L. pneumophila is a Gram-negative bacterium that encodes over 300 predicted virulence factors^{27, 28}, known as effectors. The effectors are translocated via the Dot/Icm Type IV secretion system into either the natural host, free-living amoeba, or the disease host, human alveolar macrophages, to manipulate cellular processes and establish an intracellular replication compartment known as the Legionella-containing vacuolar (LCV)^29–31. The study of L. pneumophila pathogenesis has been hindered by both the sheer number of effectors amassed by this pathogen^{27, 28} and the apparent redundancy amongst them, as illustrated in the lack of detectable growth defects upon disruption of individual or even entire chromosome islands of effector-encoding genes^{32, 33}.

Since our earlier studies had shown that silencing of up to ten genes by multiplex CRISPRi is achievable in L. pneumophila¹⁷, we built a library of 10-plex CRISPRi constructs (Table S1, MC, multiplex CRISPRi) capable of silencing more than 120 effector-encoding genes in groups of ten. The grouping of crRNA-encoding spacers into synthetic CRISPR arrays occurred based on two criteria: predicted protein function (groups 1-13, Table S1) and evolutionary conservation (groups A-D, Table S1). Protein functions were determined based on in silico predictions using Hidden Markov Model (HMM)-HMM-comparisons (HHPred; MPI Bioinformatics Toolkit)³⁴ and transmembrane predictors (TMHMM)³⁵. Evolutionary conservation information was extracted from published bioinformatics analyses of effector-encoding gene repertoires of 38 Legionella species²⁷. To confirm the efficacy of multiplex gene silencing, we used quantitative polymerase chain reaction (qPCR) to assess knockdown efficiency of gene expression by six of the seventeen MC constructs in L. pneumophila bearing a chromosomal insertion of S. pyogenes dcas9 at the thyA locus (Lp02(dcas9)¹⁷). We found that 10-plex gene silencing was reproducible for each construct and nearly all genes were knocked down at least 2-fold, with an average fold repression of one order of magnitude or more (Figure 1A).

Upon performing intracellular growth assays with this Lp02(dcas9) MC strain library in both U937 human macrophages and the amoeba Acanthamoeba castellanii, we found only a few intracellular growth phenotypes (Figure 1B). In U937 macrophages, the strains containing the MC array MC-A caused a growth defect as expected since it encoded crRNAs against lpg2815 (MavN) and lpg2300, core effectors known to be vital for intracellular growth²⁷. MC-10 improved growth of Lp02(dcas9) in U937 cells, likely due to the silencing of lpg1368 (Lgt1), lpg2862 (Lgt2), and lpg1488 (Lgt3), glucosyltransferase effectors previously known to act in a redundant fashion³⁶. MC-B caused a detectable growth defect of Lp02(dcas9) in U937 macrophages, suggesting it had silenced one or more virulence-critical effectors that had not been described before. Similar results were obtained during growth of the Lp02(dcas9) MC strain library in A. castellanii, where none of the MC arrays caused notable phenotypes. Together, these results show that silencing predetermined sets of genes, or even entire protein families, did not guarantee detection of synthetic lethal combinations of effectors.

To screen for synthetic lethality more comprehensively, we developed a protocol where CRISPR arrays were assembled de novo from oligonucleotide pairs containing crRNA-encoding spacers. Since each spacer is flanked by 36 base pair (bp) repeats of identical sequence in a canonical CRISPR array, we rationalized that repeat sequences could be split and then re-linked during CRISPR array assembly. As such, we designed complementary DNA oligonucleotides, which we call R-S-Rs (repeat-spacer-repeat), composed of a 24 bp spacer flanked by the terminal 12 bps of the upstream repeat and the starting 20 bps of the downstream repeat, each with four bp “sticky” overhangs (TGAA = a common Golden Gate cloning overhang, Figure 2A). The coming together of different R-S-R building blocks would recreate complete repeats without spurious nucleotides (12 + 20 + 4 = classic 36 bp repeat). Since the spacer sequence was buried within the repeat sequences, competitive advantage of one R-S-R over another for integration into an array should be negligeable. Furthermore, since the four bp overhangs are identical for each R-S-R, spacers can recombine into arrays of any length and any order. To cap arrays on either end with Invitrogen Gateway sites for downstream cloning, two additional “dead end” oligonucleotide pairs were designed known as attB4r-R and R-attB3r (Figure 2B). Precursor-crRNA cleavage sites were not necessary to be incorporated into the R-S-R building blocks as the precursor-crRNA would be processed into individual crRNAs by RNases endogenous to L. pneumophila.

As target genes, we chose a total of 44 predicted L. pneumophila effectors with single or multiple transmembrane domains that our above mentioned in silico predictions had identified. These 44 effectors are likely incorporated either into the membrane of the LCV or that of surrounding organelles to assist in nutrient acquisition, detoxification, and membrane fusion (Figure S1, Table S2). Despite high conservation of these transmembrane effector-encoding (TME) genes across all Legionella species^{27, 37}, only lpg2815, which encodes the metal ion transporter MavN^{38, 39}, has been shown to be virulence critical on its own. Hence, as proof-of-concept, we decided to probe this large group of genes with the randomized CRISPRi approach for synthetic lethality.

Assembly of CRISPR arrays followed simple, canonical cloning procedures (Figure 2C). Forty-four complementary R-S-R oligonucleotide pairs were incubated together with an aliquot of the attB4r-R and R-attB3r “dead end” oligonucleotide pairs and allowed to self-link after brief heat-denaturation and slow cooling. Arrays were then locked into place by treatment with T4 ligase. At this point, we found it critical to size-select arrays, as shorter arrays of only one or two spacers efficiently out-competed larger arrays during subsequent cloning steps. Size-selection proved challenging (Supplemental Note S1) but was ultimately accomplished through DNA gel purification of arrays that had first been introduced to an interim plasmid by Invitrogen Gateway cloning by way of the attB4r/attB3r sequences of the “dead ends”. To select for longer arrays, two gel sections were individually isolated (550-650 bps and 650-800 bps), corresponding to arrays of 2-4 and 4+ spacers in practice. Lastly, we used an Invitrogen Multisite Gateway Pro cloning strategy to create a final vector containing the tracrRNA-encoding sequence, and the pooled, size-selected randomized CRISPR array library flanked by a promoter (Ptet) and terminator (rrnB T1) sequence. Keeping the tracrRNA separate from crRNAs, unlike in single guide (sg)RNAs, kept the R-S-R building blocks at a manageable length. Upon navigating this assembly protocol twice, the final vector libraries were introduced into Lp02(dcas9) by electroporation and advanced to the next stage.

To assess the diversity of array composition, CRISPR array-containing vector libraries (Library 1 and 2) were harvested from two Lp02(dcas9) subpopulations, linearized, and analyzed by next-generation sequencing. Importantly, given the repetitive nature of the CRISPR arrays, canonical short-read sequencing technologies (such as Illumina or 454 sequencing) would not have been adequate as they would not have provided enough unique sequence overlap outside of the repeat sequence to map the reads back to a specific array within the mixed vector population. Instead, we made use of a high-throughput, low error rate long-read sequencing technology (PacBio Sequel). The average length of reads containing R-S-R elements was ~10,400 bp, correlating to the size of the vector backbone plus a multiplex array (sequencing metrics reported in Table S3).

We harvested vectors four times from each Library 1 and Library 2 stock to assess array diversity and library reproducibility. We designed a custom bioinformatics pipeline (discussed below) to condense data from all arrays with identical spacer content into one data set, regardless of spacer order; as such we found that Library 1 contained 784 unique spacer combinations and Library 2 contained 1251 unique spacer combinations (overlap = 250 combinations found in both). When requiring a stringent arbitrary cut-off of five or more raw (pre-sequence-depth normalized) read counts per spacer combination, a total of 231 and 519 spacer combinations were observed in Library 1 and 2, respectively, with 121 and 300 of those spacer combinations being present in each of the four replicates (A-D and E-H; Figure 3A). We found that arrays of spacers were most often 2- to 3-plex in nature, though some reached 5-, 6-, or even up to 11-plex to give an average of 3.3 spacers per array (Figure 3B, Table S4).

Each library contained a diversity of arrays that would assess silencing of nearly all 946 possible pairwise combinations of the 44 TME genes. Chord diagrams plotting each time spacers targeting two different TME genes were found in the same array provided a visualization of the remarkable pairwise comprehensiveness of each library, with the weight of the linking lines indicating frequency of observation (Figure 3C). Both libraries showed vast diversity in array composition, with Library 2 showing the greatest randomization of arrays leading to the most comprehensive pairwise coverage. Library 1, surprisingly, had an over-representation of arrays targeting the same five genes (lpg0963, lpg2223, lpg2552, lpg2888, and lpg3000) that accounted for 20% of all arrays in that library (indicated by the thick network of linking lines). Notably, from an array building vantage, the 5-plex arrays targeting these five genes were found to have seven different spacer orders, indicating that they had been generated multiple times during the assembly step. Together, these two libraries provide a vast collection of multiplex arrays to test for combinatorial TME gene contributions to L. pneumophila pathogenesis strategies.

To identify combinations of genes that, upon silencing, were detrimental to intracellular L. pneumophila replication, vectors from Lp02(dcas9) cultures bearing the pooled input array libraries were purified, linearized, and PacBio Sequel-sequenced to determine the identity and composition of arrays before infection (Figure 4A). From the same cultures, pooled bacteria were simultaneously introduced to a selective pressure, in this case host cell infection, and surviving bacteria were harvested from host cells after 72 h (for U937 macrophage infections) or 48 h (A. castellanii infections) and grown as single colonies on media plates. Vectors from arising bacterial colonies were purified, linearized, and analyzed again by long-read sequencing to determine the identity of arrays in the output vector pool.

Since each PacBio sequencing read reported the complete spacer composition of an array, barcoding was dispensable. A custom read count-based bioinformatics pipeline was designed to determine arrays that were under-represented in the output pool, and therefore, must have silenced genes vital for successful intracellular replication (Figure 4B, https://github.com/GLBRC/MuRCiS_pipeline). Taking a most stringent approach, only reads that were an exact match to the repeat sequence were analyzed. Spacers within an array were identified by searching each read for exact matches to all possible forward and reverse combinations of spacer sequences and the repeat sequence. The number of arrays with the same spacer composition were counted and an arbitrary cut-off of five raw counts in the input pool was required for further analysis of that unique spacer combination. The remaining unique array counts were normalized for run-to-run sequence depth variation, and the proportion of each unique array compared to the total number of sequenced reads containing a repeat sequence was determined for each sample. For each experiment, the proportion of normalized counts for each unique spacer composition in the input sample was compared to the normalized counts for the same spacer composition in the output sample to determine a fold reduction. The larger the reduction, the more disadvantageous silencing of the targeted genes was for intracellular L. pneumophila replication. We excluded any subset of spacers with an input vs. output reduction of less than five-fold. Together, this experimental and bioinformatics pipeline forms the basis for MuRCiS.

We first interrogated the contributions of TME genes to successful intracellular replication of L. pneumophila in human U937 macrophages. We performed four rounds of infections: two with Lp02(dcas9) bearing Library 1 (U-1A and U-1B) and two with Lp02(dcas9) bearing Library 2 (U-2A and U-2B). Each pair of infections, A and B, occurred independently of each other serving as biological replicates. To test technical reproducibility, we also collected output samples twice (T1 and T2) when performing experiments “A”. The four inputs for these experiments were A, C, E, and G (Figure 3A). In total, the four input cultures used to challenge U937 macrophages allowed for the probing of 710 of a possible 946 pairwise combinations (75%) of 44 TME when requiring a stringent arbitrary cut-off of five or more raw read counts (Figure 5A, Table S4). In fact, pairs were often in the input pool between eleven and 500 times allowing for in-experiment replication. Silencing mavN, which encodes the essential metal ion transporter, served as a control for the infection experiments. While hundreds to thousands of reads bearing the crRNA^mavN-encoding spacer were present in each input pool, each of the output samples (T1/T2 summed) had only ten or fewer reads (Figure 5B). These results indicated that MuRCiS could identify virulence-critical genes by monitoring changes in spacer abundance during host cell infection.

In total, 641 unique spacer combinations with five or more raw input reads were tested using this protocol (Figure 5C). Subsequent bioinformatic analyses found that 231 of those combinations showed a five-fold or greater reduction in read counts. All 59 combinations which contained the crRNA^mavN-encoding spacer were removed from the list, and 39 spacer combinations were identified as having a five-fold or greater read count reduction in both replicates of each library experiment (including both technical replicates). These 39 spacer combinations were examined further by gathering shorter versions of the arrays to track down the minimal number of spacers needed for the strain to show a fitness defect (Table S5). Twenty critical combinations of spacers were identified, of which most were 2- to 3-plex combinations (Figure 5D and 5E). Because of the uniqueness of Libraries 1 and 2, different gene combinations were discovered by each library to be virulence-critical, suggesting that the assembly and assay of more than one library was advantageous for comprehensiveness. Targeting of lpg2888 and lpg3000 was identified to be deleterious to L. pneumophila intracellular growth in experiments with both libraries. Table S6 shows data for each of the subsets of the 20 critical gene combinations, providing evidence that it was truly the silencing of the reported combination of genes, and not fewer or individual genes, that was needed for strain to show a fitness defect.

The nature of CRISPRi gene silencing and the novelty of the screen performed here mandated confirmation of targets by construction of strains with true chromosomal gene deletions, as the degree of target silencing can be influenced by a range of factors related to (1) crRNA design, (2) genetic environment, and on occasion (3) off-target effects. Individual and pairwise deletions of the candidate genes were made for five of the combinations identified in Figure 5E. All deletion strains were whole genome sequenced for confirmation of gene deletion and absence of background mutations and then used to challenge macrophages. Their ability to survive and grow over a period of three days was determined by plating assay.

This analysis confirmed the expected growth-inhibitory phenotype for three of the five gene pair deletions (Figure 5F), namely lpg2888-lpg3000, lpg1959-lpg2804, and lpg0096-lpg2804. The severity of the phenotype varied amongst the different pairs, ranging from a ~75-fold reduction in growth to complete inability to replicate in the host. For two pairs, lpg2552-lpg3000 and lpg0096-lpg2888, we found that deleting only one of the genes, lpg3000 and lpg2888, respectively, was sufficient for limiting L. pneumophila intracellular replication (Figure 5G). While we aimed to find critical combinations of genes, these “false positive hits” still identified genes important to L. pneumophila intracellular replication demonstrating much can be learned from this assay.

Having found critical combinations of genes required for growth of L. pneumophila in its disease host, we next interrogated the cohort of TME genes for their necessity during L. pneumophila growth in the natural host, A. castellanii. As before, we performed four rounds of infections: two with Lp02(dcas9) bearing Library 1 (AC-1A and AC-1B) and two with Lp02(dcas9) bearing Library 2 (AC-2A and AC-2B). The four inputs for these experiments were B, D, F, and H (Figure 3A). In total, these four inputs allowed for probing 672 of a possible 946 pairwise combinations (71%) of 44 TME genes when requiring a stringent arbitrary cut-off of five or more raw read counts (Figure 6A, Table S4).

This time, 571 unique spacer combinations with five or more raw input reads were tested during the experiments, leading to the identification of 71 spacer combinations with five-fold or greater reduction in read counts (Figure 6B). Of those, 21 emerged from both replicates of a given library experiment. We examined these spacer candidates further by tracking the most minimal number of spacers needed to attenuate L. pneumophila intracellular growth (Table S7). Six critical combinations of TME genes were ultimately identified, of which all were 2-plex combinations (Figure 6C and 6D, Table S8). Notably, silencing of lpg2888 and lpg3000 was again found to be deleterious to L. pneumophila growth in A. castellanii, as it was in U937 macrophages (Figure 5E). Having already constructed L. pneumophila stains bearing deletions in lpg2888 and/or lpg3000, we tested these for growth in A. castellanii. Surprisingly, in A. castellanii deletion of lpg3000 alone was sufficient to cause a growth defect as dramatic as that observed for the double mutant (Figure 6E). This contrasted with U937 macrophage experiments where only deletion of both lpg2888 and lpg3000 resulted in attenuated growth (Figure 5F).

Two of the other critical virulence-gene combinations identified in the A. castellanii experiments also contained lpg3000: lpg0086-lpg3000 and lpg0621-lpg3000, suggesting we would again observe deletion of lpg3000 sufficient of a maximum intracellular growth defect if tested. Though each library included 1-plex arrays expressing only crRNA^lpg3000 for silencing lpg3000, we did not identify those 1-plex arrays as being sufficient to attenuate virulence, likely because their decrease was not consistent between Library 1 and Library 2 experiments, although there was consistency between replicates A and B of each library (Table S8). A possible explanation was that incomplete knockdown (vs. complete knockout) of a seemingly important process was at a tipping point of phenotype extremes, and that silencing of an additional effector may have tipped the scale. If requiring a five-fold or greater reduction in only one library and not the other, an additional three (two of which are 1-plex; lpg0086 and lpg2223) and six (one of which is 1-plex; lpg0086) spacer combinations may qualify as critical to U937 macrophage and A. castellanii infection, respectively.

For any virulence-critical gene combination to be listed as hit during U937 macrophages infection (Figure 5) or A. castellanii infection (Figure 6), the CRISPR arrays silencing them had to pass the most stringent cutoff criteria, including that the spacer combination had to be represented in both biological replicates, A and B. There were instances though where CRISPR arrays were reduced in only one replicate because arrays bearing the same spacer combinations were absent in the other replicate (input overlap shown in Figure 3A). When requiring a five-fold or greater reduction in only one replicate, an additional 18 and 14 virulence-critical spacer combinations were observed in the U937 macrophage and A. castellanii experiments, respectively (Table S9). For example, in the infection experiment AC-2A, arrays silencing both lpg0096 and lpg2888 were nine-fold under-represented in the output pool after L. pneumophila growth in A. castellanii. L. pneumophila strains bearing deletions in one or both of these genes were examined for their growth in A. castellanii. Interestingly, unlike in U937 macrophages, we did observe a synergistic phenotype upon deletion of lpg0096 and lpg2888 (Figure 6F). These data represent yet another example of a host-specific phenotype and suggest that the proof-of-concept study performed here may have identified even more virulence-critical combinations of L. pneumophila genes than initially thought warranting examination of gene combinations that are considered hits in only one of the experimental replicates.

All L. pneumophila strains are listed in Table S10 and were derived from the parent strain L. pneumophila Philadelphia-1 Lp02 (thyA hsdR rpsL). Host cell lines included U937 monocytes (ATCC CRL-1593.2) which were differentiated into macrophages with 12-O-tetradecanoylphorbol-13-acetate (TPA, Sigma-Aldrich P1585) and A. castellanii (ATCC 30234).

Synthetic 10-plex MC array constructs were synthesized by GenScript and moved into pMME977 (thyA+) by Multisite Gateway Pro cloning (Invitrogen 12537-100) as previously described¹⁷. The MC arrays tested here bear a boxA sequence −58 bp upstream of the first repeat which was added after vector completion by quick-change PCR as described previously¹⁷. Final plasmids were introduced to Lp02(dcas9) by electroporation and the strains containing the plasmids were selected for on CYE plates⁴² without thymidine. Spacer composition of each array and their corresponding nucleotide sequences are listed in Table S1.

As described previously¹⁷, L. pneumophila cultures were grown overnight in AYE +Fe +Cys (10 g ACES, 10 g yeast extract per liter, pH 6.9 with 0.4 mg/ml cysteine and 0.135 mg/ml ferric nitrate) under non-inducing conditions. On the second day, cultures were subcultured twice (AM and PM, ~6-7 h apart) to OD600 0.2–0.3 with 2–3 mL fresh AYE +Fe +Cys containing 40 ng/mL anhydrous tetracycline (aTC, Clontech 631310). On the third day, cultures that had reached OD600 3-5 (post-exponential growth) were collected for mRNA analyses and/or used in host cell infections.

RNA extraction was performed on bacteria pellets after aTC induction using the Trizol Max Bacterial RNA Isolation Kit (Invitrogen 16096040). Contaminating DNA was removed using the Turbo DNA-free Kit (Invitrogen AM1907) and RNA was converted to cDNA using the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems 4368814). qPCR was performed using the SYBR Green Master Mix (Applied Biosystems 4367659) on a StepOne Plus Real-Time PCR System (Applied Biosystems) using comparative CT. qPCR primers are listed in Table S1. mRNA levels from different samples were normalized to a house-keeping gene (rpsL) and mRNA levels in CRISPRi strains were compared to that of a vector-bearing Lp02(dcas9) control strain using the ΔΔCT method⁴³ to determine fold repression.

U937 monocytes (ATCC CRL-1593.2) were maintained in DMEM (Mediatech 15-013-CV) + 10% FBS + glutamine at 37 °C. Two days prior to challenge with L. pneumophila, cells were plated on 24-well plates at 3×10⁵ cells/well with 0.1 μg/mL 12-O-tetradecanoylphorbol-13-acetate (TPA, Sigma-Aldrich P1585) to promote differentiation. L. pneumophila strains bearing each MC array were incubated under inducing conditions, as described above, and added to differentiated U937 macrophages in DMEM + 10% FBS + glutamine containing 40 ng/mL aTC at a multiplicity of infection (MOI) of 0.05. Plates were centrifuged for five minutes at 200 x g to increase bacteria-macrophage contact. After a 2 h incubation at 37 °C, extracellular bacteria were removed by washing the macrophages twice with DMEM + FBS + glutamine media containing 40 ng/mL aTC. Both 2 and 72 hours post infection (hpi) intracellular bacteria were collected upon macrophage cell lysis by addition of digitonin (0.02% final concentration), serially diluted, and spotted on CYE plates. Results are given as fold growth (CFUs harvested 72 hpi vs. 2 hpi) compared to that of the vector-bearing Lp02(dcas9) control.

A. castellanii (ATCC 30234) were maintained in PYG media⁴⁴ at 25 °C. The day before challenge with L. pneumophila, cells were plated on 24-well plates at 3×10⁵ cells/well. The morning of challenge, the media in the A. castellanii plates was changed from PYG to AC buffer⁴⁴ to promote starvation for 2 h at 25 °C. L. pneumophila strains bearing each MC array were incubated under inducing conditions, as described above, and added to the starved A. castellanii in AC buffer containing 40 ng/mL aTC at a MOI of 0.03. Plates were centrifuged for five minutes at 200 x g to increase bacteria-amoeba contact. After a 2 h incubation at 37 °C, extracellular bacteria were removed by washing amoeba twice with AC buffer containing 40 ng/mL aTC. Both 2 and 48 hpi intracellular bacteria were collected upon amoeba lysis by addition of saponin (0.05% final concentration), serially diluted, and spotted on CYE plates. Results are given as fold growth (CFUs harvested 48 hpi vs. 2 hpi) compared to that of the vector-bearing Lp02(dcas9) control.

R-S-R oligonucleotides were designed as the building blocks for array assembly. For each gene, the reverse complement of the 24 nucleotides downstream of the first PAM sequence (NGG) after the transcription start site was identified and would become the spacer sequence. This sequence was checked to be null of cut sites for SacI (GAGCTC), NotI-HF (GCGGCCGC), and AflII (CTTAAG) as these are used for downstream steps; if not, the sequence adjacent the subsequent PAM sequences was checked. Spacer sequences were then flanked on either side with one half of the repeat according to the following scheme: top oligonucleotide (5’ to 3’): tgaatggtcccaaaac-24 nt spacer-gttttagagctatgctgttt. Bottom oligonucleotide (5’ to 3’): ttcaaaacagcatagctctaaaac-24 nt complement-gttttgggacca. Standard oligos were ordered from Eurofins Genomics already resuspended in water at 100 μM concentration. Dead end oligonucleotide pairs for R-attB4r and R-attB3r did not contain a spacer but did contain the upstream or downstream repat sequence fused to the attB4r or attB3r Invitrogen Gateway cloning sequence, respectively.

For array assembly, top and bottom oligonucleotides for a single R-S-R or dead end oligonucleotide pairs were combined 1 μL:1 μL and were phosphorylated in a 50 μL reaction using T4 Polynucleotide Kinase (NEB M0201L) for 2 h at 37 °C. The reaction was stopped by addition of NaCl to a final concentration of 50 mM. Next, 2 μL of each phosphorylated R-S-R oligonucleotide pair for all 44 gene targets and 10 μL of the dead end oligonucleotides were combined in a single microcentrifuge tube. The microcentrifuge tube was placed in a heat block at 95 °C for 5 minutes and then the heat block was turned off and allowed to cool to room temperature for 2 h. Assembled arrays were preserved through ligation with T4 DNA ligase and its corresponding buffer (NEB M0202L) for 1.5 h at room temperature. The reaction was purified using a NucleoSpin PCR Clean-up Kit (Macherey-Nagel 740609.250) and eluted in 30 μL 1X TE buffer.

To size select arrays of interest, four 6.5 μL aliquots of the purified arrays were moved into the interim pDonor4r-3r in parallel by the standard Invitrogen Gateway BP reaction recipe (ThermoFisher 11789-020) which was allowed to remain at room temperature for 72 h prior to stopping the reactions with Proteinase K and transformation of all material into E. coli GC5 (Genesee Scientific 42-650) with selection on LB+Kan (30 μg/mL) plates. All colonies were scraped off plates using an inoculation loop and added directly to the A1 buffer of a Nucleospin Plasmid Purification Kit (Macherey-Nagel 740499.250) for plasmid isolation and elution in water. A Notl-HF (NEB R3189S) and AflII (NEB R0520S) double digest was performed on the four now recombined eluates, as well as one aliquot of the circular empty pDonor4r-3r plasmid, at 37 °C overnight. Next, a 1% agarose gel was used to separate the vector backbone from the excised size-ordered arrays, and we collected bands containing arrays ≥2-plex at 550-650 bps and 650-800 bps to be purified separately using a NucleoSpin Gel Purification Kit (Macherey-Nagel 740609.250). We also collected the cut vector backbone, but only from the empty pDonor4r-3r sample. For all steps forward, the 550-650 bps and 650-800 bps samples were kept separate, but procedures were carried out in parallel. As such, purified arrays were ligated back into the pDonor4r-3r backbone using T4 DNA ligase for 2.5 h at room temperature, all material was transformed with the plasmid into E. coli GC5, and re-circulated plasmids were selected for on LB+Kan plates. All colonies were scraped off plates using an inoculation loop and added directly to the A1 buffer of a Nucleospin Plasmid Purification Kit (Macherey-Nagel 740499.250) for plasmid isolation and elution in 20 μL 1X TE buffer. At this point, individual plasmids can be sequenced using a primer with the sequence: GTTTTCCCAGTCACGAC, if individually purified from a single colony, to confirm successful array assembly.

A Multisite Gateway Pro kit (Invitrogen 12537-100) was used to move arrays into the final vector. To generate the promoter-bearing donor plasmid pMME2162, the tet promoter was amplified from pMME1996 with primers containing attB5/attB4 ends for recombination into the pDonorP5-P4 via a standard Invitrogen Gateway BP reaction. To generate the terminator-bearing donor plasmid pMME2163, the rrnB T1 terminator was amplified from pMME1996 with primers containing attB3/attB2 ends for recombination into the pDonorP3-P2 via a standard Invitrogen Gateway BP reaction. Next, the size-selected arrays in the pDonor4r-3r plasmid were combined with pMME985 (tracrDNA in pDonorP1-P5r, described previously¹⁷), pMME2162, and pMME2163, and were added to the pMME977 destination vector (attR1/attR2, +thyA) by using three-times the standard Invitrogen LR reaction recipe (ThermoFisher 11791-020), with addition of β-mercaptoethanol to a final concentration of 1 mM, and allowed to remain at room temperature for 72 h prior to stopping the reactions with Proteinase K and electroporation into E. coli GC5 with selection on LB+Amp (100 μg/mL) plates. Next, all colonies (generally thousands for 550-650 bps samples and hundreds for 650-800 bps samples) were scraped off plates using an inoculation loop and added directly to the A1 buffer of a Nucleospin Plasmid Purification Kit (Macherey-Nagel 740499.250) for plasmid isolation and elution in 20 μL water. At this point, individual plasmids can be sequenced using a primer with the sequence: CAACCACTTTGTACAAGAAAGCTGGG, if individually purified from a single colony, to confirm successful construct assembly.

Lastly, 10 μL of these final plasmids were introduced into Lp02(dcas9) by electroporation and recipient cells were selected for on CYE plates. Colonies were scraped off plates using an inoculation loop, added to 15 mL AYE +Fe +Cys, vortexed, and frozen in 1 mL aliquots with glycerol (final concentration 15%) to be stored at −80°C. Freezer stocks of Lp02(dcas9) bearing each of the size-selected portions of a given library (Library 1 and 2) are listed in Table S10.

L. pneumophila bearing each size-selected portion of either Library 1 or 2 were patched from −80°C freezer stocks onto CYE plates. Two days prior to infection L. pneumophila cultures of each were grown overnight in 3 mL AYE +Fe +Cys under non-inducing conditions. On the second day, cultures were sub-cultured twice first in the AM to OD600 0.2 with 3 mL fresh AYE +Fe +Cys containing 40 ng/mL aTC and then ~6-7 h later, to OD600 0.2 with 4 mL fresh AYE +Fe +Cys containing 40 ng/mL aTC (two cultures each). On the third day, cultures pertaining to either Library 1 or 2 were combined in one tube to serve as a single culture for infection (input vector pool). For Library 1, two different 550-650 bps portions and one 650-800 bps portion was combined. For Library 2, one 550-650 bps portion and one 650-800 bps portion was combined.

U937 monocytes were maintained in DMEM + 10% FBS + glutamine and allowed to differentiate into macrophages with 0.1 μg/mL TPA for three days prior to L. pneumophila challenge. A. castellanii were maintained in PYG and starved in AC buffer 2 h prior to challenge with L. pneumophila. Ultimately, differentiated U937 macrophages or starved A. castellanii were plated in two to three 10 cm dishes at a density of 1×10⁷ cells per dish.

Infections were carried out with 40 ng/mL aTC in the host cell media at an MOI of 0.05 for differentiated U937 macrophages and 0.03 for starved A. castellanii. After addition of bacteria, the 10 cm dishes were spun down at 200 x g to increase bacteria-host cell contact and incubated for 2.5 h at 37 °C. Next, host cells were washed in their corresponding media with 40 ng/mL aTC to remove extracellular bacteria and then placed back at 37 °C for the duration of the infection, 72 h or 48 h for U937 macrophage and A. castellanii infections, respectively. To collect input vector pools, bacteria from the cultures used for infection were immediately pelleted by centrifugation and resuspended in 8 mL of RES-EF of the NucleoBond Xtra Midi EF kit prior to proceeding with the standard kit protocol (Macherey-Nagel 740420.50). Precipitated vectors were resuspended in 120 μL of water.

Following the infection duration, L. pneumophila were harvested from either the U937 macrophages or A. castellanii 10 cm dishes (15 mL vol) by addition of 300 μL of 1 % digitonin or 600 μL of 1% saponin, respectively. Next, 10 cm dishes were incubated for 10 minutes at 37 °C, followed by pipet agitation of the cells and collection of the media in 50 mL conical tubes. The conical tubes were shaken vigorously to promote host cell lysis and the L. pneumophila were pelleted by centrifugation for 10 min at 3500 x g. The resulting bacteria pellets were resuspended in a minimal amount of supernatant and plated on CYE plates. All colonies were scraped off plates using an inoculation loop and resuspended in 8 mL of RES-EF of the NucleoBond Xtra Midi EF kit prior to proceeding with the standard kit protocol for collection of the output vector pools. Precipitated vectors were resuspended in 120 μL of water.

Purified vector populations from input and output samples were linearized by addition of 24 μL of SacI restriction enzyme (NEB R0156L) with 16 μL of NEB r1.1 buffer and overnight incubation at 37 °C. To purify the linearized vectors, the AMPureXP Bead protocol (Beckman Coulter A63880) was followed using 100 μL of bead slurry. Final eluent volumes varied between samples based on the ability to get a bead-free eluent. Generally, ~10 μL of water was needed for input sample elution and ~40 μL of water was needed for output sample elution. Input DNA concentrations were generally ~200-300 ng/uL and output DNA concentrations were usually 300-700 ng/uL. Vector linearization was confirmed on a 1% agarose gel.

PacBio long-read sequence of the linearized vectors was performed by the NIH NICHD Molecular Genomics Core. Samples were prepared for sequencing on a Sequel instrument (PacBio) using the Sequencing Primer v4 and the Sequel Binding Kit 3.0. Samples were multiplexed and run on multiple SMRT cells simultaneously (# SMRT cells = # multiplex samples divided by 2 or 3) with 10 h of sequencing camera time.

PacBio sequencing files (ccs_bam files) were processed using a custom bioinformatics pipeline (https://github.com/GLBRC/MuRCiS_pipeline). Total read number and lengths were determined using Samtools⁴⁵ (version 1.9) and standard Linux commands (see GitHub for specific commands). All spacer and repeat forward and reverse orientation sequence combinations were used to identify the unique array sequence in the sequencing files using the “count_spacers_NIH.py” custom Python script (https://github.com/GLBRC/MuRCiS_pipeline). Reads lacking an exact match to the repeat sequence were removed from further analysis. Counts for unique spacers were normalized by dividing the read count plus a pseudocount of 1 by the total number of reads with repeat sequences multiplied by 10,000 in a modified Counts Per Million (CPM) equation. The 10,000 value was used instead of 1 million due to the lower number of overall reads from this PacBio sequencing.

To determine the pairwise correlation of two spacers and plot Chord diagrams, the “chord_correlation_plot_script.R” script was used. All plots were generated in R (version 4.1.0) and all scripts and command descriptions are available on GitHub (https://github.com/GLBRC/MuRCiS_pipeline).

L. pneumophila gene deletion strains were constructed using gene-specific versions of the pSR47s suicide plasmid listed in Table S10 and described recently⁴⁶. In our hands, each pSR47s plasmid was introduced into the L. pneumophila strain background of choice with the help of pRK600 by tripartite mating as described previously⁴⁷. Successful strains were confirmed by failure to grow on CYET+Kan, colony PCR, and whole genome sequencing.

For whole genome sequencing, genomic DNA of L. pneumophila was extracted using the Wizard Genomic DNA Purification Kit (Promega A1120), following the standard Gram-negative bacteria protocol, and resuspended in 10 mM Tris-HCl, 1 mM EDTA buffer. The Illumina Nextera XT DNA Library Preparation Kit (#FC-131-1096) was used to prepare libraries for sequencing on an Illumina MiSeq System by the NIH NICHD Molecular Genomics Core.

Nucleotide variants were called using HaplotypeCaller (version GATK3, Broad Institute). Integrative Genomics Viewer (version 2.4.14, Broad Institute) was used to visually confirm chromosome deletions. Final deletion strains are listed in Table S10.

Intracellular growth assays with deletion strains were performed exactly as for the 10-plex CRISPRi intracellular growth assays except that U937 monocytes were allowed to differentiate for three days prior to L. pneumophila challenge and L. pneumophila cultures were simply grown overnight without the inducer aTC. To overcome the thymidine auxotrophy of Lp02 strains, thymidine at a final concentration of 0.1 mg/mL was added directly to L. pneumophila cultures and to the DMEM + 10% FBS + glutamine media to promote growth in U937 macrophages with CFU plating on CYET plates, whereas pMME2400 (+thyA) was added to L. pneumophila strains by electroporation before growth in A. castellanii with selection on CYE plates. For U937 macrophage experiments the control strain was Lp02. For A. castellanii experiments that control strain was MML854 (Lp02 + pMME2400, G->A mutation at chromosome position 2134550).