial layer and the 200-300 nm interphase cited from Kalonia are sub-micron. That layer, even at 10x concentration, accounts for well under 1% of the mAb signal in the voxel. To cut the apparent ADC in half you need about half of the detected mAb to be in a slow pool. The paper hand-waves this with 'interact over long distances,' but there's no mechanism or measurement for a hundreds-of-micrometer slow zone. Susceptibility and concentration gradients are discussed as alternatives, but not controlled in the T2 and D measurements. Water ADC being similar is only a partial control.\n\nThat said, the paper is honest about what it can't do: it notes that MRI cannot directly see sub-visible particles and that exchange rates would need isotope labeling. The observations themselves are worth reporting, but the interpretation should be dialed back to 'localized NMR detects altered mAb dynamics near an oil-water interface' until the slow-pool fraction is quantified.\n\nWho should read this: pharmaceutical scientists working on mAb formulation stability, and NMR people developing interface-localized spectroscopy. I'd send it to peer review - the first-application claim and the method are worth a serious look - but I'd expect reviewers to demand error bars, a consistent voxel description, and either a quantitative model or a softened conclusion.","headline":"First localized NMR look at a mAb near an oil-water interface, with a solid new measurement but an overreach on the aggregate explanation.","tokens_in":17129,"tokens_out":4201,"would_cite":false,"duration_ms":40437,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Localized proton NMR shows that a model monoclonal antibody, in a voxel centered at an oil-water interface, keeps the same chemical resonances but shows line broadening, faster $T_2$ relaxation, and a roughly two-fold lower apparent…","keywords":["monoclonal antibody","NMR spectroscopy","oil-water interface","protein adsorption","T2 relaxation","diffusion-weighted NMR","self-similarity index","interfacial aggregation"],"falsifier":"A decisive check is to measure the water $T_2$ and apparent diffusion coefficient in the same interface voxel using an oil matched to the magnetic susceptibility of water, or a susceptibility-compensated sequence. If water shows the same interfacial drop as the antibody does, then susceptibility gradients, not antibody association, explain the observations; if water is unaffected while the antibody diffusion and relaxation change, the aggregation interpretation is supported. A second check would measure the mAb concentration within the voxel to rule out a local concentration increase masquerading as slower diffusion.","tokens_in":16081,"feed_emoji":"🧲","tokens_out":9938,"duration_ms":90160,"temperature":0.7,"pith_summary":"The paper tries to establish that a monoclonal antibody's interaction with an oil-water interface can be seen in localized NMR measurements even though the antibody's chemical structure appears unchanged. Approaching the interface, the antibody's proton resonances stay at the same frequencies but broaden, relax faster, and diffuse roughly half as fast as in the bulk. The authors interpret this as antibody molecules associating near the interface, forming aggregates or multilayers in a thick interphase, and possibly exchanging with the bulk phase. This matters because antibody drugs encounter such interfaces during manufacturing, storage, and delivery, and aggregation is a known risk to safety and efficacy. The contribution is a demonstration that spatially resolved NMR can probe these interfacial effects in situ at formulation-relevant concentrations.","feed_headline":"Antibodies clump where oil meets water, NMR shows","feed_subtitle":"The antibody keeps its chemical fingerprint but diffuses half as fast near the interface, a sign of aggregation.","key_machinery":"The load-bearing tool is a spatially and spectrally selective proton NMR method, relaxation-enhanced point-resolved spectroscopy (REPRESS) with VAPOR water suppression, plus a diffusion-weighted variant (DW-REPRESS), which delivers spectra, $T_2$ values, and apparent diffusion coefficients from 375-um voxels placed at increasing distances from the oil-water interface. A custom pendant-drop tensiometer independently confirms that the antibody adsorbs to the interface by showing the interfacial tension falling over hundreds of seconds. The self-similarity index (SSI), a cross-correlation score between spectra, converts line-broadening trends into a single number, and the mono-exponential fit of signal attenuation versus diffusion weighting turns the DW-REPRESS data into an apparent diffusion coefficient.","core_discovery":"At 900 MHz proton NMR on a model IgG2a-κ monoclonal antibody at 36.4 mg/mL against a proton-free fluorocarbon oil, voxels of 375 um show three mAb resonances near 3.9, 3.5, and 2.9 ppm. As the voxel moves from the bulk to the oil-water interface, these chemical shifts do not move and no new resonances appear, but the lines broaden and the self-similarity index falls; a water-only control shows a comparable SSI drop, indicating that magnetic susceptibility differences at the interface account for much of the apparent broadening. Diffusion-weighted measurements give an apparent diffusion coefficient of $6.8\\times10^{-5}\\ \\mathrm{mm^2/s}$ in the bulk and $3.7\\times10^{-5}\\ \\mathrm{mm^2/s}$ in the voxel centered at the interface, a drop of nearly a factor of two, with mono-exponential attenuation. The $T_2$ relaxation times differ across the three resonances and are shortened near the interface for the resonances at 2.9 and 3.5 ppm. The paper reads these observations as evidence that the antibody associates at the interface, possibly forming small aggregates and multilayers in a thick interphase that exchanges with bulk mAb nearby.","pith_inferences":["If the interpretation is right, the ratio of interfacial to bulk apparent diffusion coefficient could serve as a label-free screening metric for interfacial aggregation propensity in antibody formulations.","A susceptibility-matched oil control or susceptibility-compensated diffusion weighting would settle whether the factor-of-two diffusion drop is molecular or field-related; the paper's own water SSI control shows such artifacts are not negligible.","Because the voxel volume mixes adsorbed layers with bulk solution, the measured ADC is a voxel average; varying the voxel size or using diffusion times longer than the interfacial exchange time could separate a slow interfacial population from a bulk population.","The same localized approach should transfer to air-water interfaces, but the larger susceptibility mismatch there makes the susceptibility control even more important."],"forward_implications":["Localized NMR can detect antibody association near an oil-water interface even when chemical shifts show no rearrangement, giving formulations a label-free in-situ probe.","The nearly two-fold lower apparent diffusion coefficient and shortened $T_2$ near the interface imply the interfacial influence extends well beyond a monolayer, consistent with a thick interphase of associated antibody.","Magnetic susceptibility effects must be quantified before line broadening near an interface is attributed to protein structural change, since the water control shows a comparable drop in self-similarity.","Measuring the rate and extent of antibody exchange between the oil-water interface and the bulk remains open and, the authors note, would likely require isotope labeling.","Extending the approach to air-water interfaces, other mAb types, and surfactant-containing formulations is a stated next step."],"supporting_citations":[{"why":"Introduced the 1D proton NMR fingerprinting approach for comparing monoclonal antibody formulations, which this study adapts to spectra near the interface.","marker":"[30]"},{"why":"Defined the self-similarity index (SSI) used here to quantify how much the mAb spectrum changes as the voxel approaches the interface.","marker":"[35]"},{"why":"Documented thick multilayer formation and mAb exchange at solid-liquid interfaces, the picture invoked to explain the slower interfacial diffusion.","marker":"[25]"},{"why":"Measured mAb adsorption at an oil-water interface with neutron reflectivity, the closest prior structural study of this interface.","marker":"[28]"},{"why":"Showed adsorbed mAb at the air-water interface retain globular shape, framing the open question of whether interfacial adsorption alters mAb structure.","marker":"[15]"},{"why":"Supplied the VAPOR water-suppression module that removes the dominant water peak and makes the mAb resonances observable.","marker":"[48]"},{"why":"Provided the diffusion-weighted NMR signal equation and fitting approach used to extract apparent diffusion coefficients.","marker":"[49]"},{"why":"Described the pendant-drop setup and analysis used to confirm mAb adsorption via dynamic interfacial tension.","marker":"[47]"}],"fun_headline_variants":["NMR reveals antibody slowdown at oil-water interface","Antibody diffusion halves at oil-water boundary","Oil-water interface restricts antibody dynamics","NMR shows antibodies move slower near oil-water interface","Antibody rotation and diffusion change at oil-water interface"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument that slower diffusion and shorter $T_2$ near the interface mean antibody association assumes that magnetic-susceptibility gradients at the oil-water interface and local antibody concentration gradients are not responsible for the changes.","fun_headline_variants_meta":{"raw":{"variants":["NMR reveals antibody slowdown at oil-water interface","Antibody diffusion halves at oil-water boundary","Oil-water interface restricts antibody dynamics","NMR shows antibodies move slower near oil-water interface","Antibody rotation and diffusion change at oil-water interface"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000617,"raw_usage":{"total_tokens":2951,"prompt_tokens":1116,"completion_tokens":1835,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":732,"completion_tokens_details":{"reasoning_tokens":1764}},"tokens_in":732,"tokens_out":1835,"duration_ms":14051,"temperature":1.0,"reasoning_tokens":1764,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-08T18:49:11.447536+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A decisive check is to measure the water $T_2$ and apparent diffusion coefficient in the same interface voxel using an oil matched to the magnetic susceptibility of water, or a susceptibility-compensated sequence. If water shows the same interfacial drop as the antibody does, then susceptibility gradients, not antibody association, explain the observations; if water is unaffected while the antibody diffusion and relaxation change, the aggregation interpretation is supported. A second check would measure the mAb concentration within the voxel to rule out a local concentration increase masquerading as slower diffusion.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Introduced the 1D proton NMR fingerprinting approach for comparing monoclonal antibody formulations, which this study adapts to spectra near the interface."},{"cited_title":"B., Rogers, G","cited_arxiv_id":null,"evidence_quote":"Defined the self-similarity index (SSI) used here to quantify how much the mAb spectrum changes as the voxel approaches the interface."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Documented thick multilayer formation and mAb exchange at solid-liquid interfaces, the picture invoked to explain the slower interfacial diffusion."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Measured mAb adsorption at an oil-water interface with neutron reflectivity, the closest prior structural study of this interface."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Showed adsorbed mAb at the air-water interface retain globular shape, framing the open question of whether interfacial adsorption alters mAb structure."},{"cited_title":"& Gruetter, R","cited_arxiv_id":null,"evidence_quote":"Supplied the VAPOR water-suppression module that removes the dominant water peak and makes the mAb resonances observable."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provided the diffusion-weighted NMR signal equation and fitting approach used to extract apparent diffusion coefficients."},{"cited_title":"& Shoele, K","cited_arxiv_id":null,"evidence_quote":"Described the pendant-drop setup and analysis used to confirm mAb adsorption via dynamic interfacial tension."}],"review_version":1}