An open microscopy framework suited for tracking dCas9 in live bacteria

Super-resolution microscopy is frequently employed in the life sciences, but the number of freely accessible and affordable microscopy frameworks, especially for single particle tracking photo-activation localization microscopy (sptPALM), remains limited. To this end, we designed the miCube: a versatile super-resolution capable fluorescence microscope, which combines high spatiotemporal resolution, good adaptability, low price, and easy installation. By providing all details, we hope to enable interested researchers to build an identical or derivative instrument. The capabilities of the miCube are assessed with a novel sptPALM assay relying on the heterogeneous expression of target genes. Here, we elucidate mechanistic details of catalytically inactive Cas9 (dead Cas9) in live Lactococcus lactis. We show that, lacking specific DNA target sites, the binding and unbinding of dCas9 to DNA can be described using simplified rate constants of kbound→free = 30−80 s−1 and kfree→bound = 15−40 s−1. Moreover, after providing specific DNA target sites via DNA plasmids, the plasmid-bound dCas9 population size decreases with increasing dCas9 copy number via a mono-exponential decay, indicative of simple disassociation kinetics.

advent of super-resolution microscopy has been providing additional means to gain insights in biology 1 . In 33 addition to sub-diffractive imaging of cellular structures, super-resolution microscopy can be used to 34 observe the dynamics and interactions of individual proteins in living cells via single particle tracking photo-35 activatable localization microscopy (sptPALM) [2][3][4] . Compared to imaging applications, in vivo sptPALM has 36 additional experimental challenges to overcome: increased cellular background fluorescence, limited 37 photon budget of single genetically encoded fluorescent proteins and a high time resolution (millisecond 38 range) required to obtain molecular tracks. As full commercial solutions suitable for super-resolution 39 microscopy are costly and restrict the user in their choice of hardware and software, a multitude of 40 simplified custom-builds have emerged [5][6][7][8][9][10][11] . We believe, however, that the optimum between the 41 spatiotemporal demands of sptPALM, adaptability, price, and ease of installation has not yet been realized. sub-diffractive imaging as demonstrated earlier 12 , and total internal reflection fluorescence (TIRF) 47 microscopy. As we minimised the number of necessary components and used pre-aligned arrangements, 48 the miCube does not require extensive expertise in optics or engineering to set up. A detailed description 49 of all components and design choices can be found in Supplementary Note 1. 50 To demonstrate the capabilities of the miCube and to expand on previous sptPALM studies of  Cas 13,14 , we developed a novel assay that uses a heterogeneous expression system to explore the dynamic 52 nature of DNA-protein interactions in live bacteria and their dependency on protein copy numbers. To this 53 end, we expressed catalytically inactive Cas9 (dead Cas9 or dCas9) 15 fused to the photo-activatable 54 fluorophore PAmCherry2 16 under control of the heterogeneous inducible nisA promotor 17 55 Materials and Methods) in Lactococcus lactis. In this strain we introduced additional plasmids without 56 (pNonTarget) or with (pTarget) dCas9 DNA target sites (Materials and Methods). Representative cells from 57 each strain showed qualitative differences in the distribution of diffusion coefficients belonging to dCas9 58 (Fig. 1b). 59 We analysed the apparent diffusion coefficients as a function of apparent cellular dCas9 copy numbers 60 (Supplementary Note 2). The data of cells containing pNonTarget was fitted with three populations by 61 means of a global fit amongst the different segmented populations (Fig. 1c, Supplementary Figure 2). We 62 attributed the population with the highest apparent diffusion coefficient to freely diffusing dCas9 (Dfree * = 63 with a population corresponding to target-bound dCas9 (purple). The population size of the plasmid-bound 111 dCas9 decreases exponentially as a function of the cellular dCas9 copy number (inset). The error bar is 112 based on the 95% confidence interval of the histogram fits (least-squares); the dotted lines of the 113 exponential decrease is based on the 95% confidence interval of the equation fit (least-squares  (https://github.com/marcelocordeiro/medianfilter-imagej) was used to correct background intensity from 202 the movies 10 . In short, the temporal median filter determines the median pixel value over a sliding-window 203 of 50 pixels to determine the median background intensity value for a pixel at a specific position and time 204 point. This value is subtracted from the original data, and any negative values are set to 0. In the process, 205 all pixels are scaled according to the mean intensity of each frame to account for shifts in overall intensity. 206 The first and last 25 frames from every batch of 8096 frames are removed in this process. 207 Single particle localization was performed via the ImageJ 11 /FIJI 12  An apparent diffusion coefficient, D*, was then calculated for each track from the mean-squared 228 displacement (MSD) of single-step intervals 16 . In short, for every track with at least 4 localizations, the D* 229 was calculated by calculating the mean square displacement between the first four steps (taking skipped 230 steps due to a memory of 1 frame into account in the calculated distance), and taking the average of that. 231 These D* values were globally fitted with a custom-written MATLAB script, which iteratively fits (a 232 combination of non-linear regression and non-linear least-squares) a combination of 3 populations with a 233 variable D*-value and population size on multiple datasets. In short, equation (1) was globally fitted to 234 multiple histograms of the datasets. 235

Supplementary information Supplementary Note 1: Detailed description miCube
Custom-build instruments or partial instruments 17 have been presented before, such as the adaptation of commercial microscopes to add functionalities 18 , an open-frame microscope which provide dual-emission super-resolution 19 , simplified and low-cost super-resolution microscopes [20][21][22][23][24] , and a lab-on-a-chip design using polydimethylsiloxane (PDMS) rather than glass-based optics 25 . We designed the miCube to be easy to set up and use, while retaining a high level of versatility. The instrument and its design choices will be described in three parts: the excitation path; the emission path, and the 'cube' connecting the sample with the excitation and emission paths. Throughout this description, we will refer to numbered parts as shown in Supplementary Figure 1a

A. Excitation path
The excitation path is designed to be both robust and easy to align and adjust. The four laser sources located in an Omicron laser box are combined and guided via a single mode fibre towards a reflective collimator (nr. 18) ensuring a well-collimated beam. The reflective collimator is attached directly to an aperture (nr. 17), a focusing lens (nr. 16, 200 mm focus length), and an empty spacer (nr. 12). This excitation ensemble is placed in the 3D-printed piece specifically designed to hold the assembly into place (nr. 13). This holder is then attached to a right-angled mounting plate (nr. 14), which is placed on a 25mm internal reflection (TIR). The stage allows fine and repeatable adjustment of the excitation beam position on the back focal plane of the objective. By aligning the excitation beam in the centre of the objective, the microscope will act as a standard epifluorescence instrument. If the excitation beam is aligned towards the edge of the back focal plane, the miCube will operate in HiLo or TIR.

B. Cube and sample mount
The central component of the miCube is the cube (nr. 5) that connects excitation path, emission path, and the sample. The cube is manufactured out of a solid aluminium block maximising stability and minimising effects of drift due to thermal expansion (Supplementary Note 6). Black anodization of the block prevents stray light and unwanted reflections. The illumination path is further protected from ambient light by screwing a 3D-printed cover (nr. 11) on the side of the cube, as well as a door to close the cube off.
Next, the 'dichroic mirror -full mirror' part is assembled (nrs. 6-10). The dichroic mirror unit (nr. 7) consists of a dichroic mount that is magnetically attached to an outer holder. On the side of the dichroic mirror unit, opposing the excitation path, a neutral density filter (nr. 6) is placed to prevent scattered nonreflected light entering the miCube thereby minimizing background signal being recorded by the camera.
At the bottom of the dichroic mount assembly, a TIRF filter (nr. 8) is placed to remove scattered backreflected laser light from entering the emission pathway. This ensembled dichroic mirror unit is screwed via a coupling element (nr. 9) to a mirror holder containing a mirror placed at a 45° angle (nr. 10), which reflects the emission light from the objective to the camera. This completed 'dichroic mirror -full mirror' part is screwed into the backside of the miCube via two M6 screws, which hold the ensemble into place and directly in line with the excitation path (nrs. 12-18), the objective (nr. 3), and the tube lens (nr. 30).
Then, an objective (nr. 3) (Nikon 100x oil, 1.49 NA, HP, SR) is directly screwed into an appropriate thread on top of the cube. Around the objective, a sample mount (nr. 4) is screwed on top of the cube, which is designed to ensure correct height of the sample with respect to the parfocal distance of the chosen objective. We opted for using a sample mount, as it can be easily swapped for another to retain freedom in peripherals. For example, only the height of the sample mount has to be changed if the objective has a different parfocal distance as the one used here. We designed two different sample mounts (nr. 4a, 4b).
The first one can hold an xy-translation stage with z-stage piezo insert (nr. 2), to enable full spatial control of the sample (nr. 4a). The second one only holds the z-stage piezo insert, which decreases instrument cost (nr. 4b). In any case, the xy-translation stage with z-stage piezo insert, or only the z-stage piezo insert is screwed in place into corresponding threaded holes in the sample mount. A glass slide holder (nr.

1) is made from aluminium to fit inside a 96-wells-holder like the z-stage (nr. 2).
A tube lens ensemble is made (nrs. [27][28][29][30] which houses a 200 mm focal length tube lens (Thorlabs) in a 3D-printed enclosure which provides space to slot in an emission filter (nrs. 27,28). This ensemble is then attached directly to the miCube by screwing it into place with four M6 screws. The alignment of the tube lens is therefore exactly in line with the emission light, as the centre of the full mirror (nr. 10) is at the same height of the tube lens. The direction of the emission light can be aligned, which can simply be achieved by tuning the angle of the full mirror (nr. 10).
A cover (nr. 25) is attached to the tube lens ensemble to ensure darkness of the emission path, which is connected to the tube lens by a 3D-printed connector piece (nr. 26). On the other end of the cover, a 3Dprinted holder for 2 astigmatic lenses (nr. 21) is placed and screwed into place in the breadboard.
Astigmatic lenses (nrs. [22][23][24] can optionally be used to enable 3D super-resolution microscopy 26 . They can be easily changed for lenses with a different focal length or with empty holders. With this, astigmatism can be enabled or disabled, and a choice between more accurate z-positional information with a lower total z-range, or less accurate information with a larger range can be made. The Andor Zyla 4.2 PLUS camera (nr. 19) is placed behind the astigmatic lens holder, and is positioned in a 3D-printed camera mount (nr.

Supplementary Note 2: Estimating the copy number of dCas9.
The total copy number of dCas9 in a cell is not identical to the number of tracks found in each cell. Firstly, the UV illumination (405 nm wavelength) on the miCube required to photo-activate PAmCherry2 is not homogeneous over the complete field of view. To correct for this, a value for the average UV illumination experienced by each L. lactis cell is calculated. For this, a map of the UV intensity is made by placing a mirror on top of the objective and measuring the reflected scattering of the UV signal. Then, the mean UV intensity in the pixels corresponding to a cell according to the segmented brightfield images is stored. The cellular apparent dCas9 copy number is corrected for this normalized mean cellular UV intensity. Moreover, the cellular apparent dCas9 copy number was corrected for the average maturation grade of PAmCherry1, which is ~70% 28 . We assume the maturation grades of PAmCherry1 and PAmCherry2 to be similar.

Supplementary Note 3: Model simulations
The theoretical diffusion coefficient of the dCas9-PAmCherry2 construct is 36 -43 µm 2 /s in water, using the Stokes-Einstein equation 29 , and assuming a hydrodynamic radius of 5 -6 nm (PDB 5CZZ for Cas9 30 , and PDB 3KCT 28 for PAmCherry2). This is 18 -22 times higher than the found real diffusion coefficient of 1.95 µm 2 /s of dCas9-PAmCherry2. However, the increased viscosity and the molecular crowding in the cytoplasm have to be considered. For a construct this size, it is expected that the D0/Dcyto ratio is ~20. 31 Brownian motion of 33.000 tracks of particles was simulated with diffusion coefficients of 0 µm 2 /s, 0.43 µm 2 /s and 1.95 µm 2 /s in a 8:35:57 ratio, with an added 36.5 nm localization uncertainty, in a modelled cell consisting of 2 half-spheres with radius 0.5 µm connected by a straight cylinder of length 0.5 µm and radius 0.5 µm. 10 ms frames were calculated from 10 sub-frames in which the movement of the particles was Gaussian distributed according to the given diffusion coefficient. Each particle was given at least 100 ms of diffusion prior to measurement to mimic equilibrium that molecules may reach. Identical fitting was performed as in the rest of the study.
A second set of simulations was performed using only real diffusion coefficients of 0 µm 2 /s and 1.95 µm 2 /s (Supplementary Figure 3). However, this simulation was extended such that rate constants (corresponding to [free to bound] transition and [bound to free] transition) determine switching time between the states.
By increasing the rate constants, the resulting histogram of apparent diffusion coefficients becomes more convoluted. This is in line with the hypothesis that the transient interactions are a convolution of the bound population and the freely diffusion population (main text). This simulation can thus be used to approximate the rate constants (Supplementary Figure 3 g-k), although fitting this two-state simulation with the experimental data leads to larger residuals than the three-state simulation described above.

Supplementary Note 4: Unspecific interactions in the L. lactis genome
Interactions of dCas9 with genomic DNA are very complex to quantify, but we do offer some discussion on the matter. It has been reported using an in vitro assay that the DNA binding equilibrium of dCas9 shifts quickly to unbound dCas9 with an increasing number of mismatches 32 , but there is still some dCas9 bound to DNA even with minimal sgRNA-DNA complementarity. Because we are observing interactions at the short timescale of ~40 ms, short DNA complementarity will have an effect on the occupancy of this population.
Since we see a dependency of dCas9 copy numbers on the genome-bound population of dCas9 (Supplementary Figure 2), we hypothesize that the number of sites on the genome that dCas9 interacts with for longer than ~40ms is in the same order of magnitude as the highest dCas9 copy numbers (642).
If many more sites on the genome would be potentially occupied by dCas9, there would be no relationship between dCas9 copy number and genomic bound dCas9 population.
We can further estimate the amount of partially complementary DNA sites by taking the size of the genome in L. lactis into account (2.5 million base pairs). Assuming that all bases occur with similar frequency, 2-3 sites in the complete genome will have 10-bp-sgRNA-DNA complementarity, which increases to ± 800 sites for 6+ bp-sgRNA-DNA complementarity and to ± 3,000 sites for 5+ bp-sgRNA-DNA complementarity.
However, we do not know if the decrease in genomic-bound dCas9 continuous beyond 642 dCas9 copy numbers per cell, which could indicate that more partial target sites with high enough sgRNA-DNA complementarity for > 40ms dCas9 interaction are present on the genome. Moreover, this discussion does not take physical accessibility of the DNA into account. ATGGACAAAAAATACAGCATCGGCCTTGCCATCGGCACCAACAGCGTCGGGTGGGCCGTCATCACTGACGAGTATAAAGTGCCTAGTAAGAAGT  TCAAGGTGCTAGGCAACACTGACCGACACAGCATTAAGAAGAACCTGATCGGTGCTCTGCTATTCGATAGTGGTGAGACGGCCGAAGCTACAAG  ATTAAAAAGAACTGCAAGACGTAGATATACAAGACGTAAAAATCGTATTTGTTATTTACAGGAAATTTTTAGTAATGAGATGGCTAAGGTTGAT  GATAGTTTCTTTCATAGATTAGAAGAATCATTTTTAGTAGAAGAAGATAAAAAACATGAACGACATCCTATATTTGGAAATATAGTAGATGAAG  TAGCTTATCATGAAAAATATCCTACTATTTATCATTTACGTAAAAAATTAGTTGATAGTACTGATAAAGCTGATTTAAGATTGATATATTTAGC  ATTAGCACACATGATTAAATTTCGTGGTCATTTTCTAATTGAAGGAGATTTAAATCCTGATAACTCTGATGTTGATAAACTTTTTATTCAATTA  GTCCAAACTTATAATCAATTATTTGAAGAAAATCCAATTAATGCTAGCGGTGTAGATGCTAAAGCTATTTTATCAGCTAGATTAAGTAAAAGCA  GAAGACTAGAAAATTTAATTGCACAACTTCCTGGTGAGAAAAAGAATGGTTTATTTGGAAATTTGATTGCACTTAGTTTAGGTTTAACACCTAA  TTTTAAAAGTAATTTTGATTTAGCTGAAGATGCAAAACTTCAATTGAGTAAAGATACATATGATGATGATTTAGATAATTTATTAGCTCAAATT  GGTGATCAATATGCAGATTTATTTTTAGCTGCCAAAAATTTATCTGATGCTATTTTATTAAGTGATATATTACGTGTAAATACTGAAATTACTA  AAGCACCTTTATCAGCATCTATGATTAAAAGATATGATGAACATCATCAAGACTTAACATTATTAAAAGCATTAGTTAGACAACAATTACCAGA  AAAATATAAAGAAATTTTCTTTGATCAATCAAAAAATGGATATGCTGGATATATTGATGGTGGAGCATCACAAGAAGAATTTTATAAATTTATA  AAACCTATTTTAGAAAAAATGGATGGAACTGAAGAATTACTTGTTAAACTTAATAGAGAAGATTTATTAAGAAAACAAAGAACATTCGATAATG  GATCAATCCCACATCAAATTCATTTAGGTGAATTGCATGCTATTTTACGTAGACAAGAAGATTTTTATCCATTCTTGAAAGATAATAGAGAAAA  AATTGAAAAAATTTTAACTTTTAGAATTCCATATTATGTAGGACCTTTAGCACGAGGTAATTCTCGATTTGCATGGATGACACGTAAATCTGAA  GAAACAATTACACCATGGAATTTTGAAGAAGTTGTTGATAAAGGTGCTAGTGCACAATCTTTTATTGAAAGAATGACTAATTTTGATAAAAATT  TACCTAATGAAAAAGTATTACCAAAACATTCTTTATTATATGAATATTTTACTGTTTATAATGAACTTACAAAAGTAAAATATGTTACTGAAGG  AATGAGAAAACCAGCATTTTTATCAGGTGAACAAAAGAAAGCAATAGTTGATTTGTTATTTAAAACAAATCGTAAAGTTACTGTTAAACAACTT  AAAGAAGATTATTTTAAGAAAATTGAATGTTTTGATAGTGTTGAAATTTCTGGAGTTGAAGATAGATTTAATGCTAGTTTAGGTACATATCATG  ATTTATTAAAAATTATTAAAGATAAAGATTTTCTTGATAATGAAGAAAATGAAGATATTTTAGAAGATATTGTTTTAACATTAACATTATTTGA  AGATCGTGAAATGATTGAAGAACGTTTAAAAACATATGCACATTTATTTGATGATAAAGTAATGAAACAATTAAAAAGACGTAGATATACTGGA  TGGGGACGTTTATCTCGTAAATTAATTAATGGTATTAGAGATAAACAATCTGGTAAAACAATTTTAGACTTTCTAAAATCTGATGGATTCGCTA  ATCGTAATTTTATGCAATTAATTCATGATGATTCATTAACTTTTAAAGAAGATATTCAAAAAGCTCAGGTTAGTGGTCAAGGTGATAGCCTTCA  TGAACATATAGCTAACCTAGCTGGTAGTCCAGCAATTAAAAAAGGTATTTTGCAAACAGTGAAAGTAGTTGATGAACTTGTTAAAGTTATGGGT  CGTCATAAACCTGAAAACATTGTTATTGAAATGGCACGAGAAAATCAAACTACACAAAAAGGACAAAAGAATTCACGTGAACGTATGAAACGTA  TTGAAGAAGGTATTAAAGAACTAGGTAGTCAAATTCTTAAAGAACATCCAGTTGAAAATACACAATTACAAAATGAGAAATTATATTTATATTA  TTTACAAAATGGTCGTGATATGTATGTTGATCAAGAATTAGATATAAATCGCTTGTCAGATTATGATGTAGATGCAATTGTTCCTCAATCATTT  TTGAAAGATGATTCAATTGATAATAAAGTTTTGACACGTAGTGATAAAAATCGTGGTAAAAGTGATAATGTTCCTAGTGAAGAAGTTGTCAAGA  AAATGAAAAATTATTGGAGACAATTACTTAATGCTAAATTAATTACTCAACGTAAATTTGATAATTTAACAAAAGCAGAACGGGGAGGATTAAG  TGAACTTGATAAAGCTGGTTTTATAAAACGTCAATTAGTTGAAACAAGACAAATTACTAAACATGTAGCTCAAATATTAGATTCGCGTATGAAT  ACTAAATATGATGAAAATGATAAATTAATTAGAGAAGTTAAAGTTATAACATTAAAATCTAAATTAGTTAGTGATTTTAGAAAAGATTTTCAAT  TTTATAAAGTTCGTGAAATAAATAATTATCATCATGCTCATGATGCCTATCTTAATGCAGTAGTTGGAACAGCTTTAATTAAAAAATATCCAAA  ACTTGAAAGTGAATTTGTTTATGGTGATTATAAAGTCTATGATGTTCGCAAAATGATTGCTAAATCTGAACAAGAAATTGGTAAAGCTACAGCT  AAATATTTCTTTTATAGTAATATTATGAATTTCTTTAAAACTGAAATTACTTTAGCAAATGGAGAAATTAGAAAAAGACCATTAATTGAAACTA  ATGGTGAAACTGGAGAAATTGTTTGGGATAAAGGAAGAGACTTTGCAACAGTACGTAAAGTGTTATCTATGCCTCAAGTAAATATAGTTAAGAA  AACTGAAGTTCAAACAGGCGGATTTAGTAAAGAATCTATCTTACCAAAAAGAAATAGTGATAAATTAATTGCTCGTAAGAAAGATTGGGACCCT  AAAAAATATGGTGGTTTTGATTCTCCAACTGTCGCTTATTCGGTCTTAGTTGTTGCTAAAGTAGAAAAAGGTAAAAGTAAAAAATTAAAATCAG  TTAAAGAATTGTTAGGTATTACTATTATGGAAAGAAGTTCATTTGAAAAGAATCCTATTGACTTTTTAGAAGCCAAAGGTTACAAAGAGGTCAA  GAAAGACCTGATCATCAAACTGCCAAAGTACTCTCTCTTTGAATTAGAAAATGGACGTAAAAGAATGTTAGCATCTGCTGGTGAATTGCAAAAA  GGAAATGAATTAGCATTACCTAGTAAATATGTAAATTTCTTATACTTAGCTTCTCATTATGAAAAATTAAAAGGTTCACCGGAGGACAACGAGC  AGAAGCAACTTTTCGTGGAGCAACATAAACACTACCTCGACGAGATAATCGAACAAATTAGCGAGTTTTCAAAACGCGTCATCCTTGCCGATGC  CAATTTAGATAAAGTTTTATCAGCTTATAATAAACATAGAGATAAACCTATTAGAGAACAAGCTGAAAATATTATTCATTTATTTACTTTAACT  AATTTAGGTGCACCAGCTGCATTTAAATATTTCGATACAACAATTGATCGAAAAAGATATACATCAACTAAAGAAGTTTTAGATGCAACATTAA  TACATCAATCAATTACAGGATTATATGAAACACGTATTGATTTATCTCAATTAGGTGGTGATGGATCCGGAAGTTCAGCTATTATTAAAGAATT  TATGCGTTTTAAAGTTCATTTAGAAGGTAGCGTTAATGGTCATGAATTTGAAATTGAAGGAGAAGGTGAAGGTAGACCATATGAAGGTACACAA  ACAGCTAAATTAAAAGTTACAAAAGGTGGTCCATTACCTTTTGCTTGGGATATTTTGTCACCACAATTTATGTATGGTTCAAATGCTTATGTTA  AACATCCAGCTGATATTCCAGATTATTTTAAATTATCATTTCCAGAGGGTTTTAAATGGGAAAGAGTTATGAATTTTGAAGATGGTGGTGTTGT  AACAGTTACACAAGATTCATCTTTACAAGATGGTGAATTTATTTATAAAGTTAAATTAAGAGGTACTAATTTCCCTAGTGACGGACCGGTGATG  CAAAAGAAAACCATGGGATGGGAGACATTAAGTGAACGTATGTATCCTGAAGATGGTGCGTTGAAAGGTGAGCTGAAAGCTAGAACTAAATTGA  AAGATGGAGGCCACTATGATACTGAAGTAAAAACAACGTATAAGGCTAAGAAACCCGTTCAGTTACCAGGGGCATATAACGTTAATCGTAAACT  Linear drift calculations indicate that the system experiences a drift of 12.6 ± 11.6 nm/min in the lateral plane and 24.9 ± 14.5 nm/min in the axial plane. This is the average of three measurements performed on three different days (i.e. nine measurements in total). A typical drift measurement is shown in Supplementary Figure 9. The error bar indicates the 95% confidence interval of the fit of the bound population with the data (leastsquares). Uncorrected populations are used as partial fit of the apparent diffusion histograms arising from pTarget-containing cells (Fig. 1d).