Lysis‐lysogeny coexistence: prophage integration during lytic development

Abstract The infection of Escherichia coli cells by bacteriophage lambda results in bifurcated means of propagation, where the phage decides between the lytic and lysogenic pathways. Although traditionally thought to be mutually exclusive, increasing evidence suggests that this lysis‐lysogeny decision is more complex than once believed, but exploring its intricacies requires an improved resolution of study. Here, with a newly developed fluorescent reporter system labeling single phage and E. coli DNAs, these two distinct pathways can be visualized by following the DNA movements in vivo. Surprisingly, we frequently observed an interesting “lyso‐lysis” phenomenon in lytic cells, where phage integrates its DNA into the host, a characteristic event of the lysogenic pathway, followed by cell lysis. Furthermore, the frequency of lyso‐lysis increases with the number of infecting phages, and specifically, with CII activity. Moreover, in lytic cells, the integration site attB on the E. coli genome migrates toward the polar region over time, leading to more spatial overlap with the phage DNA and frequent colocalization/collision of attB and phage DNA, possibly contributing to a higher chance for DNA integration.

allowing it to react to and thrive in variable conditions, to maximize its own fitness.
The protein players involved in this cellular decision-making process have been well-characterized over decades (Court et al., 2007;Dodd et al., 2005;Oppenheim et al., 2005), and CII, Cro and and Q are among the key proteins that determine the infection outcome, mediating either the lysogenic or lytic pathways . Cro facilitates the lytic pathway by being a weak repressor for phage gene expression from both pL and pR promoters Kobiler et al., 2005;Svenningsen et al., 2005;Takeda et al., 1977), while Q activates the lytic pathway after reaching a threshold, allowing for the expression of a single transcript carrying the lysis and morphogenesis genes Marr et al., 2001).
Conversely, CII activation will inhibit the lytic pathway and establish the lysogenic pathway by activating transcription from three specific promoters Oppenheim et al., 2005). Among them, the pI promoter allows the expression of the lambda integrase, Int, which catalyzes the crucial process of integrating phage DNA into the host chromosome (Landy, 1989;Nash, 1981).
New details have emerged from higher-resolution studies of this well-established system (St-Pierre & Endy, 2008;Van Valen et al., 2012;Zeng et al., 2010). Our recent study performed at the single-cell level proposed that individual phages infecting the same cell are able to "vote" for the cell's fate independently (Zeng et al., 2010), which raised the possibilities that lytic and lysogenic pathways can happen simultaneously within the same cell, resulting from the different votes by multiple infecting phages. This coexisting lytic-lysogenic development may be naturally beneficial, serving as an intermediate state allowing for a faster and more sensitive commitment to lysis-lysogeny in a changing environment. Exploring this phenomenon requires a higher resolution of study and can yield insights into the biological process of decision-making and its evolutionary strategy.
In this study, we developed an improved reporter system at the single-DNA level to allow the visualization of phage DNA integration, in addition to the progress of the lytic and lysogenic pathways. By tracking phage and host DNA movements after infection in real-time using fluorescence microscopy, and quantitatively analyzing singlemolecule trajectories, we reveal a new biological phenomen on of "lyso-lysis" and gain further insights into the possible mechanism of cellular decision-making.

| Bacterialstrains
Bacterial strain LZ722 was constructed by inserting a DNA array containing ~200 tetO repeats into strain LZ220 (Shao et al., 2015) at ~1,500 bp upstream of attB site using lambda red recombination (Datsenko & Wanner, 2000). Plasmid pFtsKi-tetR-mCherry, which contains the tetR-mCherry under the constitutive promoter FtsKi was transformed into LZ722, resulting in LZ731. For all real-time microscopy experiments, LZ731 is used as the host, while for bulk assays (lysogenization, PCR and qPCR), E. coli strain MG1655 is used. Unless otherwise specified, phage titration assays for determining the phage concentration was done with E. coli strain LE392.

| Plasmidconstruction
To construct the plasmid pLZ729: pFtsKi-tetR-mCherry, plasmid pWX510 (Wang et al., 2014) was digested with HindIII and BamHI restriction enzymes to obtain sequences for tetR-mCherry, which was then inserted into pBR322. DNA sequences for FtsKi was PCR amplified from pWX6 (Wang et al., 2005) using primers QS15 and QS16 and inserted into the above plasmid between EcoRI and HindIII recognition sites, resulting in EcoRI-FtsKi-HindIII-tetR-mCherry-BamHI in the pBR322 backbone. When this plasmid was transformed into LZ722, the background signal (mCherry) was found to be too high, therefore we switched to another vector, pACYC177, which has a lower copy number. The piece FtsKi-HindIII-tetR-mCherry was PCR amplified using primers QS17 and QS18 and inserted into pACYC177 between SmaI and NheI, resulting pLZ729. The plasmid pZA32-dam carries the dam gene in between AvrII and KpnI in the pZA32 backbone, where the dam gene was amplified with primers QS19 and QS20, using template plasmid pGG503 (Herman & Modrich, 1981).
When phages were produced from dam + host cells containing this plasmid, pZA32-dam, the phage DNA was confirmed to be fully methylated (Fig. S1D).

| Phagestrains
The phage λ D-mTurquoise2 cI 857 bor::Kan R was obtained through recombination by infecting λ Dam cI 857 bor::Kan R on the permissive strain LE392-bearing plasmids pBR322-D-mTurquoise2-E. The recombinant (λ D-mTurquoise2 cI 857 bor::Kan R ) was selected based on its ability to titer on nonpermissive strain MG1655 and fluoresce under a fluorescence dissecting microscope. For easier selecting and counting of lysogens for λint − in the lysogenization assays, λint − -Kan was constructed following the protocol as described in (Shao et al., 2015) to replace the nonessential bor gene region of λint − with a Kan R cassette.

| Phagelysatepreparation
Fully methylated mosaic phage λWT-FP was obtained by inducing a lysogen with temperature-sensitive prophage (λ D-mTurquoise2 cI 857 bor::Kan R ) and two plasmids, plasmid pPLate-D to provide wild-type phage decorative capsid protein gpD (Zeng et al., 2010) and plasmid pZA32-dam which over produces Dam methylase after 1 mmol/L IPTG induction. Fully methylated phages λWT, λint − , λcII 68 , and λcII stable were obtained by infecting host cell LE392 carrying plasmid pZA32-dam with the corresponding phages at 42°C. This is important if the phage lysate will be used for quantifying the lyso-lysis using qPCR. We found that the phage lysate obtained through prophage induction contains nonnegligible amount of integrated phage DNA, possibly due to insufficient induction, while the phage lysate obtained through infecting the host cells contains no integrated DNA. All phage lysates used were also purified following the protocols described in (Zeng & Golding, 2011

| Bulklysogenizationassay
To measure the lysogenization frequency of the various phages, we followed the protocol as described in (Zeng et al., 2010). For easier selection and counting of lysogens, the phage λint − -Kan was used instead of λint − since bor gene was reported to be nonessential and would not affect the lysogenization frequency (Barondess & Beckwith, 1995). All the other phages used also carried an antibiotic marker by replacing the λ bor region. Briefly, 2 ml of the host cell MG1655 was grown in LBMM for overnight and subsequently diluted 1:1,000 into 12 ml of LBMM and grown to OD 600 ~ 0.4 at 37°C, centrifuged (2,000g for 10 min at 4°C) and resuspended to be ~1.5 × 10 9 cells ml −1 in prechilled LBMM (LB + 0.2% maltose + 10 mmol/L MgSO 4 ).
Thereafter, 20 μl of the resuspended cells were then infected with 20 μl of phages at different concentrations by incubation for 30 min on ice. The samples were then transferred to 35°C water bath for 5 min to allow for phage DNA ejection, followed by 10-fold dilution into prewarmed LBGM (LB + 0.2% glucose + 10 mmol/L MgSO 4 ) and incubation with shaking at 265 rpm at 30°C for 45 min. The samples were then properly diluted and plated on LB plates containing appropriate antibiotics to allow ~100 colonies on each plate.

| PCRandqPCR
Here 2 ml of host cell MG1655 was grown in LBMM overnight and was subsequently diluted 1:1,000 into 100 ml of LBMM and grown to OD 600 ~ 0.4 at 37°C. Cells were then spun down at 2,000g for 10 min at 4°C and resuspended to be ~1.5 × 10 9 cells ml −1 in prechilled LBMM. Infection was set up following the same protocol described in Bulk lysogenization assay, with corresponding phages at different concentrations for infections of different APIs, but with larger volumes depending on the number of samples to be taken later (100 μl of reaction per sample). For each time point, a quantity of 100 μl of the reaction was added to 0.9 ml prewarmed LBGM shaking at 265 rpm in 30°C shaker for various times up to 120 min.
For confirming and quantifying lyso-lysis, samples were taken at each time point and immediately filtered using 0.2 μm membrane to obtain cell-free samples. For the infection with different APIs, the samples taken at 90 min were used, and samples were diluted 10fold into dH 2 O to minimize possible PCR inhibitor effects. PCR or qPCR was performed immediately after the last sample was taken.
PCR was done using primers in (Powell et al., 1994), while qPCR was done using primers QS1 and QS2 for detecting E. coli DNA, and primers QS3 and QS4 for detecting integration (Table S2). Amplification was done using SYBR Green PCR master mix (Applied Biosystems, 4309155) with 250 nmol/L of each primer. For determining the mRNA level of int/xis/cII, infection was done following the same protocol, but with 5× volumes for each sample. Samples were taken out at different time points: 0, 6, 12, 18, 24, 30, and 40 min, and immediately poured into 5 ml ice-cold methanol. Samples were then spun down at 4,000g for 4 min, at 4°C. The cell pellet was resuspended in 1 ml of RNAprotect Bacteria Reagent (Qiagen, 76506), followed by incubation for 5 min at room temperature. Then the cells were spun down at 5,000g for 5 min at room temperature. After discarding the supernatant, the cells were kept at −20°C until all samples were collected. RNA extraction was done using RNeasy Mini Kit (Qiagen, 74104), followed by DNA digestion with TURBO DNA-free kit (ambion, AM1907) for a total of 80 min and reverse transcription using High Capacity RNA-to-cDNA kit (Applied Biosystems, 4387406). The obtained cDNA was then quantified, using SYBR Green PCR master mix. Primers QS7 and QS8 were used for quantifying cII, primers QS9 and QS10 for int and primers QS11 and QS12 for xis, while ihfB is used as a reference gene using primers QS5 and QS6 (Table S2).

| Quantifyingpercentageofmultipleprophage integration
Infection was set up as described in Bulk lysogenization assay, with the infecting phages being λWT at API of 0.1, 1 and 10. After obtaining the lysogens on the plates, 96 colonies of each infection were used to determine whether they contain single or multiple phage integration by PCR following protocols as described in (Powell et al., 1994).
The percentage of cells having multiple prophage integration is then calculated based on the PCR results.
Here 60 μl of the culture was subsequently diluted 1:100 into 6 ml M9M and grown to OD 600 ~ 0.4. 1 ml of cells were then collected by centrifugation at 2,000g for 2 min at room temperature, and resuspended in 40 μl of M9M. Thereafter, 20 μl of phage lysate was then added to 20 μl of cells to reach an API of 0.5-5, followed by incubation for 30 min on ice and another 5 min at 35°C water bath to allow DNA ejection. The sample was then diluted into M9M at room temperature by 10-fold. 1 μl of the diluted sample was used for imaging following protocols as described in (Shao et al., 2015) with 1.5% M9M agarose pad. Imaging was performed on an inverted microscope (Ti-E, Nikon, Tokyo, Japan) with a cage incubator (InVivo Scientific, St. Louis, MO) set at 30°C. Images were taken using 100× objective (Plan Fluo, NA 1.40, oil immersion) with standard filter sets and a cooled EMCCD camera (iXon 3 897, Andor, Belfast, United Kingdom). When needed, a series of 9 z-stack images with spacing of 300 nm in the CFP channel (200 ms exposure) was taken to capture all infecting phages in the initial frames, after which images were taken every 5 min through the phase-contrast, YFP, mCherry, and CFP channels (100, 200, 50 and 100 ms exposure respectively) at the focal plane to allow tracking of DNA movement and cell fate in the time-lapse movies.
Briefly, cells were first outlined using MicrobeTracker, after which spots were recognized first automatically using SpotFinderZ, then manually corrected using SpotFinderM (Sliusarenko et al., 2011). Cell lineage tracking and the calculation of minimum distance between attB and lambda DNA, Dis(λ-attB), for each cell was done using custom Matlab script in our lab. The Dis(λ-attB) is calculated as the minimum distance between all possible pairs of lambda DNA and attB in each given cell at each given time point, where the distance between lambda DNA and attB was calculated as:

| ReportersystemforphageDNAintegration: E. coliattBandphageDNAlabeling
In the lysogenic pathway, lambda DNA is integrated into the E. coli genome at the attB site through recombination by the phage-encoded integrase, Int, in the presence of the host factor IHF (Nash, 1981). To visualize this integration event, we developed a reporter system to simultaneously track the E. coli attB and the phage DNA. Specifically, the host cell LZ731 contains about 200 repeats of tetO (Lau et al., 2003) inserted upstream of attB on the chromosome (Fig. 1A, left) and a plasmid pFtsKi-tetR-mCherry, which constitutively expresses TetR-mCherry (Wang et al., 2005), therefore the tetO repeats are bound by TetR-mCherry, resulting in a distinct spot ( Fig  indicating the attB location. The phage DNA is labeled using our previously reported method (Shao et al., 2015) (also see Fig. 1A, right).
Briefly, the phage λWT-FP was produced in a host with enhanced Dam methylase activity resulting in fully methylated phage DNA packaged in its head (see Experimental Procedures). The host cell LZ731 also constitutively expresses a fluorescent SeqA chimera, SeqA-YFP (Babic et al., 2008) from the chromosome, and the host DNA is not methylated owing to a dam − mutation (methylation deficient). SeqA specifically binds to fully methylated and hemi-methylated DNA, so the phage DNA appears as a YFP spot ( Fig. 1A-C, yellow dots) once ejected into the cell. The phage DNA reporter system allows tracking of the first two copies of each initial DNA after replication, since there are only two methylated strands of DNA. For example, in Figure 1B, the yellow focus splits into two at 60 min, and no new foci appear despite continued DNA replication. Cells with more than two foci, e.g. three foci in Figure 1C at 20 min, indicated by yellow arrows, are presumably infected by more than one phage. The phage λWT-FP (here referred to as WT for simplicity and easier comparison with the int and cII mutants used later, and FP is used to indicate this phage is labeled with fluorescent proteins. See detailed genotype in Table S2) also carries a D-mTurquoise2 marker, which encodes a chimera of the gpD decorative capsid protein fused to the mTurquoise2 fluorescent protein (Goedhart et al., 2012); this enables the monitoring of the lytic development by imaging cyan fluorescence (Zeng et al., 2010), as observed in Figure 1B. With this reporter system, the location and movement of the lambda DNA and attB can be tracked over time. Cells entering the lytic and lysogenic pathways are expected to show no (or very short-term) colocalization and long-term colocalization respectively. In Figure 1B, the cell entered the lytic pathway, indicated by accumulation of gpD-mTurquoise2 (120-180 min) and cell lysis (180 min). This lytic cell occasionally showed short-term colocalization at 0 and 120 min (see also Movie S1), which could be due to random collision or just imaging artifact. In contrast, long-term colocalization was observed for cells entering the lysogenic pathway. For example, in Figure 1C (see also Movie S2), one pair of phage DNA and attB colocalized beginning at 25 min, and another pair at 40 min, showing long-term colocalization.
This cell later divided and cell growth continued, indicating that the cell entered the lysogenic pathway. Occasional apparent separation of phage DNA and attB after long-term colocalization was also observed for lysogenic cells, for example, at 80 min in Figure 1C. The lambda DNA is ~48 kbp in length, and the SeqA 5′-GATC-3′-binding sites are relatively evenly distributed across the lambda genome (Fig. S1A).
Therefore, due to the uncertainty of the sites bound by SeqA-YFP on the lambda DNA and the movement of the bound unit resulting from diffusion (Weber et al., 2010), coupled with the fact that the tetO repeats are located ~1,500 bp upstream of attB, the actual distance between mCherry/attB and YFP/lambda DNA focus is expected to vary even after integration. This is probably why the attB and lambda DNAs are sometimes seemingly separated while the integration appears to have already happened.
To quantitatively determine colocalization, we then calculated the minimum distance between lambda DNA and attB, or Dis(λ-attB) at each time point for each cell. For the lytic and lysogenic examples in Figure 1B and C, Dis(λ-attB) of the lytic cell was usually above 0.5 μm (Fig. 1B), whereas in the lysogenic cells (Fig. 1C), it generally remained below 0.5 μm after integration (here beginning at 25 min). Moreover, the distribution of Dis(λ-attB) across all time points during the timelapse movies for all lysogenic (N = 44) and lytic cells (N = 515) showed that the lysogenic cells exhibited Dis(λ-attB) in the range of 0-0.5 μm much more often than lytic cells (Fig. 1D), suggesting that 0.5 μm might be a good indicator for determining lambda/attB colocalization.
In fact, for all lysogenic cells, after the designated integration time, we found that Dis(λ-attB) largely stayed below 0.5 μm over the remaining time course of the movie (Fig. S1B); therefore we defined "spot colocalization" as having a Dis(λ-attB) below 0.5 μm.

| Lyso-lysis:celllysiswithphageDNAintegration
Interestingly and surprisingly, we observed some cells entering the lytic pathway while also showing long-term colocalization of lambda DNA and attB. An example is shown in Figure 2A (see also Movie S3), where DNA colocalization occurred from 60 min until cell lysis (130 min), suggesting that phage DNA integration might be happening. Although unexpected, this event is actually consistent with the unanimous voting model proposed recently (Zeng et al., 2010), which states that each infecting phage in a cell can make a decision toward lytic or lysogenic independently. We then termed this event as "lyso-lysis".
Before quantifying DNA integration in lytic cells, we first excluded the contribution of random collision between lambda DNA and E. coli attB particles to the observed "colocalization". Here we used phage mutant λint − as a reference/control. λint − has a mutation in the integrase, which makes it defective in integration and lysogenization (Fig.   S2). As expected, the Dis(λ-attB) distribution for λint − (N = 510) showed significantly lower frequencies at 0-0.5 μm (Fig. 1D) compared to both the λWT-FP lysogenic and lytic cells. This integrase-dependent activity suggested that the observed DNA colocalization are likely due to the real DNA integration with some background of random collision.
To our surprise, we noticed that λint − infection sometimes also led to apparent lyso-lysis events. For example, in Figure 2B (see also Movie S4), DNA colocalization happened at 90 min and lasted until cell lysis at 135 min after λint − infection. We then compared the quantitative difference between colocalization for λWT-FP and λint − infected cells.
A relaxed criterion was then set up to call out cells with apparent "integration", for both λWT-FP and λint − infections. As long as the Dis(λ-attB) is below 0.5 μm, in the last 15 min before lysis, the cell would be categorized as lyso-lysis. At the same time, the effective number of phages infecting the cell (or effective Multiplicity of Infection, or eMOI) can be obtained by counting the initial phage DNA number.
We then obtained the frequency of lyso-lysis (calculated as number of lyso-lytic cells over total cells) at each eMOI. As expected, phage λint − infections led to lower percentages of lyso-lysis at all eMOIs compared to λWT-FP, although still showing a non-negligible number of apparent lyso-lysis events (Fig. 2C). Nevertheless, it suggests that lyso-lysis does exist in λWT-FP infections, although the frequency of lyso-lysis may be overestimated due to the contribution of false colocalization events reported by the system and allowed by our criterion.

| E. coliattBmigratestowardthecellpoleinlytic cells,leadingtomorecolocalizationwithlambdaDNA
To determine the quantitative differences between phage/E. coli DNA colocalization in lysogenic cells with true integration events and those reported lyso-lysis events for λWT-FP and λint − under our criterion, we analyzed the DNA trajectories of both lambda and attB over time.
We noticed that lytic cells showed colocalization of attB and lambda DNA at the cell pole more often than at other positions, similar to that of lyso-lysis by λint − infection shown in Figure 2B, where both lambda DNA and attB migrated toward the cell pole with time and eventually colocalized near the pole. In fact, when comparing the position of colocalization between lytic cells, 15 min before lysis, and lysogenic cells, from 185 to 200 min after infection (when phage DNA has already integrated and spot-tracking stops), we observed a significant difference (Fig. 3A). Colocalization happens most frequently between mid-cell and quarter-cell positions for lysogens, while in lyso-lytic cells, the location shifts drastically toward the cell pole.
We then ask whether the spatial colocalization patterns of λWT-FP-infected lytic cells result from natural preferences in attB and lambda DNA location during lytic development. In fact, the attB location distribution for lytic, lysogenic, and uninfected cells showed that the attB position shifted gradually toward the poles in lytic cells over time ( Fig. 3D), but not in lysogenic (Fig. 3E) or uninfected cells (Fig. 3F).
Interestingly, the phage DNA preferentially locates at the quartercell region, without significant changes along time in the lytic pathway (Fig. 3G), indicating that as the lytic cycle progresses, the attB moves gradually to a region where lambda DNA preferentially locates. Consistent with this hypothesis, when comparing the average lambda DNA and attB locations along time, it was obvious that attB migrated toward the lambda DNA and subsequently crossed lambda DNA traces (Fig. 3B). Therefore, the false lyso-lysis that we detected from λint − and some of λWT-FP infection were likely due to attB and lambda DNA being in close proximity to one another, especially toward the end of the lytic cycle. If this is the case, we expect that the DNA colocalization in the false lyso-lysis events would happen later compared to the actual integration events. We then compared the apparent "integration" times (when the colocalization started) for lyso-lysis and lysogenic cells (Fig. 3C). Integration in lysogenic cells happened mostly within 20 min after infection under our experimental conditions, which agreed with previously reported data (Freifelder & Levine, 1973), although late integration was also observed, leading to an aver- The underlying mechanism remains to be investigated.

| Lyso-lysis:aprocessregulatedbyCII
We designed a PCR experiment to examine whether E. coli genomic DNA liberated from lytic cells contains evidence of phage DNA integration as a complement to our microscopy data, as our reporter system does not specifically examine covalent DNA integration. We used primers specifically targeting the junction of E. coli and lambda DNA, spanning the attL region (Fig. 4A, red arrows) to confirm integration (Powell et al., 1994).  (Fig. 4B, bottom lane), corresponding to the time when cells began to lyse under these conditions, to release their DNA into the environment to be detected. Therefore, this suggested that DNA integration and thus lyso-lysis happened in some lytic cells. We then further quantified the percentage of lyso-lysis (defined as number of integrated DNA over total E. coli DNA) in the lytic cells, using qPCR with additional primers to quantify the E. coli DNA number (Fig. 4A, black arrows). Consistently, the percentage of lyso-lysis increased significantly between 60 and 90 min to 3.5% at an API of 1 (Fig. 4C). λint − was used as a negative control and no DNA integration was detected, as expected (Fig. 4C). These results further support the notion that phage DNA integration does occur in lytic cells. The number calculated here can be an underestimation since there may be multiple copies We further explored the molecular mechanism underlying this lyso-lysis phenomenon. CII promotes the expression of int from pI promoter in addition to activating the transcription of repressor CI from the pRE promoter for establishing the lysogenic pathway and inhibiting lytic gene expression through the anti-sense transcript from paQ . It is therefore possible that transient CII activation of pI without activation of pRE or paQ leads to the lysolysis that we observe, without producing enough CI and antisense Q transcript to establish a stable lysogen. Lyso-lysis may therefore be enhanced through common factors that lead to increased CII activity, such as higher APIs. Indeed, we found that as API increased, the percentage of lyso-lysis increased for λWT infection (Fig. 4D), similar to that of lysogenization (Kourilsky, 1973;Zeng et al., 2010). Since the CII activity correlates with its protein level , more CII might lead to more Int expression and thus more lyso-lysis. We then compared the percentage of lyso-lysis of λWT to two phage mutants, λcII 68 and λcII stable . λcII 68 carries a mutation which makes CII unable to dimerize to function (Datta et al., 2005;Schwarz et al., 1978), and λcI-I stable is less susceptible to FtsH degradation and is more stable (Jones & Herskowitz, 1978). Therefore the average CII activity after infection with the same API is: λcII 68 < λWT < λcII stable . As expected, the frequency of lyso-lysis followed the same trend: λcII 68 < λWT < λcII stable (Fig. 4D). Another interesting phenomenon was that the slope of lysolysis frequency as a function API (at log-log scale) is inversely related to the effective CII levels, showing the trend: λcII 68 > λWT > λcII stable , which is the same as that of the lysogenization frequency for these F I G U R E 4 Probability of lyso-lysis increases with API and CII activity. (A) A diagram showing the primer design for PCR and qPCR. For probing the integration using PCR, the primers span the junction between E. coli chromosome and lambda DNA, the integration junction, as indicated by red arrows, amplifying 500 bp in length. For qPCR, a different set of primers with the similar design is used. Another pair of primers is used for quantifying the E. coli DNA number, as indicated by black arrows. (B) PCR shows lyso-lysis events. E. coli was infected by λWT with an API of 1, and samples were taken every 20 min after infection for PCR. PCR was done either using the sample directly (upper lane, labeled as "whole") for detecting DNA integration from the whole sample, or using filtered supernatant to detect DNA integration in the lysed content (lower lane, "sup"). The 500 bp band indicating DNA integration shows up after 20 min in the "whole" sample as expected, and after 60 min in the "sup" sample indicating the lyso-lysis events. (C) The lyso-lysis frequency of λWT and λint − along time by qPCR at an API of 1. λWT: blue, λint − : green. No amplification of DNA integration is detected for λint − infection throughout the whole infection process (0-150 min). For λWT infection, the frequency of lyso-lysis increases with time, with 60-90 min showing a drastic increase, corresponding to the time for cell lysis and releases of DNA for detection. (D) Lyso-lysis is regulated by CII and has increased probability as API increases. Combined data of three qPCR experiments were shown. The frequency of lyso-lysis for all three phages including λcII 68 , λWT and λcII stable increases with API and the effective CII level inside the cell. The frequency of lyso-lysis follows the trend of λcII 68 < λWT < λcII stable , which is equivalent to their relative CII activities, suggesting that the process is regulated by CII. (E) The lysogenization frequency of λcII 68 , λWT and λcII stable follows different trends. The data for each phage is shifted with respect to API to compare with the Poisson distribution. λcII stable (red right triangle) follows Poisson distribution of n ≥ 1 (black solid line), while λWT (blue right triangle) follows n ≥ 2 (black dashed line) and λcII 68 (green right triangle) follows n ≥ 3 (black dash dotted line), suggesting that the cell needs 1, 2, or 3 of the corresponding phages to lysogenize. phages (Fig. 4E). Altogether, this suggests that similar to the lysogenic pathway, lyso-lysis is also regulated by the CII activity.

| DISCUSSION
Cellular decision making is an important process for the fitness and survival of all organisms, and has recently attracted numerous studies (Balazsi et al., 2011). Temperate phages, one of the simplest biological systems display lysis-lysogeny decision making after infecting the host bacterium. Classically, these decisions have been described as leading to incompatible outcomes, despite that early after infection, to proceed down one pathway without going at least part way down the other is impossible since early genes favoring lytic and lysogenic pathways are expressed from the same promoters, making the two pathways interconnected (Wulff & Rosenberg, 1983). In this work, by specifically labeling the lambda DNA and E. coli attB locus to allow for the direct visualization of prophage integration, we found that some lytic cells also have lambda DNA integration, previously thought to be present only in the lysogenic pathway. This finding offers a new perspective to the fundamental process of cellular decision making by bacteriophage lambda that phages cannot only concurrently develop two distinct pathways in the early infection period, but are also able to reach both outcomes in the late developmental stage.
Our recent model proposed that phages infecting the same E. coli cell can make individual votes to determine the cell's outcome, and unanimous voting by all infecting phages is a requirement for lysogeny (Zeng et al., 2010). This means that for those cells infected with more than one phage, it is possible that some phages vote lysogenic and the others lytic, leading to cell lysis, and observations of lyso-lysis support this model. However, our data also suggest that lyso-lysis exists even reported to function largely in cis (Echols et al., 1976), suggesting that the localization of Q might be restricted. Therefore, if enough physical separation of the DNA happens early before a decision is reached, each DNA might maintain its individuality and be able to make a different decision since one phage DNA committing to the lytic pathway would not be expected to force other DNAs to vote lytic due to the restricted localization of Q.
From the perspective of phage gene expression and regulation, it is also possible that lyso-lysis is simply a result of Int expression due to low levels of CII activity during the lytic process. The phage DNA integration is a highly regulated event, depending both on the integrase level as well as its competitor, the excisionase, Xis, which can alter the direction of recombination toward excision (Landy, 1989;Miller et al., 1981;Nash, 1981). During early infection, int is expressed from the pL promoter together with xis in the same transcript , however, the int mRNA level is lower compared to xis (Fig. S3, first 5 min), due to retroregulation of int by the downstream DNA element sib (Schindler & Echols, 1981). DNA integration is therefore unlikely to happen during the very early infection period. Later, for cells committing to the lysogenic pathway, int/xis expression from pL will be shut down by CI, while the pI promoter is activated by CII, allowing only int to be expressed , resulting in DNA integration. During lyso-lysis, Int must also be expressed to reach sufficient levels for integration. It is possible that a low level of CII is present to promote Int expression from the pI promoter. If so, the level might be too low to either affect other phage DNA or activate pRE to commit to lysogeny, allowing lysis to proceed. In fact, the int mRNA level after λcII 68 infection is substantial (Fig. S3). Since λcII 68 infection leads to >99% of lytic cells, especially at a low APIs (Fig. 4E) This means that there is an excess of unused lambda DNAs in lytic cells, and integration might be a good strategy for cells to utilize the free lambda DNA as a backup, should lysis unexpectedly fail.
Lysogens were reported to have a high frequency of having multiple phage DNA integrated even at low API (Table S1), as reported previously (Freifelder & Kirschner, 1971;Freifelder & Levine, 1973) and kinetic studies on the DNA integration process suggest multiple phage DNAs can either integrate sequentially or all at once (Freifelder & Levine, 1973, 1975. Whether all the lambda DNAs inside the cell will be integrated into the genome or not and what happens to the rest of lambda DNA remain unknown. Here, our observation that some phage DNA integrated very late while the cells seemed to have entered the lysogenic pathway ( Fig. 3F) suggest that some replicated unlabeled phage DNA may integrate early while the others, that is, the labeled DNA, diffuse throughout the cell until they are also integrated.
We observe that attB migrates to the cell pole in lytic cells, where phage DNA is more enriched as shown in Figure 3G. This movement pattern is very similar to the prophage integration in lysogenic cells (Tal et al., 2014). In our system, due to the limited number of lysogens obtained, the relatively low time resolution (5 min) and long preparation times (~10 min), most of integration happens within the first 20 min under the microscope. Therefore, we are unable to observe the detailed attB and phage DNA relative movements prior to integration in cells committed to lysogeny. However, for lytic cells, it is possible that the attB migration leads to more occurrence of colocalization of attB and lambda DNA, and thus, whenever Int is present, DNA integration can happen. This attB migration happens without the presence of Int, which leaves the driving force unknown. Further analysis reveals that in lytic cells the attB number does not increase significantly (Fig.   S4C), suggesting that the lambda lytic development inhibits the host DNA replication to some extent, possibly due to the competition of limited resources by actively replicating phage DNA (Mallory et al., 1990). Moreover, the length extension of lytic cells is slower compared to lysogenic cells (Fig. S4B). Therefore, lack of cellular division, compromised host DNA replication and slower length extension might contribute together to the drastic shift of attB distribution after lytic development.

ACKNOWLEDGMENTS
We are grateful to Ryland Young, Tamas Szekely, Michael Cortes and Chenxi Qiu for commenting on the earlier versions of the manuscript.
We thank Ryland Young, David J. Sherratt, and Jeffrey F. Gardner for gifting strains, and Jennifer Herman and Xindan Wang for helpful discussions on constructing the strains of E. coli DNA reporter.
We thank Craig Kaplan for providing the real-time