Interference from carbonaceous particles in reactive oxygen species measurements: A case study with Printex 90 and BPEAnit

Enarsson, Jonas; Malmborg, Vilhelm; Miljevic, Branka; Eriksson, Axel; Wang, Hao; Ristovski, Zoran; Wierzbicka, Aneta

doi:10.5194/egusphere-2026-3807

Preprints

https://doi.org/10.5194/egusphere-2026-3807

Preprints

14 Jul 2026

| 14 Jul 2026

Status: this preprint is open for discussion and under review for Atmospheric Measurement Techniques (AMT).

Interference from carbonaceous particles in reactive oxygen species measurements: A case study with Printex 90 and BPEAnit

Jonas Enarsson, Vilhelm Malmborg, Branka Miljevic, Axel Eriksson, Hao Wang, Zoran Ristovski, and Aneta Wierzbicka

Abstract. Reactive oxygen species (ROS) and oxidative potential (OP) are strongly associated with adverse health outcomes and have emerged as a more health-relevant metric for assessing the toxicity of air pollution than particle number or mass concentration. Acellular ROS and OP measurements rely on specific chemical probes that enable quantification through fluorescence or absorbance. During measurements using a real-time ROS monitor with 9,10-bis (phenylethynyl) anthracene-nitroxide (BPEAnit) as the probe, we observed a reduction in the fluorescent response when sampling particles containing high levels of black carbon, ostensibly indicating negative ROS values, although there is scientific evidence that black carbon can generate ROS. In this study, we investigated this interference from black carbon particles in BPEAnit solution in order to understand the underlying mechanisms and propose a method of correcting for the artefacts during ROS measurements.

We suspended Printex 90 particles, known to induce OP, at three different concentrations (1 µg mL^-1, 5 µg mL^-1 and 25 µg mL^-1) in solutions containing either BPEAnit, its fluorescent and non-reactive methylated derivative BPEAnit-Me, or the solvent used in the probe solutions, dimethyl sulfoxide (DMSO), alone. The fluorescence of the suspensions was measured in the flow-through measurement cell from the real-time ROS monitor at different times after the addition of the particles. For suspension in BPEAnit and BPEAnit-Me, we found that fluorescence decreased with increasing particle concentrations, whereas no decrease was observed in solvent-only suspensions. The fluorescence response of BPEAnit with particles was consistently higher than for the same particle concentrations in BPEAnit-Me. Based on the results, we suggest that the main mechanism for the decrease in fluorescence is interactions between the particle surfaces and BPEAnit/BPEAnit-Me, and not light scattering or absorption. We attribute the difference in fluorescence between BPEAnit-Me and BPEAnit to ROS formation.

The observed interaction mechanisms are likely not specific to BPEAnit, and similar artefacts should be considered when assessing ROS or OP of carbonaceous, and possibly other insoluble particles, in both online and offline assays. For online ROS measurements with BPEAnit, we suggest regular use of BPEAnit-Me to detect and potentially correct for artefacts caused by insoluble particles.

Received: 28 Jun 2026 – Discussion started: 14 Jul 2026

Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.

Download & links

Jonas Enarsson, Vilhelm Malmborg, Branka Miljevic, Axel Eriksson, Hao Wang, Zoran Ristovski, and Aneta Wierzbicka

Status: open (until 19 Aug 2026)

Post a comment Subscribe to comment alert

CC1:
'Comment on egusphere-2026-3807', Yangyang Liu, 31 Jul 2026 reply
General comments
The manuscript egusphere-2026-3807, entitled “Interference from carbonaceous particles in reactive oxygen species measurements: A case study with Printex 90 and BPEAnit,” investigates the unexpected suppression of fluorescence observed when carbonaceous particles are analyzed using the BPEAnit-based Particle Into Nitroxide Quencher instrument. The authors compare Printex 90 suspensions containing the reactive probe BPEAnit, its fluorescent derivative BPEAnit-Me, and DMSO controls, and they attribute the decreased fluorescence primarily to particle–probe interactions rather than optical scattering or absorption. They further propose using BPEAnit-Me measurements and an empirical correction equation to recover the ROS-related signal. The topic is relevant to online aerosol ROS measurements because particle-induced analytical artefacts could produce substantial underestimation or even apparently negative ROS values. The concentration- and time-dependent experiments provide useful preliminary observations. However, the main mechanistic conclusion and the proposed correction approach rely on assumptions that are not experimentally validated. Equivalent adsorption behavior of BPEAnit and BPEAnit-Me has not been demonstrated, the calculated ROS signal is not independently verified, and the optical-interference controls are insufficient. I therefore recommend major revision. But I would leave that to the editor to decide.

Major Concerns
The proposed correction assumes that BPEAnit and BPEAnit-Me experience identical particle–probe interactions because of their structural similarity. I cannot follow this assumption without direct experimental evidence. The two compounds are used at substantially different concentrations—4 µM BPEAnit versus 130 nM BPEAnit-Me—and they differ in nitroxide functionality, electronic structure, radical character, and potentially surface affinity. Matching their initial fluorescence intensities based on quantum yields does not establish equivalent adsorption, quenching, or surface-site competition. This issue is especially important because adsorption onto carbonaceous particles is commonly concentration dependent and may become nonlinear when surface sites approach saturation. The authors themselves acknowledge that the different probe concentrations may cause differential occupation of adsorption sites and overestimation of the calculated ROS response. Therefore, the current correction equation cannot be considered quantitatively reliable.

I am confused by the relationship between Equation 1, Figure 4, and Figure 5. Equation 1 uses two normalized fluorescence ratios and therefore produces a dimensionless quantity. For the example presented in Figure 4, values of approximately 0.84 and 0.63 would yield a corrected value of roughly 0.33. However, Figure 5 reports corrected fluorescence values in arbitrary units ranging from approximately (10^4) to above (10^6). The manuscript does not explain how the dimensionless output from Equation 1 was transformed into these absolute fluorescence values. This inconsistency prevents reproduction of the analysis and raises concerns about whether the correction was applied to normalized or raw fluorescence data. The problem becomes particularly severe at 25 µg mL-1, where the BPEAnit-Me denominator approaches zero and the corrected value becomes mathematically unstable. The authors acknowledge this singularity but nevertheless present the resulting values as meaningful ROS responses.

The authors claim that the higher fluorescence observed for BPEAnit relative to BPEAnit-Me represents ROS generated by Printex 90. However, no independent measurement confirms that the difference is caused by ROS. Differential adsorption, static or dynamic fluorescence quenching, electron transfer, chemical reduction of the nitroxide, or different surface-induced changes in quantum yield could also produce unequal responses.The argument that Printex 90 generates ROS in other assays does not demonstrate that the observed difference in the present DMSO/BPEAnit system is specifically due to ROS. The manuscript currently treats this interpretation as established rather than hypothetical.

The manuscript repeatedly concludes that fluorescence suppression is primarily caused by adsorption or “ad/absorption” of the probes onto Printex 90 surfaces. However, the experiments only demonstrate a decrease in fluorescence; they do not measure adsorption or probe depletion directly. Fluorescence loss alone cannot distinguish among adsorption, fluorescence quenching without adsorption, electron transfer, chemical transformation, aggregation-induced effects, or photochemical degradation. The supporting literature on adsorption of aromatic compounds or fluorescence quenching by graphene does not establish the mechanism for BPEAnit and BPEAnit-Me under the present conditions. Nevertheless, the conclusion states that probe inactivation by particle adsorption is the primary mechanism.

The authors use DMSO with and without Printex 90 to conclude that scattering and absorption are negligible. I cannot follow this logic because DMSO itself has essentially no relevant fluorescence emission comparable to the probe. Comparing two weak DMSO spectra does not directly determine whether Printex 90 attenuates the 450 nm excitation beam or absorbs/scatters BPEAnit emission near 483 and 515 nm. Inner-filter effects must be evaluated in a fluorescent matrix, not solely in the nonfluorescent solvent. Moreover, the optical assessment was not systematically performed across all three particle concentrations. Carbon black is a strong broadband absorber, and even modest attenuation could become significant at 25 µg mL-1 or after deposition inside the flow cell.

The comparison between filtered and unfiltered BPEAnit-Me suspensions is interpreted as evidence that optical effects account for only approximately 3% of the fluorescence reduction. This experiment is not sufficient for that conclusion. First, filtration through a 0.22 µm membrane does not guarantee complete removal of Printex 90 primary particles or smaller aggregates. No post-filtration particle concentration, turbidity, absorbance, or size distribution is reported. Second, if BPEAnit-Me had already adsorbed onto particles before filtration, removal of the particles would also remove the adsorbed fluorophore. Consequently, the fluorescence would not return to its initial value even if optical attenuation had contributed substantially before filtration. The authors later acknowledge that removing particles does not restore fluorescence, but this observation cannot by itself discriminate adsorption from other irreversible processes.

The concentration of BPEAnit-Me was selected to reproduce the baseline fluorescence intensity of 4 µM BPEAnit. While this facilitates visual comparison, it does not establish equivalent analytical behavior. The authors appear to assume that matching the baseline signal and approximately compensating for the quantum-yield difference are sufficient to compare particle-induced losses. This approach neglects differences in detector linearity, fluorescence lifetime, molar absorption coefficient, adsorption affinity, concentration quenching, and the response of each compound to the local carbon surface environment. A 30-fold concentration difference may also lead to very different surface coverage. In addition, the manuscript does not report complete calibration curves, limits of detection, linear ranges, or instrument stability for either compound.

The same suspensions were measured repeatedly at 5 min, 30 min, 1 h, 2 h, 3 h, and 24 h. However, the PINQ measurement cell continuously withdraws liquid at 1 mL min-1, and the measured liquid is discarded. Each measurement lasts at least two minutes. It is therefore unclear how repeated sampling affected the remaining volume, particle concentration, probe concentration, headspace, and oxygen availability. The suspensions were also repeatedly vortexed, which can alter aggregate size, expose new particle surface area, and increase oxygen transfer. I am also unable to determine whether particle-free BPEAnit and BPEAnit-Me controls were measured at every corresponding time point. Normalization to a single initial blank would be inappropriate if the probes undergo storage-related fluorescence drift or auto-oxidation over 24 h.

The authors state that measurements performed within five minutes are comparable to PINQ measurements, although the online instrument reportedly analyzes the aerosol within approximately one minute after collection and mixing. Given that the particle–probe interaction changes rapidly during the initial period, a fivefold difference in reaction time may be analytically important. The manuscript does not provide measurements at the actual PINQ residence time. Manual suspension experiments also do not reproduce aerosol collection, mixing efficiency, particle deposition, tubing losses, cell fouling, or possible memory effects during continuous operation. Carbonaceous particles deposited on tubing or the optical cell could progressively reduce the signal and might be mistaken for solution-phase probe interactions.

The hypothesized mechanism depends on particle surface area, aggregation, surface functional groups, and dispersion behavior, yet the manuscript reports only the manufacturer, stock mass concentration, bath sonication time, and temperature limitation. No characterization of Printex 90 in DMSO is provided. Important information includes primary particle size, aggregate-size distribution after sonication, BET surface area, oxygen-containing surface groups, elemental composition, zeta potential where applicable, and suspension stability over the experimental timescale. Particle settling or progressive aggregation could change the accessible surface area and produce apparent temporal kinetics unrelated to probe chemistry. Preparing a new stock suspension for each replicate may also introduce substantial between-replicate variability if dispersion energy and aggregate breakup are not controlled quantitatively.

Reply
Citation: https://doi.org/10.5194/egusphere-2026-3807-CC1
RC1:
'Comment on egusphere-2026-3807', Anonymous Referee #1, 07 Aug 2026 reply
General Comments
This study by Enarsson et al. investigates the interference and suppression of fluorescence from carbonaceous particles in reactive oxygen species (ROS) when analyzed using the BPEAnit probe in an online instrument.
The topic of this manuscript is of interest and relevance to AMT and is useful for improving the performance of online ROS measurements by reducing the effects of experimental artefacts. The intention and objective of the study are unique.
However, I have several concerns regarding the methodological approach the authors have employed to make their inferences. Some of the authors' interpretations are not directly derived from the results and need more clarity before they can be verified. I think the manuscript needs to be revised to address these concerns and shortcomings before I can recommend it for publication.

Specific Comments
Page 2: Lines 84 – 86: Please cite a couple of papers that mention this interference or have quantified it while measuring OP or ROS. I don't think the four papers cited by the authors in the earlier sentence give interference of insoluble particles as a reason for choosing water-soluble ROS/OP measurement approaches. To my knowledge, at least two of these papers are using an online instrument with microfluidic connections (Wragg et al. 2016 and Puthussery et al. 2018) or narrow flow cells (Utinger et al. 2023). Insoluble particles in their methods can, in turn, affect the results, likely by being deposited onto the microfluidic tubing, leading to clogging, or being difficult to clean when using a narrow flow cell, which may be possible reasons for those studies to not use insoluble PM. The authors have cited studies that have shown interference for non-ROS or non-OP measurements, but this still does not provide a justification as to whether insoluble particles interfere with OP measurements.

Page 3: Lines 93 – 95; Authors are requested to substantiate the claim that BC concentrations are elevated in indoor environments due to candle burning. I don’t fully agree with their assessments as several studies have reported high BC emissions from candles only when the candles are sooting and not burning cleanly. (Fine et al., 1999; Pagels et al., 2009; Stabile et al., 2012; Subramanian et al., 2025) The authors are expected to make this distinction as well, as clean combustion of a candle releases only a very small fraction of BC (Fine et al., 1999; Pagels et al., 2009; Stabile et al., 2012; Subramanian et al., 2025).

Line 139: Have the authors investigated if the insoluble particles are deposited in the tube, as it has a pretty small diameter? This may also lead to a reduction in fluorescence. The authors are requested to provide more details validating this.

Lines 186 - 191: This is not a strong enough analogy for selecting these concentrations without providing more details on the manuscript in preparation. 150 μg/m³ of BC is very high and is likely a concentration obtained from chamber experiments, I believe. These concentrations are environmentally relevant except on rare occasions such as maybe during wildfires, or during highly polluted wintertime PM in China and India.

Moreover, I am not sure how the authors converted the mass concentrations in the air to liquid concentrations. It seems to be that 1 mL of liquid corresponds to 16.67 L of Air according to this statement. How did the authors come about this number? BC only constitutes a small fraction of ambient PM (5 - 20%). Moreover, the study the authors cite is not a study using PM in real-world DCFH-DA, albeit one using nanoparticles.

The authors are requested to cite relevant studies of ambient PM using DCFH-DA or other assays and requested to verify if the concentrations of BC they used are environmentally relevant. I think 25 μg/mL of Printex 90 is a very high concentration that may not be environmentally relevant.

Lines 229 – 235: Please be more specific about the chamber experiments and add more details about the methodology in the SI. What were the concentrations of SOA and soot in the final liquid that was analyzed using fluorescence? In Figure 1, the intensity in the first peak with SOA is only 20% of the background. Were these experiments performed in triplicate, and the LoD of the measurement ascertained? What was the sampling time point for this measurement: 5 mins, 30 mins, 1h, 2 h, 3h, or 24 h?

Lines 243 – 244: DMSO is not fluorescent. Printex 90 particles are likely not fluorescent in the presence of only DMSO, as they need a reducing agent to oxidize and release ROS. Therefore, I am not sure about the inference the authors are trying to make for this statement.

Lines 245 – 250: Why were these measurements only performed at 5 μg/mL Printex concentration and not at all three concentrations? In table A2, the relative decreases (1 – observation) would be 45 and 42%, respectively, instead of 55 and 58%. And are the authors referring to Printex 90 while mentioning particles or different particles?

Line 251: The authors are requested to be specific that they are referring to Printex 90 particles and not SOA or candle emissions here.

Lines 270 – 275 and figure 3: Have the authors statistically substantiated their claim that the fluorescence of BPEAnit increases with sampling time? To me, the error bars are overlapping, and the replicate size (n =3) is too small to make this inference statistically significant. Please validate if this statement is statistically significant.

Why are these changes in lines 275-280 for 5 μg/mL of Printex 90 very different from those in Appendix 2?

Lines 280 – 287: Have the authors validated whether the normalization at no particles holds true in the presence of particles? The authors mention that BPEAnit-Me is non-reactive in line 25, so it is likely that their quantum yields may not be the same at no particle vs. particle conditions.

Lines 313 – 318: I am not sure if the authors can make the statement that the ab/adsorption are similar between the two probes without validation.

Line 335: I am not sure if a 5% decrease makes the fluorescence decrease evident. Are these differences statistically significant?

Lines 358 – 362: I think the correction factor should not be 0.63, but rather [0.63/0.83 = 0.76] as only 84% of the BPEAnit fluorescence remains at 5 mins.

Conclusions: Adsorption/Absorption were not measured to conclude those as reasons for the reduction in fluorescence.

Reply
Citation: https://doi.org/10.5194/egusphere-2026-3807-RC1
RC2:
'Comment on egusphere-2026-3807', Anonymous Referee #2, 09 Aug 2026 reply
General Comments
This manuscript investigates an important source of artefact in fluorescence-based measurements of particle-bound reactive oxygen species (ROS), namely the suppression of BPEAnit fluorescence in the presence of carbonaceous particles. The use of BPEAnit-Me as a reference compound to characterize particle-induced signal loss is potentially useful. The experiments clearly demonstrate a concentration- and time-dependent change in fluorescence following the addition of Printex 90. However, several key aspects require further clarification before the proposed correction approach can be considered quantitatively robust, particularly the assumptions underlying Eq. (1), the mechanistic interpretation of the fluorescence loss, and the applicability of the batch experiments to online PINQ measurements. These issues should be addressed before the proposed correction can be reliably applied to quantitative ROS measurements.
Major Comments
The key assumption underlying Eq. (1) requires stronger justification. Eq. (1) assumes that BPEAnit and BPEAnit-Me experience equivalent particle-induced fluorescence suppression, and further assumes that the fluorescent product formed after reaction of BPEAnit with ROS is affected to the same extent as BPEAnit-Me. This assumption is central to the correction procedure but has not been directly validated. This is particularly relevant because BPEAnit and BPEAnit-Me were used at substantially different concentrations, 4 µM and 130 nM, respectively. The authors should provide further justification or experimental validation of this assumption. If such validation is not available, the applicability and uncertainty of Eq. (1) should be more explicitly constrained.

The proposed adsorption/inactivation mechanism remains inferential. The manuscript attributes the observed fluorescence reduction primarily to adsorption of BPEAnit/BPEAnit-Me onto Printex 90 surfaces, leading to fluorophore inactivation. However, the present experiments demonstrate particle-induced fluorescence loss rather than adsorption itself. Other mechanisms, including static or dynamic fluorescence quenching or other surface-mediated interactions, cannot be excluded. The authors should either provide more direct evidence for adsorption or consistently describe it as a proposed mechanism rather than a demonstrated one, particularly in the Conclusions.

The attribution of the BPEAnit–BPEAnit-Me difference to ROS formation requires further support. The consistently higher fluorescence response of BPEAnit relative to BPEAnit-Me is interpreted as ROS-dependent probe activation. This interpretation is plausible, but it relies on the assumption that non-ROS particle interactions are equivalent for the two compounds. Differential adsorption or quenching of BPEAnit and BPEAnit-Me could also contribute to the observed difference. The authors should provide stronger justification for attributing this difference primarily to ROS formation and discuss this alternative interpretation. Independent validation, if available, would further strengthen the conclusion.

The quantitative relationship between Eq. (1), Fig. 4, and Fig. 5 is unclear. Eq. (1) is defined using normalized relative fluorescence and therefore yields a dimensionless quantity, whereas Fig. 5 presents "Calculated ROS" in fluorescence units (a.u.). The complete calculation procedure from raw fluorescence counts to normalized fluorescence and subsequently to the values plotted in Fig. 5 should be explicitly described. If the corrected response was converted back to fluorescence counts or another quantity, the corresponding conversion and units should be clearly defined.

The representativeness of the experimental time scale for online PINQ measurements should be clarified. The earliest measurements were conducted within 5 min after particle addition, whereas the manuscript states that PINQ analysis occurs within approximately 1 min after particle collection and mixing. Because the particle–probe interaction evolves rapidly at early times, measurements obtained within 5 min cannot necessarily be assumed representative of the online PINQ time scale. The authors should either demonstrate the relevance of this time scale more clearly or qualify the applicability of the results to real-time PINQ operation.

The transferability of the batch experiments to actual PINQ operation requires a more explicit discussion. The present experiments were conducted using Printex 90 directly suspended in DMSO, whereas PINQ continuously collects aerosol particles into a BPEAnit-containing liquid. Differences in solvent composition, particle dispersion state, collection process, and contact time may influence the magnitude of the particle–probe interaction. The limitations associated with transferring the present batch-suspension results to actual PINQ operation should therefore be discussed more explicitly, and the proposed correction should be clearly framed as requiring further evaluation under online operating conditions. In addition, the manuscript should clarify the choice of 4 µM BPEAnit and whether this represents an updated PINQ operating protocol.

Basic physicochemical information on Printex 90 should be provided. Because the proposed mechanism is explicitly related to particle surface interactions, particle mass concentration alone provides limited mechanistic information. The manuscript currently reports the preparation and sonication procedure but provides little information on primary particle size, specific surface area, or other relevant properties of Printex 90. At minimum, basic physicochemical properties from the manufacturer or literature should be reported. The atmospheric relevance of the tested concentrations should also be discussed more carefully. In particular, the description of 1 µg mL^–¹ as a "low concentration" should be qualified, since this corresponds to an atmospheric particle concentration of approximately 63 µg m^–³ under the stated assumptions.

The uncertainty and validity range of Eq. (1) should be addressed. At the highest Printex 90 concentration, the BPEAnit-Me response approaches zero, making the correction in Eq. (1) mathematically unstable and strongly amplifying the calculated ROS response. The authors acknowledge this limitation, but the corresponding values are still presented and discussed quantitatively in Fig. 5. The uncertainty associated with Eq. (1) should be evaluated, and the range over which the correction can be meaningfully applied should be clearly defined or discussed.

Minor Comments
Units, numerical formatting, and figure labels should be standardized throughout the manuscript. For example, use µg mL^–¹ rather than µg/ml and 0.5 h rather than 0,5 h. Typographical errors in figure labels should also be corrected.

Printex 90 is a commercial carbon black used here as a model for soot-like particles. The observed interference should therefore not be generalized to atmospheric black carbon or other insoluble particles without appropriate qualification.

Reply
Citation: https://doi.org/10.5194/egusphere-2026-3807-RC2

Jonas Enarsson, Vilhelm Malmborg, Branka Miljevic, Axel Eriksson, Hao Wang, Zoran Ristovski, and Aneta Wierzbicka

Viewed

Total article views: 184 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	BibTeX	EndNote
138	33	13	184	16	11

HTML: 138
PDF: 33
XML: 13
Total: 184
BibTeX: 16
EndNote: 11

Views and downloads (calculated since 14 Jul 2026)

Month	HTML	PDF	XML	Total
Jul 2026	106	30	10	146
Aug 2026	32	3	3	38

Cumulative views and downloads (calculated since 14 Jul 2026)

Month	HTML	PDF	XML	Total
Jul 2026	106	30	10	146
Aug 2026	32	3	3	38

Viewed (geographical distribution)

Total article views: 142 (including HTML, PDF, and XML) Thereof 142 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 16 Aug 2026

Short summary

Reactive oxygen species and oxidative potential are strongly linked to aerosol-induced adverse health effects and have emerged as widely used metrics for assessing particle toxicity. We found that soot-like particles can interfere with methods used to measure reactive oxygen species, causing misleading results. We have investigated this artefact and propose an explanation for the interference and a correction method that improves the accuracy of reactive oxygen species measurements.


Total:	0
HTML:	0
PDF:	0
XML:	0