Formation of hydroxyl radical from San Joaquin Valley particles extracted in a cell-free surrogate lung fluid

Previous studies have suggested that the adverse health effects from ambient particulate matter (PM) are linked to the formation of reactive oxygen species (ROS) by PM in cardiopulmonary tissues. While hydroxyl radical (•OH) is the most reactive of the ROS species, there are few quantitative studies of •OH generation from PM. Here we report on •OH formation from PM collected at an urban (Fresno) and rural (Westside) site in the San Joaquin Valley (SJV) of California. We quantified •OH in PM extracts using a cell-free, phosphate-buffered saline (PBS) solution with or without 50μM ascorbate (Asc). The results show that generally the urban Fresno PM generates much more •OH than the rural Westside PM. The presence of Asc at a physiologically relevant concentration in the extraction solution greatly enhances •OH formation from all the samples. Fine PM (PM2.5) generally makes more •OH than the corresponding coarse PM (PMcf, i.e. with diameters of 2.5 to 10 μm) normalized by air volume collected, while the coarse PM typically generates more •OH normalized by PM mass. •OH production by SJV PM is reduced on average by (97±6)% when the transition metal chelator desferoxamine (DSF) is added to the extraction solution, indicating a dominant role of transition metals. By measuring calibration curves of •OH generation from copper and iron, and quantifying copper and iron concentrations in our particle extracts, we find that PBS-soluble copper is primarily responsible for •OH production by the SJV PM, while iron often makes a significant contribution. Extrapolating our results to expected burdens of PM-derived •OH in human lung lining fluid suggests that typical daily PM exposures in the San Joaquin Valley are unlikely to result in a high amount of pulmonary •OH, although high PM events could produce much higher levels of •OH, which might lead to cytotoxicity.


Introduction
Epidemiological studies have shown strong correlations between the exposure to ambient particulate matter (PM) and adverse human health outcomes such as pulmonary and cardiovascular diseases and premature deaths (Dockery et al., 1993;Pope et al., 1995Pope et al., , 2004;;Pekkanen et al., 2002;Pope and Dockery, 2006).One suggested mechanism by which PM induces toxic effects is PM-mediated oxidative stress and cell damage through the generation of reactive oxygen species (ROS) such as superoxide ( • O − 2 ), hydrogen peroxide (HOOH), and hydroxyl radical ( • OH) (Li et al., 2008;Valavanidis et al., 2008;Gonzalez-Flecha, 2004;Donaldson et al., 2003).These ROS are thought to be formed in the cell during oxidative phosphorylation, where sequential electron addition to dissolved O 2 results in the formation of • O − 2 , HOOH, and • OH, respectively (Li et al., 2003).As shown in Fig. 1, ROS can also be formed via reduction of oxygen species by the reduced forms of transition metals, which are recycled via reductants such as ascorbate.
• OH is the most reactive ROS and it can react with most organic molecules at diffusion-controlled rate constants (Held et al., 1996;Forman et al., 2010).Unlike • O − 2 and HOOH, which can be detoxified by superoxide dismutase and catalase, respectively, • OH cannot be eliminated enzymatically.• OH can cause a variety of oxidative damage to cellular macromolecules including carbohydrates, lipids, proteins, and nucleic acids, which can result in cell death and disease (Valavanidis et al., 2008;Kell, 2010).Thus in vitro and in vivo • OH formation might be useful as an indicator of the toxic potential of inhaled PM (Cohn et al., 2008).The generation of • OH from particles extracted in cell-free solutions also gives information about the oxidative potential of PM.Several groups have made these types of • OH measurements, both for ambient particles as well as specific types of particles (Shi et al., 2003;Baulig et al., 2004;Kunzli et al., 2006;Jung et al., 2006;Alaghmand and Blough, 2007; Transition-metal-catalyzed production of reactive oxygen species.TM represents transition metals, Asc is ascorbate, and (red) and (ox) represent reduced and oxidized forms.In this scheme, reduced forms of transition metals donate electrons to convert dissolved molecular oxygen to, sequentially, superoxide, hydrogen peroxide, and hydroxyl radical, with the reductant ascorbate (Asc) serving as the ultimate electron donor.DiStefano et al., 2009;Vidrio et al., 2009).These studies indicate the amounts of • OH that can be chemically generated by different particles, and generally find that transition metals play the dominant role in • OH generation.
To help characterize the chemical generation of ROS from ambient particles in a cell-free solution, we recently measured the formation of HOOH by fine (PM 2.5 ) and coarse (PM cf ) particles collected at an urban and rural site in the San Joaquin Valley (SJV) of California (Shen et al., 2011).In this current manuscript we report measurements of • OH on these same SJV particles, in order to: (1) quantify the amounts of • OH produced from the particles in the same surrogate lung fluid; (2) compare • OH generation from PM samples collected at the urban and rural site, during different seasons (summer vs. winter), and in different sizes (fine vs. coarse); (3) explore the role of added ascorbate on • OH generation; and (4) examine the role of transition metals in general, and Cu and Fe in particular, in • OH formation.

Surrogate lung fluid (SLF)
All experiments were performed in a cell-free SLF solution that contained 114 mM NaCl, 10.0 mM total phosphate (7.8 mM Na 2 HPO 4 and 2.2 mM KH 2 PO 4 ) to buffer the solution at pH 7.2 to 7.4, and 10 mM NaBA as a chemical probe to detect • OH.Prior to particle extraction, transition metals were removed from the SLF using a column filled with chelex-100.The SLF was then refrigerated and generally used within one month of preparation.In most cases, right before the start of sample extraction, freshly made Asc was added to the SLF to get a final concentration of 50 µM, similar to endogenous concentrations of the reductant (Cross et al., 1994;van der Vliet et al., 1999).

PM collection and extraction
Fine (PM 2.5 ) and coarse (PM cf ) particle samples were collected at an urban (Fresno) and rural (Westside) site in California's SJV during summer and winter between 2006 and 2009 by other researchers from UC Davis.A total of twelve samples were collected, with one PM 2.5 sample and one PM cf sample taken during each sampling period.After collection samples were kept at −20 • C until analysis.Although our storage times were quite long (approximately 1 to 4 yr; Table S1), we do not believe that this significantly reduced the ability of the particles to produce ROS since metals were the dominant redox-active species on the particles (Sect.3.3) and we added fresh Asc to the SLF on each experiment day.PM masses were determined using a Mettler Toledo XP26 microbalance with 1 µg precision.Generally, for a given sample the PM 2.5 mass concentration was greater than the corresponding PM cf .Additional information about sample PM masses, and PM mass extraction efficiencies (69-97 %), are in our HOOH study (Shen et al., 2011).
For • OH measurements, a punch of filter (PM 2.5 sample) or a piece of foil (PM cf sample) was placed in a 7-ml perfluoroalkoxy (PFA) Teflon vial that was pre-washed with nitric acid to remove transition metals.After adding 6.0 ml of SLF and, in most cases, Asc with the final concentration of 50 µM, the vials were completely wrapped with aluminum foil, placed on a wrist-action shake table (VWR OS-500, set at "5"), and shaken in the dark at room temperature for up to 24 h.For a given sample, each PM extraction was performed on a different punch (or piece) of sample cut from the same filter (or foil) and thus the number of replicates (n) in each figure represents multiple independent measurements of the same sample.For every experiment day we also "extracted" three different types of controls the same way as we treated the PM samples: (1) an SLF solution blank, (2) corresponding field blanks, i.e. filter or foil substrate that had been placed in the sampler in the field without collecting sample, and (3) 250 nM of CuSO 4 in SLF with Asc as a positive control.To examine the role of transition metals in • OH formation, in some experiments the chelator DSF was added to the SLF (for a final concentration of 1.0 mM) before adding Asc.

• OH determinations
• OH in our experiments was determined using 10 mM benzoate as a chemical probe (Anastasio and McGregor, 2001;Jung et al., 2006): as • OH is generated in the extract solution, it reacts with benzoate to form p-HBA, a stable product that is quantified by HPLC.The HPLC consisted of a Shimadzu SIL-10AF autosampler with CMB-20A controller, a Shimadzu LC-10ATVP pump, a Keystone Scientific C-18 Beta Basic reverse-phase column (3 × 250 mm, 5 µm bead) with an attached guard column, and a Shimadzu SPD-10 AV UV-Visible detector (λ = 256 nm).The eluent was 30 % CH 3 CN and 70 % H 2 O adjusted to pH 2 with HClO 4 , continuously degassed with a slow stream of helium (99.997 %), and run at a flow rate of 0.60 ml min −1 .A 500 µl aliquot of PM extraction solution was analyzed for p-HBA after 0, 1, 2, and 24 h of shaking.The extract was quickly filtered using a 0.22 µm syringe filter (Milex Millipore) and transferred into an autosampler vial (Fisher Scientific), where it was mixed with 100 µM DSF and 50 µM HSO − 3 to quench • OH generation.After 10 min in the dark, the extract was acidified to pH 2 by adding 5 µl of 1.0 M H 2 SO 4 .Samples were stored at 4 to 8 • C until HPLC analysis.
The concentration of p-HBA in each sample was quantified using a calibration curve produced from p-HBA standards in SLF run on the same day of experiment.The • OH concentration in each sample was calculated using (Jung et al., 2006): where [p-HBA] is the measured concentration of p-HBA, Y p-HBA is the molar yield of p-HBA from the reaction of • OH with BA in SLF (0.215 ± 0.018) (Jung et al., 2006), and f BA is the fraction of • OH that reacts with BA in a specific SLF.Based on published rate constants for • OH (Walling et al., 1974;Buxton et al., 1988;Zepp et al., 1992) we calculated values of f BA to be 0.9999 in the absence of Asc or DSF, 0.9972 with Asc, and 0.8175 with both Asc and DSF.
Based on the NIST compilation (Buxton et al., 1988), the average room temperature rate constants for • OH with SA and Asc are 1.6 × 10 10 M −1 s −1 and 6.4 × 10 9 M −1 s −1 , respectively, while the phosphate buffer is a negligible • OH sink.Thus, by following the procedure of Charrier and Anastasio (Charrier and Anastasio, 2011), we calculate that f SA = 0.71 in the experiments of DiStefano et al. (2009).

ICP-MS analysis of transition metals
400 µl of filtered 24-h PM extract was diluted with 3.6 ml of 3 % HNO 3 into an acid-rinsed 15-ml Corning® polypropylene centrifuge tube, and refrigerated until analysis.Samples were analyzed for Cu, Fe, V, and Mn using an Agilent 7500CE ICP-MS.A series of metal standards was prepared for quality control purposes (Shen et al., 2011).Metal concentrations for each PM sample extract were corrected for the metal amount in the corresponding field blank.

Data analysis
Two parameters were determined for each PM extract: (1) the initial rate of • OH formation, calculated using the 0 and 1 h time points, and (2) the maximum where all rates are in µM h −1 .The average positive control rate was 0.381 ± 0.061 µM h −1 .Each extract volume was 0.0060 l, while the air volumes sampled for each PM 2.5 and PM cf sample piece were 2.346 and 21.444 m 3 , respectively.Analogous equations were used to calculate the airvolume-normalized maximum • OH formation (average positive control maximum = 2.83 ± 0.24 µM) and the PM-massnormalized • OH rates and maxima.We normalized sample results to the positive control because we found that • OH generation from the positive control was covariant with sample and blank values on the same day of experiment; the positive control varied within a range of approximately −30 % to +27 % relative to its average value.
Data were analyzed using SPSS 12.0 (SPSS) and SigmaPlot 11.0 (Systat Software) and presented as means ±SD or medians and upper and lower quartiles and extremes using box and whisker plots.Comparisons of • OH generation among different PM samples were performed using oneway ANOVA followed by Bonferroni post hoc test.Differences in means were considered significant when P < 0.05.

Results and discussion
Figure 2 shows some examples of the time course of • OH generation from SJV PM, and the Cu (II) positive control, during our 24-h extraction.As shown in this figure, the solution blanks and field blanks generated very low levels of • OH.In contrast, • OH production from the positive control reached a concentration of approximately 1.7 µM at 4 h and continued to rise, though more slowly, at longer times.As also illustrated in the figure, and described in more detail below, Fresno PM was more active in forming • OH than Westside PM.As described in Sect.2.6, we used the 0 and 1 h time points to estimate the initial rate of • OH formation and reported the 24-h time point value as the "maximum" amount of • OH formed (i.e. the total • OH formed over 24 h), although more • OH is likely to be formed after this point in at least some of the samples.
We normalized the initial rate of • OH formation, and maximum • OH concentration, in each PM extract in two different ways: (1) to the air volume sampled during PM collection (e.g.nmol-• OH h −1 m −3 -air) and (2) to the extracted PM mass (e.g.nmol-• OH h −1 mg −1 -PM).The two ways of normalization are relevant to PM inhalation studies and PM instillation studies, respectively (Shen et al., 2011).

Generation of • OH in PM extracts with added ascorbate
We first quantified • OH generation from SJV PM extracted in SLF with 50 µM of added ascorbate, an important antioxidant in human lung lining fluid (Cross et al., 1994;van der Vliet et al., 1999).As shown in Fig. 1, ascorbate can also act as a pro-oxidant by recycling transition metals from oxidized to reduced forms, thus promoting ROS generation (Stadtman, 1991;Satoh and Sakagami, 1997;McGregor and Biesalski, 2006;Vidrio et al., 2008;Shen et al., 2011).
In the presence of ascorbate, the Fresno (urban) particles are generally much more reactive than the Westside (rural) particles in generating • OH, on both an air-volume and PMmass normalized basis.The initial rates of • OH formation are shown in Fig. 3a and b: on average, the Fresno fine and coarse particles are 5.5 and 11.4 times more reactive, respectively, than their Westside counterparts for air-volume normalized rates (and 4.1 and 16.1 times more effective, respectively, for PM-mass normalized rates).Based on the airvolume-normalization, the fine particles are generally more important sources of • OH than the coarse particles in the ambient aerosol (Fig. 3a); as we described earlier for HOOH, this is because the PM 2.5 mass concentration is much higher than the PM cf mass concentration during each sampling period (Shen et al., 2011).On the other hand, on a PM-massnormalized basis, the coarse particles are typically somewhat more efficient than the fine particles at generating • OH (Fig. 3b).We see the same relative importance of fine particles (dominating air-volume-normalized • OH generation) and coarse particles (more efficient on a mass-normalized basis) for the maximum • OH measured (Fig. S1).Although our sample size is small, we find no apparent seasonal difference in either the initial rate of • OH generation (Fig. 3 the maximum • OH formation (Fig. S1).The results of • OH generation in SLF with added Asc are consistent with our previous findings of HOOH formation in the same SLF: (1) the urban samples generate more HOOH than the rural samples, (2) fine PM generally makes more HOOH than coarse PM per volume of air, (3) coarse PM typically produces more HOOH than fine PM per mass unit of PM, and (4) there is no seasonal difference in HOOH generation (Shen et al., 2011).
Figure 4a and b compare the maximum • OH generation from our Fresno and Westside PM with • OH formation from Davis PM 2.5 (Vidrio et al., 2009) .While the surrogate lung fluid we used here contained 50 µM Asc, the Davis particles were extracted in a fluid containing 200 µM Asc and 300 µM   (Vidrio et al., 2009).The southern California PM were extracted for 45 min at 37 • C in a pH 6.4 solution containing 500 µM ascorbate and 500 µM salicylate as the • OH probe (DiStefano et al., 2009).Each box and whisker plot shows the median, upper and lower quartiles, and upper and lower extremes.
H. Shen and C. Anastasio: Hydroxyl radical from San Joaquin Valley particles citrate (Cit), conditions that reduce the effectiveness of Cu at generating • OH but increase the effectiveness of Fe (Charrier and Anastasio, 2011).Despite the differences in SLF composition, the maximum amount of • OH generated by our Fresno PM 2.5 is comparable to the Davis PM 2.5 results (Fig. 4), although the Davis samples show a clear seasonal difference in • OH generation, with the spring/summer PM 2.5 much more efficient in producing • OH than the winter PM 2.5 (Vidrio et al., 2009).We can also compare our initial rates of • OH generation in Fresno and Westside PM extracts with values determined from southern California quasi-ultrafine PM (PM 0.18 ), which were extracted in a pH 6.4 solution containing 500 µM Asc (DiStefano et al., 2009).The PM-mass-normalized initial rates of • OH generation for our Fresno (and Westside) PM are much lower than the • OH rates for the southern California PM 0.18 (Fig. 4c), with median values of 18, 21, 1225, and 1218 nmol h −1 mg −1 for the Fresno PM 2.5 , Fresno PM cf , Riverside PM 0.18 , and Claremont PM 0.18 , respectively.Thus, the southern California PM 0.18 are approximately 60 times more reactive than the Fresno fine and coarse PM in generating • OH.However, the extraction conditions in these two studies were quite different: the southern California PM 0.18 were extracted at a much higher temperature (37 • C, compared to room temperature in our experiments) and with a 10fold higher concentration of ascorbate (500 µM, compared with 50 µM in our experiments).These methodological differences can probably account for much of the difference in • OH rates seen between our Fresno PM and the southern California PM 0.18 : (1) based on results in a more complicated surrogate lung fluid (200 µM Asc, 300 µM Cit, 100 µM glutathione, and 100 µM uric acid), we find that the rate of • OH production from dissolved Fe is approximately 5 times faster at 37 • C compared to room temperature (J.Charrier, personal communication, 2011), (2) the rate of • OH generation in a 500 µM Asc solution is likely close to 10-times faster than in a 50 µM Asc solution, and (3) if these effects are multiplicative, the southern California PM 0.18 rates of • OH production should be approximately 50 times faster than the Fresno PM solely because of extraction differences, which is close to the observed factor of 60 (Fig. 4c).Thus, while the southern California PM 0.18 particles are likely somewhat more reactive than the Fresno PM 2.5 and PM cf , the difference is probably greatly magnified in Fig. 4c because of the variation in extraction conditions.

Generation of • OH in PM extracts without added ascorbate
• OH production in the SJV PM extracts above were all in SLF containing 50 µM ascorbate, which mimics the lung lining fluid concentration (Cross et al., 1994;van der Vliet et al., 1999).To examine the importance of ascorbate on • OH generation, we also measured • OH formation in PM extracts without added Asc.As shown in Fig. 5, there was essen-tially no • OH generation in the fine and coarse PM extracts within the first 1 h in the absence of Asc, with one exception -Fresno Winter 2009 PM 2.5 .In this sample the initial rate of • OH formation in SLF without Asc (Fig. 5) was 11 times lower than the rate measured in SLF with added Asc (Fig. 3).However, given that this sample had the highest rate of • OH formation in the absence of ascorbate, it might be underestimating the typical impact of Asc on • OH generation.Indeed, as shown in Fig. 6, in this sample the presence of ascorbate had the weakest amplifying effect on the rate of • OH generation (a factor of 11), compared to factors ranging from 21 to 4500 in the other samples (with an overall median value of approximately 50), independent of whether • OH results are normalized to air volume or PM mass.These results are consistent with our previous results on HOOH generation by SJV PM, where the presence of ascorbate also greatly amplified HOOH formation, with a median enhancement of a factor of 19 (Shen et al., 2011).
As with the initial rate of • OH formation, the maximum amounts of • OH formed in SLF without added Asc (Fig. S2) were also much lower than those in SLF with added Asc (Fig. S1).The presence of ascorbate amplified the maximum • OH formation from the SJV PM by factors of 6 to 258 (Fig. S3), with a median value of approximately 60.In the absence of Asc, the Fresno winter 2007 coarse PM generated the highest • OH maximum, followed by the Fresno summer 2008 and winter 2009 coarse PM (Fig. S2).The relatively high production of • OH by the Fresno winter 2007 coarse PM is especially pronounced on a PM mass-normalized basis (Fig. S2b), and could be due to the role of redox-active organic compounds such as quinones (Dellinger et al., 2001;Rodriguez et al., 2005;Valavanidis et al., 2008).The generation of • OH in the absence of ascorbate by several samples suggests these particles contain unidentified reductants that can reduce oxidized forms of metals and/or organics to form • OH (Fig. S2).However, while • OH generation in the absence of ascorbate in these few samples is interesting, as we describe below, • OH generation in our PM samples is dominated by soluble transition metals utilizing ascorbate as the reductant.
We can use our results with and without ascorbate to discern the relative importance of the different acellular mechanisms by which particles can produce • OH and HOOH during aqueous extraction.There are at least three of these mechanisms: (1) dissolution of particle-bound ROS such as peroxides (HOOH, ROOH, ROOR') into solution, (2) reactions of particle-bound ROS precursors, e.g.reduced forms of redox-active species such as Fe(II), to make ROS in solution, and (3) redox-cycling reactions where particle components (e.g.Cu) interact with reductants in the extraction solution (e.g.ascorbate) to form ROS. Comparing the amount of ROS formed in the presence of Asc (where all three mechanisms contribute) to the amount formed in the absence of Asc (where only mechanisms (1) and ( 2) contribute) indicates the relative importance of these mechanisms.For our Fresno samples, the median ratio of the • OH formation rate in the presence of Asc to the • OH formation rate in the absence of Asc is 47 (Fig. 6).The same picture holds for HOOH, where the analogous median ratio is 42 (Shen et al., 2011).These ratios strongly suggest that redox reactions involving endogenous reductants (mechanism (3)) are the dominant chemical sources of ROS from particles deposited in the lungs.

SLF-soluble transition metals, especially Cu, play a dominant role in • OH generation from SJV PM
As an initial step to explore the mechanisms of • OH generation from particles extracted in the presence of ascorbate, we performed parallel extractions in the presence of DSF, a strong metal chelator, in order to remove • OH generation by transition metals.As shown in Fig. 7, DSF is exceptionally effective at reducing • OH generation in extracts of PM from both sites: on average, adding DSF reduces the initial rate of • OH formation by 94 (±8) % for the fine PM and by 100 (±0.5) % for the coarse PM.Similarly, DSF decreases the maximum • OH formation by 98 (±2) % and 98 % (±1) % for the fine and coarse PM, respectively (Fig. S4).These results indicate that essentially all • OH generation in the PM 2.5 and PM cf extracts involved transition metals.As we reported previously, transition metals also dominated HOOH generation from these particles, although to a lesser extent compared to • OH: DSF reduced the initial rate of HOOH formation by 83 (±16) % and 73 (±13) %, and the maximum HOOH formation by 78 (±12) % and 63 (±14) %, for the fine and coarse particles, respectively (Shen et al., 2011).
As our second step in understanding the mechanisms for • OH formation in the San Joaquin Valley particles, we specifically examined the contributions of SLF-soluble Cu and Fe.We chose to focus on these metals since our Cu positive control is very effective in generating • OH (e.g.Fig. 2) and previous studies have shown that both Cu and Fe can be effective sources of ROS (Vidrio et al., 2008(Vidrio et al., , 2009;;DiStefano et al., 2009;Rushton et al., 2010;Wang et al., 2010;Nawrot et al., 2009).A regression analysis shows that the air-volume-normalized initial rate of • OH formation by Fresno fine and coarse PM is strongly linearly correlated with SLF-soluble Cu (R 2 = 0.98) (Fig. 8), suggesting that Cu plays a major role in • OH formation in the Fresno particles.We also find a strong, but non-linear, relationship between the maximum amount of  PM samples (Fig. S5).For the rural Westside particles there is no correlation between the initial rate of • OH formation (or maximum amount of • OH formed) and SLF-soluble Cu (Fig. 8 and Fig. S5), but this is a very small sample set.
In contrast to the strong correlations with Cu seen for the Fresno particles, SLF-soluble Fe is not correlated with the initial rate of • OH formation (R 2 = 0.05) or the maximum • OH concentration (R 2 = 0.20).Similarly, we find no correlation between • OH formation by the Fresno particles and either SLF-soluble V (R 2 = 0.08 and 0.24 for initial rate and maximum • OH) or Mn (R 2 = 0.13 and 0.19 for initial rate and maximum • OH).
Our results are consistent with previous papers that have examined relationships between • OH generation and soluble transition metals in ambient particle extracts.For example, Cho and co-workers (Distefano et al., 2009) also found that soluble Cu is strongly correlated with the rate of • OH generation for PM 0.18 from southern California, while there were no correlations between • OH generation and soluble Fe, V, or Mn.In contrast, for PM 2.5 from Davis, CA, Vidrio et al. (2009) found no correlation between soluble Fe or Cu and the amount of • OH formed by PM extracted for 24 h in SLF containing 200 µM ascorbate and 300 µM citrate.Despite the lack of correlation, a more mechanistic examination -involving quantifying • OH generation from dissolved Fe and Cu in the SLF extracts of PM -revealed that soluble Fe could account for the bulk of • OH generation from Davis PM (Vidrio et al., 2009).
In order to quantitatively understand the contributions of Cu and Fe towards • OH generation in our SJV particles, we also applied the technique of Vidrio et al. (2009) to our samples.This determination involves four steps: (1) making "calibration curves" that quantify the initial rate (and maximum concentration) of • OH generated from known concentrations of dissolved Cu and Fe in SLF containing 50 µM  As shown in Fig. 9, Cu dominates • OH formation in the Fresno PM samples.On average, Cu accounts for 89 ± 18 % of the initial rate of • OH generation in the Fresno PM 2.5 extracts and 89 ± 23 % in the Fresno PM cf extracts.Similarly, Cu can account for 156 ± 23 % and 107 ± 27 % of the maximum • OH generated in the Fresno fine and coarse PM extracts, respectively (Fig. S6).While Fe also contributed to • OH generation, it played a smaller role, accounting for less than 30 % of the • OH rate or maximum in the Fresno fine and coarse PM extracts (Figs. 9 and S6).For the Westside PM samples the picture is less clear, in part because the • OH production was generally much smaller and thus less certain, but Cu was also the dominant source of • OH in these samples (Figs. 9 and S6).
Together, SLF-soluble Cu and Fe can account for all of the • OH formed in our PM extracts (Figs. 9 and S6).For the Fresno samples the average sum of ratios for the Fresno fine and coarse PM are 1.07 ± 0.41 and 0.96 ± 0.49 for the rate of • OH formation, and 1.84 ± 0.67 and 1.19 ± 0.43 for the maximum • OH, respectively.For the Westside samples, Fe and Cu can also account for measured • OH generation, although the results are quite noisy: the average sum of ratios for the Westside fine and coarse PM are 1.3 ± 1.1 and 2.3 ± 2.6 for the rate of • OH formation, and 1.4 ± 0.7 and 2.1 ± 5.2 for the maximum • OH, respectively.As we described previously, Fresno PM cf generally contains higher levels of PM-mass-normalized soluble copper (i.e.ng-Cu µg −1 -PM) than the Fresno PM 2.5 (Shen et al., 2011), which helps to explain why the Fresno coarse PM is generally more reactive in generating • OH than the corresponding fine PM on a PM-mass-normalized basis (Fig. 3b, Fig. S1b).
Our current finding that soluble Cu can account for essentially all of • OH generation from the Fresno PM 2.5 is an interesting contrast to our previous finding that dissolved Fe dominates • OH formation from PM 2.5 collected in Davis, CA (Vidrio et al., 2009).This difference is likely due to the fact that the SLF in Vidrio et al. study contained both ascorbate (200 µM) and citrate (300 µM), while the SLF in this work contained only ascorbate (50 µM).We recently reported that, compared to an SLF containing only ascorbate, adding citrate enhances the ability of Fe to generate • OH and inhibits the ability of Cu to make • OH (Charrier and Anastasio, 2011).Thus the composition of SLF used to extract particles can significantly, and differentially, affect the roles of different transition metals in • OH generation.
More broadly, our finding that transition metals dominate • OH formation by the SJV PM adds support to the link between particulate transition metals and PM-induced adverse health effects that has been found by previous studies (Costa and Dreher, 1997;Donaldson et al., 2003;Valavanidis et al., 2008;Lippmann and Chen, 2009;Gerlofs-Nijland et al., 2009).Our finding that Cu is responsible for the majority of • OH generation is also consistent with the particulate-Cu mediated toxic effects found in numerous in vitro and in vivo studies, including ROS generation and oxidative stress, protein and DNA oxidative damage, inflammation and tissue injury (Shi et al., 2003;Gasser et al., 2009;Wallenborn et al., 2009;Rushton et al., 2010).Considering that our Fresno (urban) site is close to a major highway and multiple surface streets while the Westside (rural) site has very little nearby traffic, vehicular brake wear emissions, which contain relatively high copper concentrations (Gasser et al., 2009;Bukowiecki et al., 2009), are likely an important source of Cu in our PM samples.

Implications and uncertainties
To examine if the amounts of • OH produced in aqueous extracts of SJV PM might be significant for human health, we first estimate the expected daily PM-mediated • OH load in the lung lining fluid based on our measured maximum • OH levels, using the procedure of Vidrio et al. (2009): Using the average of the maximum • OH production over the 24 h extraction (4.0 nmol m −3 for PM 2.5 and 0.7 nmol m −3 for PM cf ; Fig. S1), an inhaled air volume of 20 m 3 per day, and assuming 30 % of inhaled PM 2.5 and 70 % of inhaled PM cf deposit in lungs (Sarangapani and Wexler, 2000), we calculate that the average • OH lung burden from aerosol inhalation is 34 nmol • OH d −1 in Fresno, with 71 % of • OH formation from PM 2.5 .The same calculation for the Westside particles produces an average • OH lung burden of 16 nmol • OH d −1 , with 73 % of • OH formation from PM 2.5 .For individual samples, particle-mediated • OH burdens in lung lining fluid range from 26 to 50 and 11 to 19 nmol d −1 for Fresno and Westside PM, respectively.
Since lung lining fluid antioxidants provide an important defense network to protect against • OH-mediated cellular damage, we compare the estimated • OH burdens to the total amount of antioxidants in lung lining fluid, which is approximately 15 000 nmol (Vidrio et al., 2008).Since the total antioxidant level is much higher than the levels of PM-mediated • OH from the Fresno and Westside particles, the amounts of • OH generated might not be significant for human health.Even peak PM 2.5 events in Fresno likely produce relatively low amounts of • OH.For example, the maximum 24 h average PM 2.5 concentration was ∼100 µg m −3 in both 2006and 2007(California Air Resources Board, 2010), which is 3 times higher than our average Fresno PM 2.5 concentration (33 µg m −3 ) (Shen et al., 2011).Assuming a linear response between PM 2.5 mass and • OH generation, this peak PM 2.5 level corresponds to a daily • OH burden in the lung lining fluid of ∼70 nmol, which is still quite small compared to the antioxidant pool.However, relatively small amounts of • OH can lead to much greater levels of ROS, and oxidative damage, in vivo by initiating lipid peroxidation (Leibovitz and Siegel, 1980).In addition, additional ROS -including • OH, will be formed indirectly by PM via activation of oxidant-dependent signaling pathways in lung epithelial cells (Gonzalez-Flecha, 2004).Furthermore, other ambient air pollutants, such as ozone, can act additively or synergistically with PM to increase aqueous-phase • OH production at physiological pH (Valavanidis et al., 2009).
While our results suggest that the chemical generation of • OH by inhaled ambient particles might lead to toxic effects under some circumstances, there are some large uncertainties.First, our results are for cell-free solutions that do not include biological responses that either increase (e.g.PMmediated generation of ROS by activated macrophages and epithelial cells) or decrease (e.g.decomposition of • O − 2 and HOOH by superoxide dismutase and catalase, respectively) cellular oxidative stress.Secondly, while we only included one antioxidant (ascorbate) in our surrogate lung fluid for PM extraction, recent studies suggest that other antioxidants (e.g.glutathione) and endogenous substances (e.g.citrate) in lung fluid can effectively inhibit the generation of • OH by Cu, while increasing • OH from Fe (Vidrio et al., 2008;Charrier and Anastasio, 2011).Lastly, because the 50 µM of ascorbate used in our PM extraction solution is at the lower end of human lung lining fluid Asc levels (Cross et al., 1994;van der Vliet et al., 1999), we expect more • OH production at higher ascorbate concentrations.

Conclusions
We have quantified the formation of • OH in cell-free aqueous extracts of PM from an urban and rural site in the San Joaquin Valley of California.Although the sample size is small, our results show that: (1) in general, the urban (Fresno) samples generate more • OH than the rural (Westside) samples; (2) normalized by air volume, the fine PM generally makes more • OH than the corresponding coarse PM; (3) normalized by PM mass, the coarse PM typically generates more • OH than the fine PM; (4) the presence of a physiologically relevant level of ascorbate in the extraction solution greatly enhances the formation of • OH, and (5) transition metals, especially SLF-soluble Cu, play a dominant role in • OH generation from the SJV PM.While it is difficult to extrapolate from our acellular results to possible in vivo effects, an estimate of the lung lining fluid • OH burden suggests that • OH generation from inhaled particles could potentially cause toxicity at high particle levels.

Figures Figure 1 Fig
Figures

Fig. 9 .
Figure 9 • OH formed after 24 h of extraction (i.e. the total amount of • OH formed during the 24 h).Our initial rate of • OH formation likely underestimates the true value since we used 1 h instead of an earlier time point for the rate calculation.Likewise, since samples typically formed Field Blank Rate Daily Positive Control Rate − Daily SLF Blank Rate • OH throughout the extraction, the 24-h • OH value will often underestimate the true maximum.The rate of • OH formation in each PM extract was blankcorrected, positive-control-normalized, and expressed relative to the sampled air volume using www.atmos-chem-phys.net/11/9671/2011/Atmos.Chem.Phys., 11, 9671-9682, 2011 9674 H. Shen and C. Anastasio: Hydroxyl radical from San Joaquin Valley particles Corrected Rate of • OH (nmol h −1 m −3 ) (3) = Sample Rate −

DiStefano et. al., 2009)  OH Rate (nmol/h/mg)
Correlation between the air-volume-normalized initial rate of • OH generation in SLF with Asc and the accompanying SLFsoluble Cu concentration.Values are means ± SD. n = 6 except for a few of the Cu concentrations where n = 4.The initial rates of • OH formation by Fresno fine and coarse PM were strongly correlated with the SLF-soluble Cu concentrations in corresponding PM extracts: y = 2.08x +0.03, R 2 = 0.98.No correlation was observed between the SLF-soluble Cu concentrations and the initial rates of • OH formation from the Westside PM (R 2 = 0.03).