Second harmonic scattering investigation of bacterial efflux induced by the antibiotic tetracycline

Abstract

Efflux pumps are a key component in bacteria’s ability to gain resistance to antibiotics. In addition to increasing efflux, new research has suggested that the antibiotic, tetracycline, may have larger impacts on bacterial membranes. Using second harmonic scattering, we monitor the transport of two small molecules across the membranes of different Gram-positive bacteria. By comparing our results to a simple kinetic model, we find evidence for changes in influx and efflux across both bacterial species. These changes, however, are probe-dependent, opening new questions about the localization of the drug’s effects and the specificity of the efflux pumps involved.

I. INTRODUCTION

The plasma membrane is an intricate and dynamic bilayer encasing the delicate cytoplasm of the bacterial cell, providing structure and protection from harmful external forces.^1–4 In order to protect themselves from antibiotics, bacteria can alter their membrane in multiple ways to suppress the influx and accelerate the efflux of foreign molecules.^5,6 Efforts to overcome these protection mechanisms in order to develop new, or alter current, medicinal classes require understanding how exposure to small molecule drugs alters bacterial membranes.

Tetracycline (Tc) is a small-molecule antibiotic [Fig. 1(a)] that has been the basis of multiple alteration-resistance cycles for decades.^7,8 Discovered in 1953, resistance to Tc arose within a few years of its introduction.⁹ Since then, multiple structural variations have led to more successful derivatives,^10,11 some of which are still in clinical use at present.^12,13 Despite these successes, work continues to identify new directions for improvement to broaden their application and address emerging threats to their efficacy.¹⁴

FIG. 1.

Chemical structures of (a) tetracycline (Tc), (b) FM 1-43, and (c) malachite green (MG) molecules. For the SHS experiments, the bacterial cells are incubated with different concentrations of Tc before the probe molecule, FM 1-43 or MG, is added for the measurement.

Tc’s primary mechanism of action is to interrupt protein synthesis by binding to the ribosome.⁹ Recently, Wenzel et al. provided evidence for a second mechanism of Tc action, where it directly perturbs the bacterial plasma membrane.¹⁵ To combat Tc, bacteria have developed multiple resistance pathways, including ribosomal protection and, most notably herein, upregulation of efflux.^10,11,14

Efflux pumps are proteins embedded within bacterial membranes that eject foreign substances from the inner membrane leaflet or cytoplasm to the cell exterior. There is a wide range of different proteins that act as efflux pumps and vary not only in their energy sources but also in their ability to accept substrates from the cytoplasm and/or inner membrane leaflet and deposit them to the outer leaflet and/or extracellular space.^16–18 Efflux is a key component in a wide range of different antibiotic resistance mechanisms,⁶ but despite its importance, less than half of the multidrug efflux pumps in the Gram-positive bacterium, Staphylococcus aureus, have been studied, with the majority of the function and specificity of these systems remaining unknown.¹⁹ In fact, despite decades of research, recent candidates have been identified within the last year.^20,21

Most methods to measure efflux rely on the use of fluorescent probes whose properties change depending on their immediate environment.²² For instance, efflux activity can be measured directly by pre-loading a dye in the cell colony and monitoring the fluorescence of a probe molecule before and after efflux pumps are activated.²² An indirect measurement consists of recording fluorescence intensity before and after dye addition to cells and measuring their respective steady-states.²² As efflux activity increases across strains, the steady-state accumulation decreases. In this case, the rate of influx may disrupt efflux quantification, making it less favorable than direct measurement. Despite the multitude of available methods, however, none are capable of providing quantitative measurements for both influx and efflux with a single technique.

As an alternative to fluorescence-based approaches, we show here how second harmonic scattering (SHS) provides an opportunity to observe the impacts of both small-molecule influx and efflux in a single time trial. SHS has been heavily utilized to study a wide range of biological membranes and continues to offer unique insights into understanding interfacial interactions, including eukaryotic systems,^23–25 bacterial cells,^26–35 and model membranes.^29,36–42 The SHS signal requires a noncentrosymmetric surface to upconvert two photons of incident light into one photon of twice the frequency, yielding unique surface-specificity.⁴³ This phenomenon has been detailed elsewhere.^28,29

In this work, SHS is used to monitor the impact of Tc on the transport of small molecules through the membranes of two different Gram-positive species, Bacillus subtilis and S. aureus. Comparison of the experimental results to a simple kinetic model shows how alterations in influx and efflux manifest in distinct temporal signatures. Perturbations to the membrane transport behavior are shown to be dependent on the small molecule employed, prompting questions regarding the specificity and localization of the Tc impact.

II. MATERIALS AND METHODS

A. Cell preparation

S. aureus (ATCC 27217) and B. subtilis 168 cells were grown on brain heart infusion (BHI) agar plates and LB agar plates, respectively, and single colonies were inoculated overnight in BHI broth with a 5:1 headspace ratio statically and aerobically at 250 rpm, respectively, at 37 °C. After overnight incubation to the stationary phase, the cells were then diluted in fresh BHI or LB with a 5:1 headspace to an optical density at 600 nm (OD₆₀₀) of 0.01. After reaching an OD₆₀₀ of ≈0.200, the cells were centrifuged to create pellets (11 ml for 10 min at 7000 rpm for S. aureus, 3000 rpm for B. subtilis), which were washed twice with phosphate saline buffer (PBS) in the same manner and then resuspended in a BHI solution containing 2% oxyrase.²⁶ The final OD₆₀₀ of the solution was ≈0.200. For cells influenced by Tc, the antibiotic was added prior to washing and incubated at room temperature for 30 min.

Prior to cell introduction, the flow cuvette with a path length of 2 mm was treated with 1M HCl for ∼30 min to cleanse the sample volume, which was then flushed with 50 ml of sterile Milli-Q water and 10 ml of Tris buffer (pH adjusted to 7.4). The sample volume was then flushed and filled with 10 ml of bovine serum albumin (BSA), which consisted of 50 mM BSA in 10 ml of cold Tris buffer to ensure BSA stabilization, and sterilized with a 0.2 μm filter. After 10 min, a solution consisting of 10 ml of cold Tris and 1% glutaraldehyde was flushed and allowed to sit in the cuvette for an additional 10 min. After a final flush with 10 ml of Tris, the tubing was attached with a pump, and PBS was allowed to flow through the tubing and cuvette before replacement with the cell solution. This BSA/glutaraldehyde cross-linking passivation method has been previously described in the literature.⁴⁴

B. Tc and dye preparation

Tc stock solution was prepared in an 80:20 H₂O:DMSO solvent to a final concentration of 3.8 μM (S. aureus) and 4.5 μM (B. subtilis) minimal inhibitory concentration (MIC) and 2×MIC in solution.^15,45 The FM 1-43 stock solution was prepared in an 80:20 H₂O:DMSO solvent for a final concentration of 16 μM in solution, while the MG stock solution was prepared in H₂O for a final concentration of 25 μM in solution.^28,35 For stocks requiring a solvent, the final DMSO concentration within the cell solution was 0.2% after addition. FM 1-43 was obtained from Biotium. MG and Tc were obtained from Sigma Aldrich. All chemicals were used as received.

C. Time-resolved second harmonic scattering

Time-resolved SHS signals were collected on a home-built spectroscopy instrument with a Griffin Ti:sapphire oscillator (KM Labs) centered at 800 nm with a repetition rate of 80 MHz and a pulse duration of ∼55 fs. The average power at the sample position was 100 mW. A 750 nm longpass filter (Edmund Optics) was utilized alongside a focusing lens used to prohibit second harmonic response from routing optics from being collected and increase the photon density at the sample position, respectively. The resulting scattering was collected and filtered with a 725 nm shortpass filter (Edmund Optics) and a 450 nm dichroic mirror (Edmund Optics). The second harmonic scattering signal was then filtered with a 400/25 nm bandpass (Edmund Optics) filter prior to detection with a photon-counting photomultiplier tube (Hamamatsu, H7421-40). Time trials consisted of 4000 s collection periods with a 250 ms integration time. Cell background signal was collected for 200 s prior to probe addition. This background signal arises predominantly from hyper-Rayleigh scattering from the cells in solution and is subtracted prior to the normalization of the trSHS data. The changing SHS signal, arising from the electronic enhancement from the probe molecules, was collected over time as the probe molecules interacted with the bacterial cell membranes and analyzed with in-house LabVIEW and MATLAB codes. All experimental trSHS figures were plotted with a binning time of 20 s. An average of three trials is shown with error bars, as described in the caption for each figure. All experiments were conducted at room temperature.

III. RESULTS AND DISCUSSION

A. Small molecule transport in Tc-exposed B. subtilis

To monitor Tc’s influence on the B. subtilis membrane, probe molecule FM 1-43 was utilized. While other tetracycline species have previously been directly observed with second harmonic spectroscopy,⁴⁶ the SHS signal from Tc was found to be insufficient when performing measurements on living cells. In contrast, FM 1-43 is an amphiphilic molecule [Fig. 1(b)] that produces a strong, resonantly enhanced SHS signal, which has facilitated previous documentation of its movement within the lipid bilayer of bacteria.^28,47 As such, after a 30-min incubation with different concentrations of Tc, FM 1-43 was introduced and tracked with time-resolved SHS. It should be noted that the cells are washed after the Tc incubation to minimize any potential interaction between the antibiotic and probe molecule. While some Tc will remain in the cells after this washing step, it is expected that a significant fraction will move from the membrane to the ribosomal target,⁹ and any remaining in the membrane will not contribute to our signal due to its low SHS response.⁴⁶ With no Tc incubation, FM 1-43 favors translocation into the inner leaflet of the membrane after its initial adsorption. This can be seen by the light blue trend in Fig. 2, where the initial rapid signal increase is first caused by the adsorption of FM 1-43 into the outer leaflet. The subsequent decrease in signal is due to the increase in probe population within the inner leaflet of the B. subtilis membrane as the molecule translocates. After incubation with Tc, there is an alteration in FM 1-43’s behavior, where the orange trend in Fig. 2 represents FM 1-43’s membrane dynamics after incubation with Tc at the MIC for our B. subtilis strain.¹⁵ When compared to untreated cells, FM 1-43’s dynamics change dramatically in two ways. First, it is apparent that the rate by which the SHS signal decreases is slowed. Second, the minimum SHS signal achieved is at a higher intensity, suggesting a reduced population on the inner leaflet and less signal cancellation. Both of the features are further exacerbated at 2xMIC, represented by the yellow trend in Fig. 2.

FIG. 2.

Normalized trSHS experimental data points with 16 μM FM 1-43 on untreated B. subtilis (light blue) and B. subtilis treated with 1xMIC Tc (orange), 2×MIC Tc (yellow). For each condition, N = 3. The shaded regions represent the error of the averaged values.

A reduced population of FM 1-43 on the membrane inner leaflet could arise from two likely sources. First, translocation of the molecule from the outer leaflet could be hindered. This hypothesis is supported by the work of Wenzel et al., who suggested that Tc resides within the hydrophobic space between the lipid leaflets in B. subtilis perturbing the membrane.¹⁵ Alternatively, FM 1-43 molecules may be removed from the inner leaflet via increased efflux to result in a lower overall inner leaflet population. While it is well-established that efflux is increased by bacteria to aid in resistance to Tc,^9,11 many of these protein pumps are specific for a molecular substrate.^48,49 To better visualize how these two different processes would be expected to manifest in the SHS signal dynamics, we turn to a simple kinetic model.

B. Kinetic model for SHS of small molecule transport in Gram-positive bacteria

Our approach to modeling the impact of influx and efflux rates was first described by Clarke⁵⁰ and is pictorially summarized in Fig. 3. It closely follows the approach previously taken by others to simulate time-resolved SHS signals.^30,36,51,52 In comparison to the kinetic model of Wu et al. for SHS of molecular transport into unilamellar liposomes,³⁶ the most significant difference is the separation of rates for molecules moving from one leaflet to the other. In our model, the rate of efflux describes the movement from the inner leaflet to the outer leaflet. It should be noted that multiple studies have shown how efflux pumps can deposit their cargo into the outer leaflet of the membrane in addition to the extracellular space.^16,17 In addition, the molecular cargo can be extracted from the inner leaflet of the membrane.^17,18 Given the lipophilic nature of FM 1-43, it is appropriate to assume that the majority of its population will reside in the membrane environment as opposed to free in solution.⁵³ As such, our efflux model only considers movement between leaflets.

FIG. 3.

Cartoon schematic depicting the simple kinetic model of the binding and membrane transport steps and their associated rate constants for an FM 1-43 molecule within the Gram-positive lipid bilayer. The variables k_+/− are adsorption/desorption rate constants, respectively, while k_in/out are the influx and efflux rate constants. The values for S_o/i are the number of molecules in solution outside and inside the cell, while N_o/i represents the number of molecules in the outer and inner membrane leaflets, respectively. In our simulation, we are assuming that the rate of k_out is increased by an efflux pump, which is depicted in blue. Created with BioRender.com.

The migration of a small molecule into and across the membrane is composed of multiple distinct steps. The first step is molecular adsorption of the probe molecule from the bulk solvent environment into the outer leaflet of the cell membrane [Eq. (1)]. Once in the membrane, the molecule can translocate to the inner leaflet [Eq. (2)]. Molecules in the inner leaflet can either be effluxed back to the outer leaflet [Eq. (2)] or desorb into the cytoplasm [Eq. (3)]. The equations for these processes are as follows:

$$S_o+E_o\underset{{\mathrm{k}}_{-}}{\overset{{\mathrm{k}}_{+}}{\rightleftharpoons }}{N}_o,$$ (1)

$$N_o+E_i\underset{{\mathrm{k}}_{\text{out}}}{\overset{{\mathrm{k}}_{\text{in}}}{\rightleftharpoons }}{N}_i+E_o,$$ (2)

$$S_i+E_i\underset{{\mathrm{k}}_{-}}{\overset{{\mathrm{k}}_{+}}{\rightleftharpoons }}{N}_i,$$ (3)

where S_o is the concentration of molecules in the bulk solution, N_o is the concentration of molecules in the outer leaflet, N_i is the concentration of molecules in the inner leaflet, and S_i is the concentration of molecules in the cell interior. E_o and E_i are the concentrations of empty sites on the outer and inner leaflets, respectively. The empty sites can be calculated by subtracting the number of occupied sites from the maximum available (e.g., $E_o=N_o^{\max }-N_o$). In addition, we assume that the number of empty sites is the same in both membranes $(N^{\max }=N_o^{\max }=N_i^{\max })$. Such an approximation has been used previously for bacterial systems.³⁰

The dynamic concentrations of the probe molecules in the different positions outlined above can be calculated through the following set of coupled differential equations:

$$\frac{\mathrm{d}{S}_o}{\mathrm{d}t}=-k_{+}{S}_o{C}_c(N^{\max }-N_o)+k_{-}{N}_o,$$ (4)

$$\frac{\mathrm{d}{N}_o}{\mathrm{d}t}=k_{+}{S}_o{C}_c(N^{\max }-No)-k_{-}{N}_o-k_{in}{N}_o+k_{\mathit{out}}{N}_i,$$ (5)

$$\frac{\mathrm{d}{N}_i}{\mathrm{d}t}=k_{+}\frac{S_i}{N_A{V}_i}(N^{\max }-N_i)-k_{-}{N}_i+k_{in}{N}_o-k_{\mathit{out}}{N}_i,$$ (6)

$$\frac{\mathrm{d}{S}_i}{\mathrm{d}t}=-k_{+}\frac{S_i}{N_A{V}_i}(N^{\max }-N_i)+k_{-}{N}_i.$$ (7)

In addition to the variables described above, C_c is the concentration of cells, and V_i is the volume of the intracellular space. Finally, while the adsorption and desorption rates are not assumed to be the same (k₊ ≠ k₋), they are linked through the dissociation constant $(K_d=\frac{k_{-}}{k_{+}})$, which has been previously measured with SHS.^27,28 Given that the initial SHS signal intensities between treated and untreated cells are within error of each other (Fig. S1), we assume that Tc treatment does not significantly alter the dissociation constant.

Given the interface-specific nature of SHS, the time-resolved SHS signal can be calculated from the N_o and N_i populations,

$$I_{\mathit{SHS}}\sim |(N_o-N_i){|}^2.$$ (8)

Simulations with the model are carried out using variables extracted from the literature reports. The concentration of cells is calculated from an expected cell density for our experimental OD₆₀₀ of 0.2.⁵⁴ The adsorption rate (k₊) of FM 1-43 onto lipid vesicles has previously been measured to be $\sim 200$ mM⁻¹ s⁻¹.⁵⁵ Assuming a rate of 6 × 10⁻⁴ s⁻¹ for k_in,³⁰ we can see the impact of different efflux rates (k_out) on the SHS signal in Fig. 4. While many of these chosen values arise from experiments on model lipid vesicles, we are unable to fit them directly to the experimental data as the model is overdetermined and does not produce a unique fit. While the simulation captures the increased long-time SHS signal as the efflux rate is increased, the initial decay rates from the peak signal do not match what is observed in the experimental data. The simulation shows that all efflux rates maintain a similar signal decrease after the initial signal from molecular adsorption. In contrast, higher Tc concentrations result in a slower decrease in FM 1-43 SHS signal from its peak intensity with B. subtilis (Fig. 2).

FIG. 4.

Simulated time-resolved SHS signal for different values of efflux, k_out, using the described kinetic model. The k_out values were set as fractions of k_in to best match the experimental data in Fig. 2 and are k_out = 12k_in (yellow), k_out = 10k_in (orange), and k_out = 5k_in (blue). Increasing the efflux rate shows little change in the initial signal decay rate but increases signal plateaus at long times.

Altering the initial decay rate is possible if k_in changes, as shown in Fig. 5. In this case, however, the long-time SHS signal remains more sloped than seen in the experiment or the efflux-varying simulation. As such, the data suggest that Tc is altering both efflux and translocation to the inner leaflet, as expected by the membrane perturbation Wenzel et al. noted.¹⁵

FIG. 5.

Simulated time-resolved SHS signal for different values of influx, k_in, using the described kinetic model. The k_in values were set as multiples of k_out to best match the experimental data in Fig. 2 and are k_out = 0.1k_out (yellow), k_out = 0.15k_out (orange), and k_out = 0.2k_out (blue). Decreasing the influx rate yields a slower signal decay throughout the measurement time period.

C. Small molecule transport in Tc-exposed S. aureus

To assess the broader implications of these findings, we performed additional experiments altering both the bacterial species and SHS probe molecule. Figure 6 shows the SHS response for the impact of Tc on the movement of FM 1-43 through the bacterium, S. aureus. Similar to B. subtilis, S. aureus is a Gram-positive species having a single lipid membrane, but there are significant differences between the membranes of these bacteria. For example, B. subtilis contains a thicker cell membrane than S. aureus,^56,57 and $\sim 90\%$ of fatty acids (FAs) in B. subtilis are branched as compared to $\sim 50\%$ of FAs in S. aureus.^58,59 In addition, B. subtilis contains protein machinery that has been shown to facilitate regions of increased fluidity.⁶⁰

FIG. 6.

Normalized trSHS experimental data points with 16 μM FM 1-43 on untreated S. aureus (blue) and S. aureus treated with 1×MIC Tc (orange), 2×MIC Tc (yellow). For each condition, N = 3. The shaded regions represent the error of the averaged values.

The strain of S. aureus we chose for these experiments, 502 A, is known to exhibit some Tc-resistance, which is immediately apparent when comparing its FM 1-43 SHS response without Tc addition (Fig. 6, blue) to that of B. subtilis (Fig. 2, light blue). While the FM 1-43 signal decreases in both bacterial samples after its adsorption peak, the S. aureus signal stabilizes after 20 min and then begins to rise again while the signal in B. subtilis continues to decrease over the entire course of the measurement. The mutations associated with the S. aureus strain have been previously linked to efflux pump expression,⁶¹ which explains why the signal without Tc more closely matches simulations with higher rates of efflux. As Tc is introduced to the S. aureus cells, the FM 1-43 SHS signal begins to more closely match that observed in B. subtilis. It can be seen that a higher Tc concentration is correlated with slower signal decay and a higher minimum in the SHS intensity observed. Overall, the mechanism by which Tc exposure alters FM 1-43 transport appears to be conserved between these bacterial species.

While others have used liposomes for comparison to bacterial trSHS responses,⁶² these models are significantly limited in their ability to reproduce the compositional asymmetry and lateral heterogeneity of lipid species in natural membranes.^63,64 In contrast, cell starvation is a routine method for efflux pump activity determination.^22,65,66 In such experiments, bacteria are exposed to a minimal media environment to deplete available energy in the system used to initiate carrier-mediated transport.⁶⁷ This method was applied to this study to examine the impact of energy depletion on FM 1-43 dynamics before and after Tc exposure. Here, cells were grown and incubated as in previous studies but were then washed and resuspended in PBS that was supplemented with 2% oxyrase immediately before conducting the trSHS measurements.⁶⁷ The results shown in Fig. 7 display three key points. First, the rate of FM 1-43 moving into the cells is the same for untreated cells, cells starved in PBS, and cells in PBS that were previously incubated with Tc. Second, the rise observed at later times in the untreated cells is not present in the starved trials, indicating that this organization process is energy-dependent. Third, changes induced by incubation with Tc are unable to alter the FM 1-43 transport dynamics if cells are starved. This final point further supports our previous assignment of the Tc-altered signals arising from upregulated efflux.

FIG. 7.

Normalized trSHS experiment with 16 μM FM 1-43 on untreated S. aureus (blue) in BHI, as well as starved S. aureus cells in PBS with (red) and without (yellow) Tc. For untreated, N = 3. For both starvation averages, N = 2. The shaded regions represent the error of the averaged values.

We next turn to consider whether the cell’s response to Tc impacts all small molecule transport through the membrane. Malachite green (MG) is a small, positively charged dye [Fig. 1(c)] that has been extensively involved in membrane research regarding model^36,40,42,68 and eukaryotic systems,²³ as well as Gram-negative bacteria such as Escherichia coli.^{30,33,52,69,70} In regards to its behavior on Gram-positive bacterial membranes, MG tends to passively diffuse across the bilayer more rapidly than FM 1-43.⁴⁷ Our results in Fig. 8 show that MG transport is unaffected by Tc exposure in S. aureus. As the molecule crosses the lipid bilayer in S. aureus cells, the initial spike in intensity that is caused by adsorption to the outer leaflet of the membrane dramatically decreases as the molecule transports through the bilayer. There is no change in the minimum SHS, which decays back to the baseline for both conditions, and the timescale of the signal decrease is also unchanged. This observation differs from previous SHS experiments of MG with E. coli, where MG has been used to study efflux behavior.⁷¹

FIG. 8.

Normalized trSHS experimental data points with 25 μM MG on untreated S. aureus (blue) and S. aureus treated with 1×MIC Tc (orange). For each condition, N = 3. The shaded regions represent the error of the averaged values.

Understanding the different responses of FM 1-43 and MG warrants further discussion. Wenzel et al. noted localized aberrations in the membrane of Tc-exposed cells arising from membrane invaginations.¹⁵ In our previous work with MG and FM 1-43 and S. aureus cells, we presented evidence that these probes interact with different localized regions of the bacterial membrane.⁴⁷ As such, these Tc findings further support the distinct transport and interaction behavior between FM 1-43 and MG and the complementary information that can be provided by employing both in future studies.

When considering the probes’ efflux response, MG is a quaternary ammonium compound (QAC) with known efflux pump binding in S. aureus;^72,73 however, we see no evidence of active efflux of this compound in our studies or impact of Tc on its behavior. In contrast, we are unaware of any previous work linking FM 1-43 with bacterial efflux. There are, however, many possible efflux pumps for which FM 1-43 may be a suitable substrate. A plurality of multidrug efflux pumps have been identified as being induced by tetracycline,⁷⁴ but over half of the prospective efflux pump proteins in S. aureus have yet to be studied sufficiently to assess their substrate specificity.²¹

IV. CONCLUSIONS

In summary, our SHS data provide evidence of a Tc concentration-dependent alteration in the membrane dynamics of FM 1-43 with two Gram-positive bacteria strains, S. aureus and B. subtilis. A simple kinetic model shows that observed alterations are consistent with both hindered influx and increased efflux, in agreement with known Tc impacts on bacteria.^9,15 As such, our SHS approach provides a new way to monitor both influx and efflux in a single experiment. An important element of this approach is the use of the FM 1-43 probe. While it is unknown what protein pump is responsible for its efflux, its behavior is conserved over the two bacterial species studied. In contrast, the more commonly studied probe, MG, does not exhibit sensitivity to Tc exposure in these systems. Future studies will explore the capability to extend this approach to other antibiotics, including more recently developed Tc derivatives as well as those from different structural classes.

SUPPLEMENTARY MATERIAL

The supplementary material contains a figure of raw trSHS intensities.

AUTHOR DECLARATIONS

Conflict of Interest

The authors have no conflicts to disclose.

Author Contributions

Eleanor F. Page: Conceptualization (equal); Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Software (equal); Visualization (equal); Writing – original draft (equal); Writing – review & editing (equal). Mikala F. Blackmon: Investigation (supporting). Tessa R. Calhoun: Conceptualization (lead); Formal analysis (equal); Funding acquisition (lead); Methodology (equal); Project administration (lead); Software (equal); Supervision (lead); Writing – original draft (equal); Writing – review & editing (equal).

DATA AVAILABILITY

The data that support the findings of this study are available from the corresponding author upon reasonable request.