Mammalian histones facilitate antimicrobial synergy by disrupting the bacterial proton gradient and chromosome organization

First proposed as antimicrobial agents, histones were later recognized for their role in condensing chromosomes. Histone antimicrobial activity has been reported in innate immune responses. However, how histones kill bacteria has remained elusive. The co-localization of histones with antimicrobial peptides (AMPs) in immune cells suggests that histones may be part of a larger antimicrobial mechanism in vivo. Here we report that histone H2A enters E. coli and S. aureus through membrane pores formed by the AMPs LL-37 and magainin-2. H2A enhances AMP-induced pores, depolarizes the bacterial membrane potential, and impairs membrane recovery. Inside the cytoplasm, H2A reorganizes bacterial chromosomal DNA and inhibits global transcription. Whereas bacteria recover from the pore-forming effects of LL-37, the concomitant effects of H2A and LL-37 are irrecoverable. Their combination constitutes a positive feedback loop that exponentially amplifies their antimicrobial activities, causing antimicrobial synergy. More generally, treatment with H2A and the pore-forming antibiotic polymyxin B completely eradicates bacterial growth.

A ntibiotic resistance is a worldwide epidemic. To develop new treatments, a better understanding of natural defenses may be helpful. As a first line defense, neutrophils mediate the host's response partly through establishment of neutrophil extracellular traps (NETs) [1][2][3][4][5] . NET formation is stimulated by virulent microorganisms that interfere with phagosomal killing, such as aggregates of pathogenic bacteria 6 , including Pseudomonas aeruginosa 7 , Escherichia coli 8 , and Staphylococcus aureus 9 , and fungal hyphae 10 . Histones and antimicrobial peptides have potent antimicrobial activity in NETs, but how the individual and combined effects of these components inhibit bacterial growth has not been determined 11,12 .
Histones, originally proposed as antibacterial agents 13,14 , are essential for NET-mediated antimicrobial activity 1 . However, extracellular histones can have toxic effects, triggering autoimmune and inflammatory responses 15 , mediating mortality in sepsis 16 , inducing thrombosis 17 , and activating pro-inflammatory signaling through the toll-like receptors TLR2 and TLR4 18 . Thus, levels of extracellular histones must be tightly controlled.
Importantly, the histones' bacterial killing mechanism is unclear. The bulk of histone antimicrobial activity has been observed in low-ionic non-physiological solutions. At physiological magnesium levels, histones are less effective/ineffective at killing bacteria [19][20][21][22][23][24][25] . Since histones contribute critically to NET activity 1 , but are not effective alone under physiological conditions, it is likely that histone antimicrobial activity requires coordination with other immune cell components 26 .
Antimicrobial peptides (AMPs) are broad-spectrum antimicrobials 27 that co-localize with histones in NETs 1 . Many AMPs kill bacteria by forming transient pores that induce permeabilization of microbial membranes 12,[28][29][30][31] . The co-localization of AMPs and histones suggests joint function. Both histones and AMPs are comparable in size, between 14-18 kDa 32,33 , are cationic, contain a high proportion of hydrophobic amino acids, and possess the ability to form alpha helices. However, the ability of histones to condense mammalian DNA, a property that LL-37 lacks, raises the possibility of additional separate antimicrobial functions.
Here, we show that histone H2A and AMP LL-37 have distinct antimicrobial effects, and that together they constitute a selfamplifying, synergistic antibiotic mechanism. LL-37 forms pores that enable the entry of H2A into bacteria. H2A enhances the pores, stabilizing them and allowing entry of additional LL-37 and H2A. Once inside, H2A reorganizes bacterial chromosomal DNA, and inhibits transcription, which kills bacteria directly. Importantly, these activities are observed under physiological conditions. The combined LL-37/H2A effects are much greater than their individual effects, resulting in a synergistic antimicrobial interaction. This selfamplifying mechanism is general in nature, extending to other histones, including H3, and other AMPs, including magainin-2.

Results
H2A antimicrobial activity requires membrane permeabilization. Cations stabilize the outer membrane of bacteria. We hypothesized that decreasing Mg 2+ concentrations would destabilize the bacterial membrane, increasing H2A entry and bactericidal activity. Based on free magnesium levels in human plasma and extracellular fluids 34 , killing experiments were performed using two magnesium concentrations: 1 mM (physiological concentration) and 1 μM (low concentration). We assayed antimicrobial activity using 10 μg/mL histone H2A based on the finding that 15 μg/mL histones are detected in the blood plasma of baboons after E. coli challenge 16 .
H2A treatment of E. coli or S. aureus decreased colony-forming units (CFUs) on agar plates at low magnesium ( Supplementary   Fig. 1A), but not at physiological magnesium ( Supplementary  Fig. 1A). Similarly, in liquid cultures, H2A inhibited bacterial growth only at low magnesium (Fig. 1a, b), as measured by optical density. We note that low magnesium decreased total bacterial growth, consistent with previous reports. The inability of S. aureus to recover in low magnesium environments may be due to a higher sensitivity to histones in low magnesium environments.
We investigated whether H2A disrupts membranes by using propidium iodide (PI), which fluoresces upon binding nucleic acids and does not permeate the outer membranes of viable bacteria. H2A induced PI fluorescence in E. coli in low magnesium ( Fig. 1c and Supplementary Fig. 1B), but no PI fluorescence increase was observed at physiological magnesium ( Fig. 1c and Supplementary Fig. 1B), suggesting that H2A inhibits growth in low magnesium by enhanced membrane permeabilization. However, H2A-induced PI fluorescence could in principle reflect a bacterial response that induces cell death, where membrane permeabilization could be a secondary effect.
We reasoned that increased membrane destabilization due to low magnesium facilitated H2A entry. If so, membranepermeabilizing agents could similarly increase histone entry. LL-37 is a human cathelicidin AMP that co-localizes with histones in NETs, exhibits broad-spectrum microbial activity, and disrupts lipid bilayers by forming toroidal pores 30 . LL-37 production is elevated in tissues that are exposed to microbes, such as skin and mucosal epithelia, for rapid defense against microbial infections 35 . We hypothesized that LL-37 pores could increase H2A entry.
We treated E. coli with LL-37 and H2A at physiological magnesium (1 mM) to avoid membrane stress from low ionic conditions. Treatment with 2 μM LL-37, a concentration reported to be the bulk minimum inhibitory concentration (MIC) of E. coli after 12 h 36 and a concentration below that found in inflamed epithelial cells 37 , decreased the growth rate and slightly extended the lag time (Fig. 1d). H2A alone had no effect on E. coli growth. However, cultures treated with both H2A and LL-37 had significantly decreased growth rates compared to untreated or LL-37-treated samples. Similar effects on growth were observed using S. aureus (Fig. 1d), suggesting that treatment of Grampositive or Gram-negative bacteria with LL-37 enhances the antimicrobial activity of H2A. Treatment using both H2A and LL-37 increased the PI fluorescence of E. coli after 1 h, indicating that increased membrane permeabilization accompanies the enhanced antimicrobial activity of H2A (Fig. 1e). Synergy is defined as an effect that is greater than the sum of each of the constituents. LL-37 and H2A are synergistic: the combined treatment inhibited growth to a larger degree than the two individual effects combined. Synergistic killing was also observed using LL-37 and histone H3 in place of H2A ( Supplementary  Fig. 1C), suggesting that synergy is a general property between histones and AMPs. The synergistic killing effect was diminished using citrullinated H3, which suggests histone citrullination could affect antimicrobial synergy.
Bacterial growth was not completely inhibited by treatment of LL-37 and H2A, with renewed growth observed after~15 h (Fig. 1d). We suspect a small fraction of resistant mutants or phenotypic variants give rise to this 38,39 . The lack of complete growth inhibition was similarly observed in treatments with the bacteriostatic antibiotic chloramphenicol and bactericidal antibiotic kanamycin ( Fig. 1f and Supplementary Fig. 1D), indicating that a lack of complete growth inhibition is not specific to H2A and LL-37 and may be a general property of antibiotic treatments in liquid cultures. To determine whether the combined treatment of LL-37 and H2A was bactericidal or bacteriostatic, E. coli were treated for 1 h and plated on agar plates that did not contain LL-37 or H2A (Fig. 1g). A significant decrease in CFUs was observed, suggesting the combined H2A/LL-37 treatment is bactericidal.
The synergistic H2A/LL-37 effects were striking at the subcellular level, as measured via scanning electron microscopy (SEM) (Fig. 1h). In E. coli treated with either LL-37 or H2A, few cell morphological differences were observed. However, the combined LL-37/H2A treatment caused dramatic cellular damage, including cell aggregation and extensive production of insoluble components to the outer surface of the membrane and to the surrounding surfaces. In some cells, the dual treatment   Fig. 1E), suggesting the LL-37/H2A combination induces membrane damage. We note that blebs were not observed in all cells and attribute this to the transient nature of the observed blebs, which is discussed below. H2A also frequently induced aggregation of cells by fusing the poles of cells together (Fig. 1h). The linkage cannot be attributed to inhibition of cell division, as aggregates contain many more cells linked together than can be duplicated through bacterial replication during the course of treatment ( Supplementary Fig. 1F). Previous reports noted that positively charged molecules accumulate at the bacterial cell poles, where the Gaussian curvature is highest 40 . The bacterial aggregation here may be explained by large positive charge accumulation on the outside of the cells at the poles. LL-37 treatment did not induce aggregation and instead caused a significant reduction in cell size, consistent with induction of membrane permeabilization ( Supplementary Fig. 1G). We investigated potential synergy between H2A and the aminoglycoside kanamycin or the amphenicol chloramphenicol at physiological magnesium conditions. These antibiotics inhibit growth through protein translation inhibition 41 and cause membrane disruption 42 . Treatment of E. coli with either antibiotic alone increased the growth lag time (Fig. 1f). Combining H2A with either antibiotic had no additional antimicrobial effect, which suggests that these antibiotics do not synergize with H2A. Similarly, treatment using kanamycin concentration well above the MIC did not synergize with H2A ( Supplementary  Fig. 1D). Although aminoglycosides are reported to increase membrane permeabilization, this mechanism did not appear to form pores that enabled PI entry into E. coli. In addition, the combination of H2A and kanamycin had no effect on intracellular PI fluorescence of E. coli after a one-hour treatment (Fig. 1e). These results suggest that kanamycin and H2A do not synergize because kanamycin does not enable entry of H2A into cells. It is possible that incubation with kanamycin at higher concentrations, for longer times, or under different growth conditions would enable H2A entry into cells and produce synergy.
To determine if the H2A/LL-37 synergy was representative of a more general mechanism, we investigated the activity of H2A with the membrane-permeabilizing AMP magainin-2 (MAG2), an α-helical peptide belonging to a class of AMPs from the African claw frog (Xenopus laevis) 43 . Similar to LL-37, MAG2 is cationic and forms amphipathic α-helical structures in membranes. The 23-amino acid AMP forms a 2-3 nm toroidal pore, disrupting the ion gradient and inducing membrane permeabilization 44 .
H2A or 10 μM MAG2 alone, a concentration below the MIC for E. coli, had no effect on E. coli growth (Fig. 1i). However, cultures treated with both H2A and MAG2 extended the lag time significantly. Furthermore, combined H2A/MAG2 treatment significantly increased intracellular PI fluorescence after 1 h ( Supplementary Fig. 1H), indicating membrane permeabilization accompanies the enhanced antimicrobial activity of H2A.
H2A enters the cytoplasm and enhances LL-37 uptake. We fluorescently-labeled H2A with AlexaFluor488 (AF-H2A) to track H2A localization. AF-H2A bactericidal activity was confirmed via growth inhibition assays (Supplementary Fig. 2A). Treatment of E. coli and S. aureus using AF-H2A alone produced little or no cellular fluorescence (Fig. 2a, b), indicating lack of H2A uptake, consistent with Fig. 1c, e. Under low magnesium, enhanced AF-H2A uptake was observed (Fig. 2c, d), consistent with the ability of H2A to increase membrane permeabilization in low magnesium (Fig. 1c). Treatment using kanamycin or chloramphenicol did not induce AF-H2A uptake (Fig. 2b), consistent with the inability of these antibiotics to enable the entry of H2A. Importantly, membrane permeabilization by LL-37, polymyxin B (PMB), or MAG2 significantly enhanced AF-H2A uptake (Fig. 2a, b and Supplementary Fig. 2B, C). Coupled with the growth dynamics findings above (Fig. 1d, f), these results indicate that H2A-mediated growth inhibition is concomitant with the uptake of H2A into the cell.
While H2A (at 10 μg/mL) itself does not induce membrane permeabilization at physiological magnesium (Figs. 1e and 2b), it might enhance uptake through membrane pores formed by LL-37. We thus measured effects of H2A on LL-37 uptake using fluorescently-labeled LL-37 (5-FAM-LL-37). Fluorescence was observed in the cytoplasm using a treatment of LL-37 alone (Fig. 3a, b), consistent with a previous report 45 . Importantly, H2A significantly increased LL-37 uptake over the course of 1 h (Fig. 3a-c and Supplementary Fig. 2D). Furthermore, H2A increased the localization of LL-37 to the membrane ( Fig. 3c and Supplementary Fig. 2D). H2A did not shift LL-37 entirely to the membrane, as this would have been discernible as a rim pattern, such as that produced by the membrane dye FM4-64 ( Fig. 3c and Supplementary Fig. 2D). Together, these results suggest H2A enhances the effect of membrane pores by enabling greater uptake of LL-37 into the cytoplasm and inducing membrane localization of LL-37.
The enhancement by H2A of LL-37 membrane pores could significantly impact ion gradients across the membrane, disrupting ATP production. We thus measured the bacterial proton motive force (PMF) using the proteorhodopsin optical proton sensor (PROPS), which increases fluorescence with a loss in PMF due to electrical depolarization 46 . The change in fluorescence is due to protonation of a Schiff base on the proteorhodopsin inside Fig. 1 Histones and the antimicrobial peptides LL-37 and magainin-2 increase killing efficacy against bacteria. a, b Growth profiles, measured by optical density, of E. coli and S. aureus treated with H2A in media containing a low (1 μM) magnesium (n = 33 for each condition) and b physiological (1 mM) magnesium (n = 11 for each condition). c Intracellular propidium iodide (PI) fluorescence intensities of H2A-treated E. coli in 1 μM and 1 mM concentrations of magnesium after 1-h treatment (n = 3 for each condition). d Growth profiles of E. coli (n = 18 for each condition) and S. aureus (n = 8 for each condition) treated with 10 μg/mL H2A, 2 μM LL-37, or both in medium containing 1 mM magnesium. e Intracellular propidium iodide (PI) fluorescence intensities of E. coli treated with 10 μg/mL H2A, 2 μM LL-37, both H2A and LL-37, 10 μg/mL kanamycin (Kan), or H2A and Kan, in medium containing 1 mM magnesium (n = 3 for each condition). PI fluorescence of LL-37-treated E. coli and Kan-treated E. coli was normalized to H2A-treated cells. f Growth profiles of E. coli treated with 10 μg/mL H2A, H2A and 10 μg/mL chloramphenicol (Cam), or H2A and 10 μg/mL Kan in medium containing 1 mM magnesium (n = 6 for each condition). g Colony-forming units (CFU) of E. coli that were untreated (n = 5) or treated with 10 μg/mL H2A (n = 5), 2 μM LL-37 (n = 3), or both (n = 3) in minimal medium containing 1 mM magnesium. Bacteria were treated for 1 h before plating on non-selective LB agar plates. CFUs were normalized to H2A-treated E. coli CFUs in Supplementary Fig. 1A. h Scanning electron microscopy (SEM) images of E. coli treated with 10 μg/mL H2A, 1 μM LL-37, or both in medium containing 1 mM magnesium (n = 3 for each condition). i Growth profiles of E. coli treated with 10 μg/mL H2A, 10 μM MAG2, or both in medium containing 1 mM magnesium (n = 4 for each condition). Data shown as mean ± standard error of the mean (SEM) and are representative of biologically-independent experiments. One-way ANOVAs were performed. No adjustments were made for multiple comparisons. ns > 0.05. Scale bars represent 3 μm. the membrane. The fast dynamics of the reporter has captured electrical spiking in bacterial membranes due to changes in the PMF 46 . Treatment of E. coli with H2A had no effect on the PMF (Fig. 3d), consistent with the lack of membrane permeabilization by H2A (Fig. 1e). Cells treated with both H2A and LL-37 or with both H2A and PMB exhibited significantly higher PROPS fluorescence than cells treated with LL-37 or PMB alone, indicating that H2A further depolarizes the membrane and disrupts the PMF. We note that PROPS fluorescence is affected by the growth phase and pH 47 . All the cells were harvested at mid-exponential growth phase and were cultured in buffered medium, which minimizes potential changes in pH.
H2A inhibits bacterial recovery and membrane repair. We next tested bacterial recovery from H2A-induced damage. If H2A enhances uptake by membrane pores and depolarizes the membrane, these mechanisms would inhibit bacterial recovery even in the absence of AMPs and H2A. We quantified the extent of membrane repair in a strain of E. coli that expresses CFP under the control of a constitutively-active ompA promoter. E. coli were treated with LL-37 alone or in combination with H2A, washed to remove the treatments, and recovered in fresh medium lacking treatments. During the recovery period, cells previously treated with H2A or LL-37 alone resumed growth and division and retained expression of CFP (Fig. 4a). In contrast, cells treated with H2A and LL-37 formed membrane blebs at the mid-cell position (Fig. 4a). Within 10 min of the formation of membrane blebs, cells lost CFP fluorescence, indicating rapid leakage of cytoplasmic contents into the surrounding medium (Fig. 4a). Blebs were observed for only 10-20 min during the recovery period ( Fig. 4a), indicating that the structures were temporary. The formation of membrane blebs in dual-treated cells is consistent with the observation of membrane blebs in the SEM images ( Supplementary Fig. 1E). Membrane blebs were not observed in all cells, which we attribute to the transient nature of the blebbing events.
We quantified CFP fluorescence in cells during the recovery period to measure the extent of recovery following treatment with H2A or AMPs. To ensure that the analysis focused specifically on the ability to recover, cells were also stained with propidium iodide (PI). Cells that demonstrated significant membrane damage by the treatments (PI positive) were excluded from the analysis during the recovery period. CFP fluorescence increased or remained the same during the recovery period following treatment with only LL-37 or MAG2 (Fig. 4b,  high across the recovery period, consistent with the lack of antimicrobial activity from H2A alone. In contrast, during recovery following a combined treatment of H2A with either LL-37 or MAG2, CFP expression remained depressed (Fig. 4b, c), indicating persistent damage from which the cell cannot recover. The inability to recover from cell damage may reflect the specific combination of H2A with AMP-induced pores. To test this, we performed recovery experiments using low-magnesium medium in place of AMPs as an alternative method of increasing membrane permeability ( Supplementary Fig. 2E) and enabling initial AF-H2A uptake (Fig. 2c, d). Treatment with H2A in the low magnesium condition significantly decreased CFP fluorescence relative to the physiological magnesium condition (Fig. 4d).
However, CFP fluorescence was restored following a 60 min recovery period and increased to levels comparable to growth in a physiological magnesium environment (Fig. 4d). Thus, H2A causes persistent cell damage that is specific to AMP-induced pores, and causes only transient cell damage when the membrane is permeabilized using a non-AMP method such as growth in low magnesium medium.
Bacterial recovery from H2A was further assessed by monitoring PI fluorescence. After 1 h of recovery, PI fluorescence in LL-37-treated cells was significantly lower compared to cells that were treated with both H2A and LL-37 (Fig. 4e). H2A-treated cells showed consistently low PI fluorescence across the recovery period, indicating the lack of membrane damage by H2A. These results suggest that bacteria can repair membrane pores induced by LL-37 but the presence of H2A inhibits repair of LL-37induced pores. Similar results were observed using MAG2 in place of LL-37 (Fig. 4f). In addition, cells that were treated with H2A in low magnesium repaired membrane permeabilization within 30 min of the recovery period ( Supplementary Fig. 2F), which supports the model that inhibition of repair by H2A is specific to AMP-induced pores.
Thus, H2A enhances the permeabilizing effects of AMPinduced membrane pores by facilitating H2A and LL-37 uptake, and by inhibiting repair of AMP-induced pores. We refer to this effect as pore stabilization by H2A. In further support of the porestabilizing effects of H2A, we observed that the dual treatment of H2A and LL-37 caused a dramatic decrease in cell size ( Supplementary Fig. 2G). This result is consistent with a model in which H2A-stabilized pores enable the efflux of cytoplasmic components out of the cell and disrupts the membrane. Furthermore, the result suggests a potential mechanism for the production of cellular debris in the dual-treated SEM images ( Fig. 1h and Supplementary Fig. 1E).
H2A disrupts DNA organization and suppresses transcription. We next investigated the role of H2A subsequent to cellular entry. E. coli were electroporated with H2A and cultured in the continued presence of H2A. We confirmed initial entry of H2A into E. coli via electroporation using fluorescently-labeled AF-H2A ( Supplementary Fig. 3A). The electroporation of H2A into E. coli had a striking inhibitory effect on growth, increasing the lag time of the cultures comparable to those treated with LL-37 alone (Fig. 5a) and decreasing CFUs to nearly undetectable levels on non-selective media following a 1 h treatment (Fig. 5b). The growth inhibition was not due to the process of electroporation or the presence of H2A alone, as these conditions had relatively minor impacts on growth (Fig. 5a, b). The dramatic decrease in CFUs on non-selective media indicate that the effects of H2A, whether through introduction into the cell by electroporation or through co-treatment with LL-37, is bactericidal and that the mechanism is rapid. Together, these results indicate that H2A has a growth-inhibitory effect in the cytoplasm and suggest that H2A affects an intracellular bacterial target.
We note that during the period that followed the electroporation of H2A, the concentration of H2A in the cytoplasm decreased in E. coli (Fig. 5c) despite the fact that additional H2A was present in the extracellular medium. This suggests that the presence of H2A inside the cytoplasm does not induce further membrane pore formation. In contrast, the dual treatment of LL-37 and H2A caused H2A levels to rise within the cytoplasm (Fig. 5c) and produced a striking irrecoverable effect on growth (Fig. 5a, b). This data further supports the hypothesis that H2A stabilizes the membrane pores induced by LL-37. We note that dual LL-37/H2A treatment had more growth impact than previously observed (Fig. 1d), likely due to the low magnesium present during electrocompetent cell preparation.
Although histones bind eukaryotic DNA, their ability to interact with bacterial DNA has not been characterized. We hypothesized that H2A complexes with microbial DNA, perturbing replication and transcription. To measure potential H2Abacterial DNA interactions, we performed non-denaturing polyacrylamide gel electrophoresis of purified E. coli genomic DNA with H2A ( Supplementary Fig. 3B, C). Increasing H2A levels inhibited DNA migration, indicating interactions between H2A and bacterial DNA ( Supplementary Fig. 3B, C). LL-37 exhibited less retention of bacterial DNA for a comparable range of concentrations ( Supplementary Fig. 3D, E).
We then measured the effect of H2A on E. coli chromosomes in live cells using Sytox Green, which fluoresces upon binding to DNA but does not inhibit bacterial growth at low concentrations 48 . Cells were pre-treated with H2A, LL-37, or with a combination of H2A and LL-37 for 3 h. Fluorescence was distributed uniformly in untreated cells, indicating a diffuse bacterial chromosome (Fig. 5d). Treatment using H2A alone produced no change in localization, which was expected since H2A treatment does not induce H2A entry into the cytoplasm (Fig. 2a, b). Treatment using LL-37 also produced no significant change in localization, suggesting that the chromosome is not significantly perturbed by LL-37. The addition of H2A to LL-37treated cells induced a webbed pattern in the chromosome (Fig. 5d), suggesting that H2A entry induces chromosomal reorganization. We performed principal component analysis to identify changes in the chromosomal organization over several hundred cells in each condition. This analysis provides a visualization of how similar or dissimilar chromosomal patterns are for all cells in a single condition. The clustering of cells together along the two principal components (PC1 and PC2) indicates similarities in chromosomal organization whereas those that are distributed further apart indicate dissimilarities in organization. We plotted the principal components for the four treatment conditions as normalized density plots (Fig. 5e). Most cells in untreated, H2A-treated, and LL-37-treated conditions clustered together near a single position. However, treatment using both H2A and LL-37 caused a broad distribution of chromosomal organization patterns. Importantly, this pattern of the chromosomal organization was distinct from that produced through treatment of H2A or LL-37 alone.
To further characterize H2A impact, we visualized HupA localization using a HupA-mRuby2 construct 45 . HupA is a bacterial DNA-binding protein that reports on chromosomal organization. HupA distributions in untreated and H2A-treated cells were largely diffuse and comparable (Fig. 5f). Treatment with LL-37 condensed the fluorescence near the center of the cell (Fig. 5f), distinct from the fluorescence of untreated or H2A-treated cells (Fig. 5g). Treatment with both H2A and LL-37 caused fluorescence condensation and asymmetric localization towards the periphery of the cell (Fig. 5f), a pattern distinct from untreated, H2A-treated, or LL-37-treated cells (Fig. 5g). This result further supports the hypothesis that H2A rearranges the bacterial chromosome.
To characterize H2A's transcriptional effects, we quantified mCherry expression in an E. coli strain containing a tetracycline promoter (activated by anhydrotetracycline) that was transcriptionally fused to a gene encoding mCherry. Similar constructs have measured transcription in other studies 49,50 . We note that inhibition of translation could also affect mCherry fluorescence. Following 1-h pre-treatment with H2A, LL-37, or both, cells were induced for transcription using anhydrotetracycline for 1 h (Fig. 6a). Pre-treatment with H2A or LL-37 had little effect on mCherry fluorescence, indicating the H2A or LL-37 do not Fig. 4 Histone H2A inhibits membrane recovery by stabilizing LL-37-induced pore formation. a Fluorescence and phase-contrast time-lapse images and b CFP fluorescence intensities of E. coli that constitutively express CFP. E. coli were initially treated for 1 h with 10 μg/mL H2A, 1 μM LL-37, or the combination of 10 μg/mL H2A and 1 μM LL-37, and then recovered for a 1-h period without the treatments. Arrows indicate the formation of membrane blebs. c CFP fluorescence intensities of E. coli that constitutively express CFP that were initially treated for 1 h with 10 μg/mL H2A, 10 μM magainin-2 (MAG2), or the combination of 10 μg/mL H2A and 10 μM MAG2, and then recovered for a 1-h period without the treatments. d CFP fluorescence intensities of E. coli that constitutively express CFP that were initially treated for 3 h with 10 μg/mL H2A in media containing low (1 μM) magnesium and physiological (1 mM) magnesium, and then recovered for a 1-h period without the treatments. e Intracellular propidium iodide (PI) fluorescence of E. coli over a 1-h recovery following a 1-h treatment with H2A, LL-37, or the combination of H2A and LL-37. f Intracellular propidium iodide (PI) fluorescence of E. coli over a 1-h recovery following a 1-h treatment with H2A, MAG2, or the combination of H2A and MAG2. Data shown as mean ± SEM and are representative of three independent experiments. For comparison of LL-37 treatments at 0 and 60 min in (b), a one-way ANOVA was performed using LL-37 data only. Two-way ANOVAs were performed for all other comparisons. No adjustments were made for multiple comparisons. p values in e, f are indicated for comparisons between LL-37 and LL-37 + H2A and between MAG2 and MAG2 + H2A, respectively. ns > 0.05. Images are representative of three independent experiments. Scale bars represent 2 μm.
H2A's transcriptional effect was further analyzed through measurements of overall RNA production. A decrease in RNA yield was observed after 30 min of treatment with 10 μg/mL H2A and 1 μM LL-37 (Fig. 6b). Additionally, the total RNA yield was dramatically decreased after 30 min of treatment with 50 or 100 μg/mL H2A alone (Fig. 6c), which are concentrations that partially inhibited growth (Supplementary Fig. 3F). These results To understand the bacterial transcriptional response to H2A, we performed RNA-seq using increasing histone concentrations, instead of using both H2A and AMPs, since the latter condition convolves the effects of both molecules. Further, higher histone concentrations, such as 50 and 100 μg/mL H2A, may occur locally in NETs or upon release from lipid droplets.
H2A upregulated genes belonging to the colonic acid cluster, including wza, wzb, wzc, wcaABCDEFGHIJKL, gmd, and wzx, (Fig. 6d), a 19-gene cluster encoding genes for surface polysaccharide and O-antigen production 51 . Notably, these products are recognized by host cells as hallmarks of bacterial infection and modulate pro-inflammatory responses 18 . Colonic acid also assists in maintaining the membrane potential 52 . Since H2A increases the disruption in PMF (Fig. 3d), this transcriptional response may reflect the cell attempting to repair PMF disruption. The greatest transcriptional changes induced by 10 μg/mL H2A were also the greatest changes induced by 100 μg/mL H2A (Supplementary  Table 1). In particular, wza was upregulated 73-fold in cultures treated with 100 μg/mL H2A for 30 min. The gene cluster is tightly regulated by the rcs phosphorelay system, where transcription of the H-NS-regulated rcsA regulator increased 22-fold due to H2A treatment. We validated rcsA upregulation using a transcriptional coli that were untreated, electroporated in the absence of any treatment (empty electroporation), treated with 10 μg/mL AF-H2A, 2 μM LL-37, or both, or electroporated with 10 μg/mL AF-H2A and cultured with the same concentration of AF-H2A. Cells were cultured in minimal medium containing 1 mM magnesium. AF-H2A was mixed with unlabeled H2A (1% AF-H2A, combined concentration of 10 μg/mL) (n = 4 for each condition). b CFUs per microliter of cultures that were treated for 1 h under identical conditions as (a) and plated on non-selective LB plates (n = 4 for each condition). c AF-H2A fluorescence intensities of E. coli that were cultured under identical conditions as (a) (n = 3 for each condition). d Representative phase contrast, SYTOX fluorescence, and merged images of E. coli that were untreated or treated with 10 μg/mL H2A, 2 μM LL-37, or both. The fluorescence images are displayed using the full range of pixel values in the images. e Corresponding principal component analysis (PCA) of images of SYTOX-stained E. coli (n = 3 for each condition). Each cell is represented as a point on a PCA plot, which is then transformed into a density plot. The color scales indicate normalized cell densities. f Representative phase contrast, HupA-mRuby2 fluorescence, and merged images of E. coli that express HupA-mRuby2 that were untreated or treated with 10 μg/mL H2A, 2 μM LL-37, or both. g Corresponding PCA analysis of E. coli expressing HupA-mRuby2 that has been transformed to density plots (n = 3 for each condition). For imaging, cells were immobilized for 3 h on agarose pads containing 2 μM LL-37, 10 μg/mL H2A, or both. For SYTOX analysis, pads additionally contained 5 μM SYTOX. Bars and points are shown as mean ± SEM of biologically-independent experiments. One-way ANOVAs were performed in (b) and a two-way ANOVA was performed in (c) in which the 30-min data were excluded the analysis. See Supplementary information for raw statistical data. Images are representative of three independent experiments. Scale bars represent 2 μm. c 10   fusion of YFP to the rcsA promoter in cells co-expressing CFP under control of the ompA promoter ( Supplementary Fig. 4A). Importantly, cells that were membrane-permeabilized (PI-positive) exhibited a significant increase in rcsA-yfp expression and decrease in ompA-cfp expression ( Supplementary Fig. 4B). Thus, H2A induces selective upregulation of membrane biogenesis components while globally decreasing transcription. An rcsA mutant showed increased sensitivity to treatment with H2A and LL-37 ( Supplementary Fig. 4C), indicating its key role in promoting survival in response to H2A.
Histone-AMP synergy due to a positive feedback loop. Together, histones and AMPs constitute a positive feedback loop: histone entry into bacteria facilitates the uptake of AMPs, which further increases histone uptake (Fig. 7a). Feedback loops provide exponential amplification of small signals. We modeled AMP/ histone dynamics using first order differential equations, encoding passive diffusion of AMPs and histones into the cytoplasm, the increase of AMP entry by histones, and the increase of histone entry by AMPs (full model details are described in the Methods). In our simulations, we observed that feedback between AMPs and histones exponentially amplifies the uptake of both ( Fig. 7b and Supplementary Fig. 5A). Thus, low concentrations of histones and AMPs together can trigger an exponential uptake of both, effectively lowering the MIC of these molecules and resulting in rapid bacterial killing. This is consistent with the all-or-none bimodal membrane permeabilization phenotype that we observed among dual-treated populations ( Supplementary Fig. 5B). If interactions between the two lacked positive feedback, cells would take up far less of each ( Fig. 7b and Supplementary Fig. 5A), and a continuous distribution of membrane permeabilization would instead be observed. Importantly, while bacteria can respond to a dual histone/AMP attack by increasing the expression of outermembrane repairing machinery, this response does not defeat the exponential nature of the feedback loop and only serves to elevate the threshold concentration required to activate it. The combination of histones, or histone fragments, with a pore-forming agent could provide a new strategy to kill bacteria. We investigated possible synergy between H2A and the cationic antibiotic polymyxin B (PMB) 53 , a pore-forming antibiotic typically used as a last-resort drug. Recent reports show that polymyxin-resistant strains of E. coli have emerged 54 . PMB permeabilizes the bacterial membrane and enables uptake of the peptide itself 24 , a mechanism similar to LL-37. Treatment of E. coli with 1 μg/mL PMB slightly inhibited growth (Fig. 7c). Based on our model, supplementing PMB treatment with H2A should lead to synergistic killing. Indeed, concurrent treatment of H2A and PMB completely eradicated bacterial growth (Fig. 7c) and was more effective at inhibiting growth than kanamycin or chloramphenicol at near-MIC or above-MIC concentrations using otherwise identical growth conditions ( Fig. 1f and Supplementary Fig. 1D). Our model thus demonstrates that incorporating a natural defense strategy can have a potent impact on antibiotic efficacy. Identifying other synergies in natural host defenses is likely to yield important insights that can be incorporated into future antimicrobial designs.

Discussion
The histone antimicrobial mechanism has remained elusive for decades. The work here discovers that histone H2A kills bacteria in conjunction with AMPs by inducing depolarization of the membrane potential, enhancing the effects of pore formation by AMPs, reorganizing bacterial chromosomal DNA, and repressing transcription. Importantly, the main activity of H2A is not observed in physiological environments unless a membrane poreforming agent is present. These findings place into context the previous findings of limited histone activity in physiological conditions [19][20][21][22][23][24][25] and demonstrate that the antimicrobial effects of histones are unmasked when histones combine with other actors. While AMP-mediated pore formation has significant bactericidal activity, our results show that bacteria can largely recover from these effects. However, when both H2A and AMPs are present, killing is synergistic and irrecoverable.
Innate immune responses require concerted action among multiple components; the AMPs/histones activity described here represents such a mechanism. The role of AMPs in immune responses would thus appear to be the formation of membrane pores that facilitate the entry of other antimicrobial molecules such as histones into bacteria. Once inside, such molecules may target different bacterial growth mechanisms. AMPs and histones thus function as two components of a multi-step innate immunity antimicrobial mechanism.
H2A has antimicrobial activity at the membrane and within the cytoplasm. At the membrane, H2A enhances LL-37-induced pores, increasing LL-37 and H2A uptake, inhibiting repair of AMP-induced membrane pores, increasing destruction of the NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-17699-z ARTICLE gradient required for ATP production, and facilitating the release of cellular contents. Repair inhibition was not observed in membranes that were weakened through growth at low magnesium concentrations, which suggests that pore stabilization by H2A is specific to AMP-induced pores. H2A could enhance the AMP-induced pores through two mechanisms: impeding pore repair by making AMP membrane removal more difficult, or by increasing membrane tension, proposed as a mechanism of action by some AMPs 55 , which would facilitate creation of AMPinduced pores and increase the difficulty of closing them. Inside, H2A targets multiple processes providing an additional level of assault. H2A's transcriptional inhibition diminishes the bacterial response, preventing repair of cell damage caused by AMP-induced pores. However, H2A's antimicrobial activity is not fully dependent on AMP pore formation, as both growth with H2A in low magnesium conditions and the electroporation of H2A into the cytoplasm inhibited growth and were bactericidal, as judged by decreases in CFUs. We thus propose transcriptional inhibition as a potential third mechanism by which H2A stabilizes pores. We note that the electroporation of H2A inhibited growth to a less extent than the combined treatment of using both H2A and LL-37. The greater inhibition in the latter case may be attributed to the persistent pore formation by LL-37, as opposed to the transient pores formed during electroporation. Future work is required to evaluate the relative impacts of H2A acting at the membrane and within the cytoplasm on bacterial death.
H2A and LL-37 constitute a self-amplifying mechanism that significantly lowers the effective minimum inhibitory concentration of both molecules. As H2A shares significant structural and chemical similarities with other histones 33 and LL-37 shares similarities with other cathelicin-derived AMPs and defensin AMPs 32 , the results and model suggest potential synergy between other histones and AMPs. In support of this, synergy was observed between histone H3 and LL-37, and between H2A and magainin-2. In certain environments, such as a lesion or NETs, proteases may cleave histones, producing histone-derived peptides 56 , which could also synergize with AMPs. Importantly, as AMP antibacterial activity can be modulated, peptides may be designed in the future to increase membrane permeabilization, translocation, or synergy with other AMPs 57,58 . In principle, a single molecule that has both AMP-like and histone-like properties, such that it both induces pores in membranes and inhibits transcription, could have a self-amplifying effect by itself.
Given histone toxicity to the host [15][16][17][18] , the therapeutic potential of histones and AMPs should be considered in the context of natural host defenses, including NETs and lipid droplets. The binding of histones and AMPs to NETs may prevent generalized histone spread and avoid off-target effects. Further, histone localization to lipid droplets for targeted release to kill bacteria 20 could provide an effective delivery mechanism while limiting off-target effects.
The ubiquitous co-occurrence of histones and AMPs in the immune system suggests that this antimicrobial mechanism is present in a wide range of cell types. In addition to the formation of NETs by neutrophils, histone-and AMP-rich extracellular traps form in macrophages (METs) and in dendritic cells 59,60 . Extracellular traps (ETs) have been observed in other immune cells, including mast cells and eosinophils, suggesting a role for histones and AMPs in antimicrobial activity in these cells 61 . Recent reports also demonstrate that different types of NETs are established through distinct citrullination-dependent or citrullination-independent pathways 62 . Our data suggest that the citrullinated form of histone H3 has less antimicrobial activity. The level of histone citrullination could represent a mechanism for the cell to tune the level of antimicrobial activity while balancing autoimmune activation. Future work will need to investigate the impact of different NET formation pathways on antimicrobial activity and autoimmunity and will need to investigate the general co-occurrence of histones and pore-forming agents within and beyond the innate immunity system, where they may function effectively as a two-component antimicrobial mechanism.
Bacterial strains. Experiments were performed using the E. coli strain MG1655 (seq) 64 , which is devoid of the bacteriophage lambda and F plasmid, and the S. aureus strain RN4220, which was originally derived from NCTC8325-4 was used for experiments involving Gram-positive bacteria 65 . The proton gradient was measured using pJMK001 in the E. coli strain XL1 Blue (Addagene, Watertown, MA), which expresses the PROPS protein under the control of the arabinose promoter 46  Agar plate assay. To quantify the effects of histone treatment in low ionic conditions, overnight cultures of stationary phase E. coli or S. aureus were diluted 1:1000 into the minimal medium containing 1 μM or 1 mM MgSO 4 and cultured with or without 10 μg/mL histone H2A. Bacteria were cultured for 1 h at 37°C in a shaking incubator at 225 rpm. Bacterial suspensions were diluted 1:1000 into fresh minimal medium with either 1 μM or 1 mM MgSO 4  . See 'Code Availability' section below for code. After treating bacteria with antimicrobial agents, 5 µl of culture was plated on 1% agarose-minimal medium pads and imaged immediately, which is described in ref. 69 . A minimum of 100 cells were imaged and analyzed in each experiment.
Propidium iodide staining. To visualize membrane permeability of stationary phase E. coli in low and physiological magnesium concentrations, overnight cultures of MG1655 were grown overnight to saturation and diluted 1:1000 into minimal medium with 1 μM or 1 mM MgSO 4 , with or without 10 μg/mL H2A. 30 μM propidium iodide was co-incubated with the solution of bacteria for 1 h at 37°C in a shaking incubator at 225 rpm before plating on 1% agarose-minimal medium pads. Data were collected using the mCherry filter. To visualize membrane lysis of mid-exponential-phase E. coli, overnight cultures of MG1655 were grown overnight to saturation, sub-cultured in fresh minimal medium containing 1 mM MgSO 4 and grown to an OD 600 of 0.2. Exponential-phase bacteria were diluted 1:20 into fresh minimal medium with 1 μM or 1 mM MgSO 4 , and supplemented with H2A, LL-37, magainin-2, kanamycin, chloramphenicol, or polymyxin B. Bacteria were cultured at least 1 h at 37°C in a shaking incubator at 225 rpm before plating on 1% agarose-minimal medium pads. 30 μM propidium iodide was co-incubated with the solution of bacteria for at least 15 min before imaging. Data was collected using the mCherry filter.
PROPS fluorescence analysis. The E. coli strain XL-1 Blue containing the PROPS plasmid pJMK001 were grown in LB in a shaking incubator at 33°C, induced with arabinose and 5 μM retinal, and incubated in darkness for 3.5 h. The culture was spun down and resuspended in M9 minimal medium 46 . E. coli were back-diluted into fresh MinA minimal medium, cultured to an OD 600 of 0.2, diluted 1:20 into fresh MinA minimal medium, treated with 10 μg/mL H2A, 1 μM LL-37, both H2A and LL-37, 1 μg/mL PMB, or both H2A and PMB, and incubated for 1 h. Cells were immobilized on a 1% agarose pad and imaged using a Cy5 filter.
Electroporation of E. coli with H2A. Electrocompetent MG1655 were prepared by culturing in SOB to an OD 600 of 0.2 to 0.5, washing with 10% chilled glycerol 4 times, resuspending to an OD 600 of 0.2, and freezing at −80°C. For electroporation, 10 µg/ml of H2A or the equivelant volume of water was added to 50 µl of electrocompetent E. coli, transferred to a 1 mm electroporation cuvette, and shocked using the "Ec1" setting on a Bio-Rad Micropulser (Bio-Rad, Hercules, CA). Cells were resuspended in cold MinA minimal medium with 1 mM MgSO 4 in a final volume of 1 mL containing 10 µg/ml of H2A, 2 µM of LL-37, or both H2A and LL-37, and cultured at 37°C. To count CFUs, cultures were diluted serially using minimal medium containing 1 mM MgSO 4 and 25 μL of the dilutions were plated on non-selective LB-Miller agar plates. Plates were incubated for 18 h at 37°C and assessed for CFUs by counting the number of colonies present.
Timelapse of E. coli recovery. To quantify the time-course of recovery in E. coli treated with H2A alone, LL-37 alone, or the synergistic combination of LL-37 and H2A, MAL204 was cultured to mid-exponential phase in MinA minimal medium, treated with 10 μg/mL H2A, 1 μM LL-37, or 1 μM LL-37 with 10 μg/mL H2A, and incubated for 1 h. The solution was filtered through a 0.22 μm filter to remove excess LL-37 and H2A and cells were resuspended in fresh minimal medium. Cells were immobilized on a 1% agarose pad and imaged over an hour time period.
Time course of membrane healing. To quantify the time-course of membrane repair in bacteria treated with H2A alone, AMPs alone, or the synergistic combination of AMPs and H2A, MAL204 was grown to mid-exponential phase in MinA minimal medium, diluted 1:20 with 10 μg/mL H2A, 1 μM LL-37, 1 μM LL-37 with 10 μg/mL H2A, 10 μM MAG2, or 10 μM MAG2 with 10 μg/mL H2A, and incubated for 1 h. The solution was filtered through a 0.22 μm filter to remove excess LL-37 and H2A and cells were resuspended in fresh minimal medium. Cells were allowed to recover for 0, 30, and 60 min before the addition of 30 μM propidium iodide for 15 min prior to performing fluorescence microscopy. Intracellular propidium iodide fluorescence and CFP fluorescence were quantified. To quantify the time-course of membrane repair in bacteria treated with H2A in low and physiological environments, MAL204 was grown to mid-exponential phase, diluted 1:20 into minimal media with 1 μM or 1 mM MgSO 4 , with or without 10 μg/mL H2A, and incubated for 3 h. The solution was filtered through a 0.22 μm filter to remove excess H2A and cells were resuspended in fresh minimal medium. Cells were allowed to recover for up to 60 min before the addition of 30 μM propidium iodide for 15 min prior to performing fluorescence microscopy. Intracellular propidium iodide fluorescence and CFP fluorescence were quantified.
Cell aggregate size and cell size analysis. MG1655 were cultured to an OD 600 of 0.2, diluted 1:20 into fresh MinA minimal medium, treated with 0-4 μM LL-37 or 0-100 μg/mL H2A, and incubated for 1 h. Cells were immobilized on an agarose pad, imaged using phase contrast microscopy, and analyzed using our own customwritten image analysis tools 67,68 version 1.1 that was written in in Matlab (Version R2017b; Mathworks, Natick, MA). See the "Code availability" section below to download code. The total pixel area of each individual cell was determined by computing the mask area and converting from pixels to μm 2 by multiplying the mask area by a factor of 0.00422 μm 2 /pixel to account for the microscope camera pixel size and objective magnification.
Chromosomal analysis using SYTOX and HupA-mRuby2. MG1655 or XL-1 Blue expressing HupA-mRuby2 was cultured to OD 600 of 0.2, diluted 1:20 into fresh MinA minimal medium, and treated with 2 μM LL-37, 10 μg/mL H2A or both LL-37 and H2A for 30 min. For SYTOX visualization, MG1655 were stained with 3 μM SYTOX Green nucleic acid stain (ThermoFisher, Waltham, MA) for 10 min. Cells were immobilized on agarose pads containing 2 μM LL-37, 10 μg/mL H2A, or both LL-37 and H2A, and remained on the pad for 3 h before imaging using the fluorescence microscopy. For SYTOX analysis, pads additionally contained 5 μM SYTOX Green. Raw images were analyzed through principal component analysis using our own custom-written image analysis tools 67,68 version 1.1 and modified in Matlab (Version R2017b; Mathworks, Natick, MA). See 'Code Availability' section below for code. Individual cells were identified in phasecontrast images using canny edge detection and using SuperSegger 70 . Images of LL-37-treated cells and of cells treated with both LL-37 and H2A were pooled together, rotated such that the major axis of the cell was parallel to the x-axis and resized to 30 × 100 pixels. The covariances between corresponding pixels of different cells were computed using the 16 bit intensity values from the rotated and resized fluorescence images and for the same images rotated by an additional 180 degrees. The orientation that gave the lower covariance was used for the analysis. Principal components for the covariance matrix were computed using~400 cells and the principal components that gave the two largest eigenvalues were plotted. Density plots were created by binning points in principal component space in a 15 × 15 bivariate histogram plot. The size of each bin was determined by subtracting the minimum principal component score from the maximum principal component score and dividing that by the number of bins along that principal component. Histogram bins were normalized as the fraction of the total cell population.
Bacterial DNA purification. Overnight MG1655 cultures were grown to saturation in MinA minimal media. DNA purification was performed using a Miniprep kit (Qiagen, Germantown, MD). A three-second sonication step was performed after lysis to isolate genomic DNA.
Non-denaturing nucleic acid PAGE. 10 μL mixtures containing 1 μg purified DNA from MG1655 were incubated with 0-1.4 μg Histone H2A or LL-37 for 25 min at 25°C. Gel loading sample buffer (5×, Bio-Rad, Hercules, CA) was added to a final concentration of 1× and the products were separated by native PAGE on a 5% TBE gel (Bio-Rad, Hercules, CA) at 100 V for 60 min. The gel was stained with 1X SYBR safe (Invitrogen, Carlsbad, CA) in TBE buffer 71 for 30 min before visualization using a EOS Rebel T5 DSLR camera with an f/3.5-5.6 18-55 mm lens (Canon, Huntington, NY) and a DR46B Transilluminator (Clare Chemical, Dolores, CO).
In vivo transcription assay. To determine how histone entry into the bacterial cell impacts transcription, MAL190 was cultured to mid-exponential phase, diluted 1:20 in MinA minimal medium, treated with 10 μg/mL Histone H2A and/or 2 μM LL-37, incubated for 1 h, and induced for transcription using 50 ng/mL of anhydrotetracycline. The fluorescence of mCherry was measured after 1 h using fluorescence microscopy.
RNAseq. MG1655 were cultured to saturation overnight in MinA minimal medium, diluted 1:1000 into the same medium, cultured to an OD 600 of 0.2, diluted 1:20 in pre-warmed medium, and supplemented with 10 μg/mL H2A, 1 μM LL-37, or both. 10 mL of culture was harvested at 0, 30, and 60 min, filtered through a 0.8 μm filter, washed with 2 mL H 2 O, and resuspended in 600 μL Total Lysis Solution (TE 8.0 (10 mM Tris-HCl, 1 mM EDTA), 0.5 mg/mL lysozyme (Sigma), and 1% SDS). Samples were incubated for 3 min at room temperature before snap freezing in liquid N 2 . Samples were kept in −80°C until nucleic acid extraction with a hot phenol-chloroform extraction and ethanol precipitation 67 . RNA yield was measured using a Nanodrop 2000 (Thermo Fisher, Waltham, MA). Samples were digested with DNase (Ambion, Waltham, MA) and treated with RiboZero (Illumina, San Diego, CA). A NEBNext Ultra Directional Library kit (NEB, Ipswich, MA) was used to construct a cDNA library, which was sequenced by the Princeton University Genomics Core Facility with a depth of at least 10 M read per experimental condition. Sequencing data were analyzed using our own software written in Python version 2.7.16 and using R version 3.4.3 (The R Foundation, Vienna, Austria). Sequences were aligned to the MG1655 genome (U00096.3) using Bowtie2 72 version 2.2.4.
Transcripton of rcsA. MAL204, which contains YFP fused to the promoter of rcsA and constitutively expresses a transcriptional fusion of CFP to ompA was grown to mid-exponential phase. Bacteria were diluted into warmed MinA minimal medium with increasing concentrations of H2A. In addition, 30 μM PI was added to the culture to specifically measure fluorescence intensities in membrane-permeabilized cells. After a 30-minute incubation period, cells were immobilized on a 1% agarose pad and YFP, CFP, and PI fluorescence values were analyzed using fluorescence microscopy.
Statistical analysis. Statistical analysis was performed by running Welsh t-tests or ANOVA and Tukey's post-hoc tests using R 3.4.3 (Kite Eating Tree), Image J (v1.51k), Microsoft Excel version 16.36, or our own custom-written MATLAB scripts version 1.1. See the "Code availability" section below for code.
Histone-AMP positive feedback model. We developed a mathematical model to describe the dynamics of histone and AMP uptake into bacterial cells. Histones and AMPs enter passively using simple diffusion: are the rate constants associated with the passive entry of histones and AMPs into the cell, respectively. Molecules of histones and AMPs can leave the cell through a number of ways including cell division, shedding of cell components, and transport through drug efflux pumps. We describe these combined effects on histones and AMPs using the rate constants k Hisexit and k AMPexit , respectively. To encode the behaviors that histones increase the intracellular AMP concentration and that AMPs increase intracellular histone concentrations, potentially through pore-stabilization, we defined the rate constants k Hisstab and k AMPstab , arriving at the equations: d½His in dt ¼ k Hisentry His out ½ Àk Hisexit His in ½ þk Hisstab AMP in ½ d½AMP in dt ¼ k AMPentry AMP out ½ Àk AMPexit AMP in ½ þk AMPstab His in ½ In our simulations, we set the initial histones and AMP concentrations inside the cell to 0. The concentration of histones and AMPs outside of the cell remained constant, which describes an environment in which there is an excess of histones and AMPs. We set the permeation rates of k Hisentry and k AMPentry to 0.004 s −1 based on permeation measurements of the charged antibiotic tetracycline into bacterial cells 73 . The rate constants k Hisexit and k AMPexit were set to correspond to a doubling time of 30 min, which is a conservative estimate of the rate of histone and AMP removal from the cell that does not require the existence of an export mechanism. We simulated the synergy condition by setting k Hisstab and k AMPstab to 0.1 s −1 and simulated the non-synergistic condition by setting these rate constants to 0 s −1 . For the uptake dynamics figure, we set the concentrations of histones and AMP outside of the cell to 1 and computed the total intracellular concentration of these molecules as a function of time. Density plots were constructed by computing the total intracellular concentration of histones and AMPs following 60 min of exposure to a range of histones and AMPs concentrations outside of the cell.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
Raw data for all figures that contain ANOVA or t-test analyses are available in the Supplementary Information Source Data file. The RNA-Seq data is freely available under the National Center for Biotechnology Information Gene Expression Omnibus accession number GSE142755. Additional raw data that support the findings of this study are available from the corresponding authors upon request. Source data are provided with this paper.