{"id":"c16f914f-0569-4363-9f8e-6df25fec1bf7","arxiv_id":"2506.00389","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"A neural-network-calibrated thermometer that removes viscosity effects puts the mitochondrial temperature ceiling near 42-43 °C, not the 50 °C reported earlier.","lead":"Researchers calibrated a fluorescent dye with a computer model that separates temperature from viscosity, then mapped temperatures inside living cells. They conclude mitochondria stay below 42-43 °C, contradicting earlier claims of 50 °C mitochondria and supporting the idea that organelle temperature is capped by enzyme stability.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The fixed-cell calibration is the load-bearing step: the matrix is corrected and validated on uniform fixed cells, so any environment-specific offset that differs in living mitochondria transfers directly into the reported 42–43 °C ceiling.","rationale":"I read the paper as an attempt to settle the 'hot mitochondria' controversy with a two-parameter readout and staged calibration. The staged calibration from solution to GPMVs to fixed cells is a real improvement over solution-only calibration, and the paper offers a falsifiable quantitative claim. The main text, however, contains no independent accuracy check for living cells. Fixed cells are chosen because their true temperature is known, but this is exactly what makes them unsuitable as the sole validation target: a correction optimized on fixed cells can absorb any fixed-cell-specific artifact. The stated ±0.05 °C is a residual on the calibration set, not an out-of-sample error, and the abstract's 'eliminating viscosity' wording overstates what the control experiments can establish. I also note an internal ambiguity: the method is described as a neural network, but the operative object is a coefficient matrix; if the network is only the linear 2×2 map implied by the matrix equation, the neural-network framing adds no decoupling beyond linear inversion, and the conditioning of that inversion is not reported. Still, the paper does not make its central claim circularly: the calibration is anchored to external temperature standards, and the ceiling is stated as an upper limit with a few-degree margin over observed maxima. For that reason I do not recommend rejecting the manuscript; the conditional verdict with medium confidence is appropriate. The proposed checks would settle whether the fixed-cell transfer assumption is sound; if they fail, the verdict on the mitochondrial-ceiling claim should move to REJECT or UNVERDICTED.","tokens_in":11584,"tokens_out":6144,"duration_ms":67088,"concrete_test":"Test whether the fixed-cell calibration is environment-neutral by comparing, on the same live-cell images, the mitochondrial temperatures produced by the coefficient matrices after each calibration stage (solution, GPMV, fixed-cell). If switching from the fixed-cell matrix to the solution or GPMV matrix shifts mitochondrial temperatures by more than ~1 °C in a structured way, the final correction is dominated by the fixed-cell environment rather than by a validated physical model. In addition, co-stain living cells with an independent membrane-potential-insensitive thermometer (e.g., a nanodiamond or FRET-based reporter) and compare its mitochondrial-region reading under FCCP/Ca2+ stimulation; if the independent reporter reads above 43 °C or disagrees with the matrix output by more than 2 °C, the viscosity-decoupling claim fails.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim is that, after viscosity decoupling, mitochondrial temperature has an upper limit of 42–43 °C. For this to be true, the coefficient matrix must map the measured pair (R, τ) to (T, η) inside living mitochondria. The final correction of that matrix is performed on fixed cells (Fig. 2d–f), whose true temperature is assumed uniform and equal to ambient. The same fixed-cell dataset is used both to correct and to validate the matrix, so the quoted ±0.05 °C is an in-sample residual, not an out-of-sample error. The transfer to living cells therefore rests on the assumption that the only relevant environmental difference between fixed and living cells is viscosity. That assumption is not demonstrated. The probe is cationic and partially accumulates in mitochondria via membrane potential; living mitochondria have a membrane potential that fixed cells lack, and mitochondrial microenvironments differ in pH, ionic strength, crowding, and local dye concentration. Any of these can change R or τ, and a fixed-cell correction would absorb that change as if it were a temperature offset. The manuscript points to Supplementary Note 6 for pH/ionic-strength controls, but solution-level controls do not establish that the fixed-cell correction is environment-neutral at the living mitochondrial membrane-potential regime. A 2–3 °C hidden offset in either direction would be enough to erase or create the claimed ceiling, and the paper provides no propagated uncertainty and no independent in-cell validation. Thus the 42–43 °C bound is not established by the calibration evidence as presented.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This manuscript reports a neural-network-aided fluorescent thermometry scheme that attempts to decouple temperature and viscosity in live-cell imaging. The probe provides a ratiometric intensity signal R and a fluorescence lifetime τ, and a two-dimensional response matrix is used to map (R, τ) to (T, η). The coefficient matrix is initialized in solution, then successively corrected using giant plasma membrane vesicles (GPMVs) and fixed cells, achieving a claimed precision of ±0.05 °C in fixed cells. The authors apply the method to living HeLa cells and report an intracellular temperature gradient of 4–5 °C, with mitochondria 1–4 °C warmer than the cytoplasm and lysosomes about 1 °C cooler. Under Ca2+ shock, FCCP, and bacterial infection, mitochondrial temperatures increase but reportedly never exceed 42–43 °C, which the authors interpret as an upper limit consistent with enzyme thermal stability. The paper also reports a correlation between mitochondrial fusion/fission morphology and temperature, and dynamic temperature/viscosity changes during S. aureus infection.","tokens_in":11693,"tokens_out":2189,"duration_ms":23315,"significance":"If the central claim holds—that mitochondrial temperature is bounded by ~42–43 °C once viscosity artifacts are removed—this would directly address a long-standing controversy regarding reports of 50–53 °C mitochondria and would support the idea that prior high readings were viscosity-induced artifacts. The calibration chain from solution to GPMVs to fixed cells is a genuine methodological improvement over single-curve calibration, and the use of a two-dimensional (T, η) response is a sensible approach to a known confound. The manuscript also provides a concrete falsifiable prediction (mitochondrial temperature ceiling bounded by enzyme inactivation temperature) and connects temperature to mitochondrial morphology, which could be of broad interest to the biophysics and cell biology communities. However, as discussed in the major comments, the validation of the key quantitative claim rests on an in-sample fixed-cell correction whose transfer to living mitochondria is not independently established.","major_comments":[{"comment":"The ±0.05 °C cellular accuracy is an in-sample residual: the coefficient matrix is corrected using fixed cells at known ambient temperatures and then validated on the same fixed-cell dataset. The text states that after correction 'the temperature measurement accuracy within the cells also reached ±0.05 °C' (Fig. 2f), but this is not an out-of-sample test. To support the claimed accuracy, the authors should use a held-out set of fixed-cell images, cross-validation, or an independent validation target (e.g., GPMVs after fixed-cell correction). Without this, the quoted precision does not establish the accuracy of the matrix in living cells.","section":"Fig. 2d–f and 'The evolution and correction of coefficient matrix'"},{"comment":"The transfer of the corrected coefficient matrix to living cells assumes that viscosity is the only living-cell-specific variable that shifts the probe response. The authors acknowledge the probe is a cationic dye that accumulates in mitochondria partly via membrane potential, and that living mitochondria differ from fixed cells in pH, ionic strength, crowding, and local dye concentration. Solution-level pH and ionic-strength controls (Supplementary Note 6) do not reproduce the living mitochondrial microenvironment, particularly the membrane-potential regime. Any environment-dependent change in R or τ that differs between fixed and living mitochondria would be absorbed as a temperature offset during fixed-cell correction, directly biasing the reported 42–43 °C ceiling. The manuscript should provide a test that distinguishes temperature from such environmental effects in living cells, or explicitly quantify the possible magnitude of this bias.","section":"Fig. 2d–f and transfer to living cells"},{"comment":"The neural-network method is described only schematically. Equation (1) defines a matrix operation, but neither the coefficient matrix nor the training procedure is specified: no architecture, loss function, regularization, number of parameters, training set size, or uncertainty propagation. This makes it impossible to judge whether the claimed decoupling of T and η is genuine or whether the network is overfitting the calibration data. The authors should provide a complete description of the neural network, including the exact relation between the 'coefficient matrix' and network weights, and an analysis of how uncertainties in R and τ propagate to T and η.","section":"Eq. (1) and 'The evolution and correction of coefficient matrix'"},{"comment":"The claim that 'the highest measured temperature do not exceed 42 °C' under maximal Ca2+ shock is based on an unspecified number of cells and pixels, and the text does not report a statistical upper bound or confidence interval. Given that the measured mitochondrial average under stimulation is 40.1 °C and the claimed measurement precision is 0.05 °C, the statement of a hard 42–43 °C ceiling needs a clear statement of the number of independent measurements, the pixel-level distribution, and how the 'highest measured temperature' was derived from that distribution. Without this, the ceiling claim is not quantitatively supported.","section":"Results, 'Temperatures on different organelles'"}],"minor_comments":[{"comment":"There are typographical and grammatical errors, e.g., 'the question on of' and 'promotes better understanding' should read 'promote a better understanding'.","section":"Abstract"},{"comment":"The phrase 'the highest measured temperature do not exceed 42 °C' has a subject-verb agreement error ('temperature do not exceed' should be 'temperatures do not exceed').","section":"Results, 'Temperatures on different organelles'"},{"comment":"The temperature color scale is not shown consistently across panels; adding a unified scale bar and color map would improve readability.","section":"Fig. 4d–e"},{"comment":"The statistical tests behind the asterisks in Fig. 5e are not described in the Methods; the authors should specify the test used and the number of biological replicates.","section":"Results, 'Mitochondrial temperature and function'"},{"comment":"The Discussion states that lysosomes are '2-6 °C lower than mitochondria', which is a wider range than the ~2.5 °C difference reported in the Results; the manuscript should reconcile these values.","section":"Discussion"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is within scope for a biophysics journal and addresses an important controversy. The main concern is whether the fixed-cell calibration genuinely transfers to living mitochondria; the authors should be pushed to provide an independent validation or a quantitative bound on environmental cross-sensitivities. The neural-network methodology also needs sufficient detail for reproducibility."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper is worth your attention because it takes the mitochondrial temperature controversy seriously and does something concrete about it: instead of the usual single-curve calibration in solution, they build a two-parameter response matrix (ratio and lifetime) and evolve it stepwise through solution, GPMVs, and fixed cells. That is a real methodological improvement, and the headline claim—an upper bound near 42-43 °C after viscosity decoupling—is anchored to absolute ambient standards rather than built into the calibration. The result is also externally consistent with the enzyme-inactivation literature they cite (ref 10), which is a meaningful sanity check. I credit the calibration chain as a genuine attempt to address the known critique from Baffou et al.\n\nThe soft spots are real but not fatal. The ±0.05 °C cellular accuracy is computed on the same fixed-cell data used for the final matrix correction, so it is in-sample; the out-of-sample error on living cells is not reported. The transfer to living cells assumes viscosity is the only living-cell-specific confounder that shifts the probe response. That is arguable: the probe is cationic and partially accumulates via membrane potential, and fixed cells have no membrane potential, so pH, ionic strength, or crowding differences could be absorbed by the fixed-cell correction as if they were temperature offsets. The main text's description of the neural network as a linear coefficient matrix also leaves the conditioning of the inversion over the relevant 15 °C and 4-fold-viscosity range unexamined. These concerns do not sink the central conclusion because the paper states the result as an upper limit with a 2-3 °C margin over observed maxima, and the direction of plausible artifacts (e.g., viscosity contamination) would tend to inflate apparent temperature, making the bound conservative. What is missing is a propagated uncertainty estimate on the 42-43 °C ceiling and an independent in-cell validation, such as a different probe or a known heat stimulus with a modeled response.\n\nWho is this for? Anyone working on cellular thermometry, mitochondrial physiology, or the 'hot mitochondria' debate. It deserves a serious referee: the calibration strategy is novel, the data are extensive, and the conclusion is falsifiable and consistent with enzymology. I would encourage review, with the expectation of revision focusing on the uncertainty propagation and a clear discussion of the fixed-cell-to-living-cell transfer assumptions.","headline":"A better-calibrated fluorescent thermometry study that probably gets the mitochondrial temperature ceiling right, but the headline error bar and environment-transfer assumptions need scrutiny before I'd trust the 42-43 °C bound as stated.","tokens_in":12459,"tokens_out":598,"would_cite":true,"duration_ms":7696,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"A neural-network-aided fluorescent thermometer that separates viscosity from temperature finds mitochondrial temperature peaks near 42–43 °C, not the 50–53 °C reported earlier.","keywords":["fluorescent thermometry","mitochondrial temperature","viscosity decoupling","neural network","ratiometric probe","fluorescence lifetime","intracellular temperature gradient","cellular thermogenesis"],"falsifier":"Measure mitochondrial temperature under the same maximal Ca2+ shock using a viscosity-insensitive thermometer, such as nanodiamond or plasmonic thermometry, and check whether the highest local reading still stays below 43 °C; a reading above 43 °C would refute the claimed ceiling. Alternatively, independently vary cytoplasmic viscosity at fixed temperature and see whether the reported temperature changes—if it does, viscosity has not been fully decoupled.","tokens_in":1682,"feed_emoji":"🌡️","tokens_out":1840,"duration_ms":67635,"temperature":0.7,"pith_summary":"This paper tries to settle how hot mitochondria get inside living cells by removing an error source—viscosity—that has muddled fluorescent temperature readings. The authors build a neural-network-assisted thermometer that records two signals (a fluorescence ratio and a lifetime) and converts them, through a two-dimensional calibration matrix refined from solution to cell-like vesicles to fixed cells, into simultaneous maps of temperature and viscosity. With viscosity removed, they report that mitochondria run 1–4 °C warmer than cytoplasm but never exceed about 42–43 °C even under maximal calcium-stimulated heat production, and that lysosomes sit about 1 °C cooler than cytoplasm. If correct, this would mean earlier reports of 50–53 °C mitochondrial temperatures were viscosity artifacts, and it would reconcile cellular heat output with the known ~43 °C enzyme-stability limit.","feed_headline":"Neural-network thermometry caps mitochondria at about 43 °C","feed_subtitle":"A two-signal fluorescent probe separates viscosity from heat, casting doubt on earlier 50 °C mitochondrial readings.","key_machinery":"The load-bearing object is a two-dimensional coefficient matrix relating two measured probe outputs—the ratio R of rhodamine-to-hemicyanin emission and the fluorescence lifetime tau—to two unknowns, temperature T and viscosity eta. A neural network supplies and refines the matrix weights, and the matrix is corrected sequentially against glycerol solutions, giant plasma membrane vesicles, and fixed cells with known uniform temperatures before being applied pixel-by-pixel to living cells. This matters because each of R and tau responds to both temperature and viscosity, so a one-dimensional calibration curve conflates the two; the matrix is what lets the method separate them.","core_discovery":"The central claim is a measurement: once cellular viscosity is explicitly measured and removed, living-cell temperature maps show a persistent 4–5 °C intracellular gradient, with mitochondria averaging 39.0 °C at a 37 °C ambient (40.1 °C after maximal Ca2+ shock), cytoplasm near 37.6 °C, and lysosomes near 36.7 °C; the highest local mitochondrial reading never exceeds 42 °C. The paper interprets this as evidence that the true physiological ceiling for mitochondrial temperature is 42–43 °C, matching the temperature at which mitochondrial respiratory complexes begin to degrade, and that previous estimates of 50–53 °C arose from viscosity-induced errors in one-dimensional fluorescence calibrations.","pith_inferences":["If the 42–43 °C ceiling is real, it implies an active mitochondrial heat-dissipation or thermoregulatory mechanism that keeps the organelle below enzyme-damage threshold even under uncoupling stress; a testable prediction is that stronger heat stimulation would trigger degradation or stress responses rather than higher temperatures.","The same two-dimensional calibration logic could be extended to decouple other confounds—such as pH, ionic strength, or oxygen—by adding more probe channels; the paper's approach predicts this would further improve accuracy in living cells.","The fixed-cell calibration assumes viscosity is the only living-cell-specific factor that shifts the probe response; an independent cross-check with a viscosity-insensitive thermometer under the same calcium-shock protocol would test that assumption directly.","The morphology–temperature correlation could be causal through reactive oxygen species: fragmented mitochondria produce more ROS and waste heat, so antioxidant treatment should lower the temperature of fragmented mitochondria if the link is ROS-driven."],"forward_implications":["Reports of mitochondrial temperatures near 50–53 °C are likely artifacts, and the true ceiling is consistent with enzyme inactivation around 43 °C.","Intracellular heat flow appears directional—mitochondria warmer, cytoplasm intermediate, lysosomes cooler—supporting active mitochondrial thermogenesis and transfer to the rest of the cell.","Mitochondrial morphology and temperature are linked: fused networks run cooler, fragmented mitochondria run hotter, so the fission–fusion balance shapes local heat production.","One-dimensional solution-calibrated fluorescent thermometers risk errors of order 0.5 °C or more inside cells, so simultaneous viscosity measurement is needed for reliable intracellular thermometry.","Bacterial invasion triggers a measurable thermal and rheological stress response, with cell temperature rising about 2 °C and viscosity rising about 1.3 mPa·s within 10–15 minutes."],"supporting_citations":[{"why":"One of the recent organelle-specific thermometry reports in the 50–53 °C range that this paper re-examines as viscosity-affected.","marker":"[4]"},{"why":"The 2018 claim that mitochondria are physiologically maintained near 50 °C—the high-temperature reading this paper's ceiling directly contradicts.","marker":"[5]"},{"why":"The critique that fluorescent temperature imaging in single cells can confuse other physicochemical factors with temperature; motivates the need to remove viscosity.","marker":"[7]"},{"why":"Defines the motion-induced change in emission (MICE) mechanism that makes the probe's rhodamine module sensitive to both temperature and viscosity.","marker":"[9]"},{"why":"Supplies the in vitro result that mitochondrial respiratory complexes degrade above 43 °C, the biochemical limit against which the measured 42–43 °C ceiling is compared.","marker":"[10]"},{"why":"Provides giant plasma membrane vesicles as a cytoplasm-like calibration standard without organelles, used to evolve the coefficient matrix from solution toward cellular environments.","marker":"[25]"},{"why":"Earlier cascade organelle-targeting thermometry mapped temperature transfer from mitochondria to lysosomes, supporting the paper's organelle temperature-gradient interpretation.","marker":"[37]"}],"fun_headline_variants":["Neural net thermometry caps mitochondria at 42°C","Viscosity-aware probes revise mitochondrial heat limit","AI-corrected thermometry: mitochondria stay below 42°C","Mitochondrial temperature cap corrected to 42°C via AI","Neural network thermometry debunks 50°C mitochondria"],"cache_read_input_tokens":14336,"weakest_assumption_plain":"The load-bearing assumption is that correcting the probe's coefficient matrix against fixed cells—whose temperature is assumed uniform and equal to the ambient—removes all living-cell-specific distortions except viscosity; if pH, membrane potential, crowding, or ionic strength shift the probe differently in live cells, the reported 42–43 °C ceiling could be biased.","fun_headline_variants_meta":{"raw":{"variants":["Neural net thermometry caps mitochondria at 42°C","Viscosity-aware probes revise mitochondrial heat limit","AI-corrected thermometry: mitochondria stay below 42°C","Mitochondrial temperature cap corrected to 42°C via AI","Neural network thermometry debunks 50°C mitochondria"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000154,"raw_usage":{"total_tokens":1167,"prompt_tokens":860,"completion_tokens":307,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":476,"completion_tokens_details":{"reasoning_tokens":221}},"tokens_in":476,"tokens_out":307,"duration_ms":3751,"temperature":1.0,"reasoning_tokens":221,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T12:07:36.267131+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure mitochondrial temperature under the same maximal Ca2+ shock using a viscosity-insensitive thermometer, such as nanodiamond or plasmonic thermometry, and check whether the highest local reading still stays below 43 °C; a reading above 43 °C would refute the claimed ceiling. Alternatively, independently vary cytoplasmic viscosity at fixed temperature and see whether the reported temperature changes—if it does, viscosity has not been fully decoupled.","supporting_citations":[{"cited_title":"& Lee, J","cited_arxiv_id":null,"evidence_quote":"One of the recent organelle-specific thermometry reports in the 50–53 °C range that this paper re-examines as viscosity-affected."},{"cited_title":"-T., Jastroch, M., Jacobs, H","cited_arxiv_id":null,"evidence_quote":"The 2018 claim that mitochondria are physiologically maintained near 50 °C—the high-temperature reading this paper's ceiling directly contradicts."},{"cited_title":"& Jullien, L","cited_arxiv_id":null,"evidence_quote":"The critique that fluorescent temperature imaging in single cells can confuse other physicochemical factors with temperature; motivates the need to remove viscosity."},{"cited_title":"L., Wang, L., Liu, X","cited_arxiv_id":null,"evidence_quote":"Defines the motion-induced change in emission (MICE) mechanism that makes the probe's rhodamine module sensitive to both temperature and viscosity."},{"cited_title":"A., Ferná ndez-Silva, P","cited_arxiv_id":null,"evidence_quote":"Supplies the in vitro result that mitochondrial respiratory complexes degrade above 43 °C, the biochemical limit against which the measured 42–43 °C ceiling is compared."},{"cited_title":"-J., Baumgart, T., Schwille, P., Simons, K","cited_arxiv_id":null,"evidence_quote":"Provides giant plasma membrane vesicles as a cytoplasm-like calibration standard without organelles, used to evolve the coefficient matrix from solution toward cellular environments."},{"cited_title":"Spatiotemporally mapping temperature dynamics of lysosomes and mitochondria using cascade organelle -targeting upconversion nanoparticle s","cited_arxiv_id":null,"evidence_quote":"Earlier cascade organelle-targeting thermometry mapped temperature transfer from mitochondria to lysosomes, supporting the paper's organelle temperature-gradient interpretation."}],"review_version":1}