Development of a non-invasive method to detect

Development of a non-invasive method to detect pericellular spatial oxygen gradients using FLIM

Neveen Amera Hosny 2011

This thesis is submitted for the

Degree of Doctor of Philosophy in Medical Engineering

School of Engineering and Material Science, School of Biology and Chemical Science, Queen Mary, Unversity of London

Declaration I confirm that the work presented in this thesis is the author’s own and that the copyright of this thesis rests with the author and no quotation from it or information derived from is may be published without the prior written consent.

Neveen Hosny

ii

Acknowledgements I’m thankful to say we have reached the end of our relationship. There have been many ups and downs during our short lived fling, but alas it was never meant to be, one day I would always have to stop writing and just print you off ready for binding. The experiences has been exhilarating, painful and damn right depressing, but I have gained a wealth of knowledge about new techniques and met a great set of colleagues and friends in the process. Initially I would like to thank my supervisors Martin Knight and David Lee for obtaining the grant and employing me to conduct this work. The assistance with completing this thesis has been most welcomed and would not have been possible without your help. I would also like to extend my thanks to the people who provided me with exciting opportunities to test my experiments on their systems. Robin Maytum in the biology department offered the use of absorption spectrometer. Ignacio Hernandez and Bill Gillin in the physics department facilitated the use of their luminescence system. Klaus Suhling at Kings provided the use of their single-photon confocal TCSPC system for comparison studies. Boris Vojnovic provided intellectual stimulation and continued interest in the development of the project, without which all enthusiasm might have been lost, for which I am indented to you. The progress of the work took place in two places the laboratory and the office. Here two groups of people have kept the bounce in my step and push in my stride! They were the ones that made this PhD entertaining and a big thanks will have to go out to the lab group: Angus Wann, Afshin Anssari-Benam, Kirsten Legerlotz, Hannah Heywood and Lauren Shor. Also the most important group being the ‘office boys’ without whom life would not have been as interesting or amusing. These righteous people consist of Percy Song, Steffanie Hunk, Fed Christ, Milan Rayamajhi, Osman, ‘the boy’ Henry Clarke and his treacle Clarise Sarell. I would like to thank my family Mum, Dad, Shereen and Georgina for always being around and providing me with lost of love and support! I would also like to thank my extended family Pekka and Anna who are always so interested in what is happening and always offer their support and love, Thank you! Finally and by no means least I will thank Raisa, who without which I would never have endeavoured on this crazy journey or managed to complete it without you constantly being there for my every need. The last mention goes to Indy Croft who is the coolest mini man on the planet and always knows how to cheer you up with a wiggle. Well that’s me signing off and for whoever picks up this thesis ‘you’re dangerous!’

iii

Abstract Extracellular oxygen concentrations affect cellular metabolism and influence tissue function. Detection methods for these extracellular oxygen concentrations currently have poor spatial resolution and are frequently invasive. Fluorescence Lifetime Imaging Microscopy (FLIM) offers a non-invasive method for quantifying local oxygen concentrations. However, existing FLIM methods also show limited spatial resolution >1 µm and low time-resolved accuracy and precision, due to widefield time-gate. This study describes a new optimised approach using FLIM to quantity extracellular oxygen concentration with high accuracy (±7 µmol/kg) and spatial resolution ( ≅ 0.3 µm). An oxygen sensitive fluorescent dye, tris(2,2′-bipyridyl)ruthenium(II) chloride hexahydrate [Ru(bipy)3]+2, was excited with a multi-photon laser and fluorescence lifetime was measured using time-correlated single photon counting (TCSPC). The system was fully calibrated with optimised techniques developed for avoiding artefacts associated with photon pile-up and phototoxicity, whilst maximising spatial and temporal resolution. An extended imaging protocol (1800 sec) showed no phototoxic effects on cells at dye concentrations of 0.99 and 880 nm excitation was fitted with linear model

xv

adjusted R2 >0.99. Inset graph represents zoomed in section of graph identified by the blue box..................................................................................................................... 135 Figure 75 – The calculated reduced chi-square of mono-exponential lifetime fitting [Ru(bipy)3]2+ at varying concentrations between 0.334 mM – 3.34 mM and increasing EOM gain at an excitation 780 nm. The inset graph represents the full range of lifetimes measured and the red box represents the zoom in of the full size graph. The error bars represent the standard deviation of lifetime fitting. The dashed line dictates χr2 ≤1.17 and the values above the data points are the count rate percentage. ................................................... 136 Figure 76 – The calculated reduced chi-square of mono-exponential lifetime fitting [Ru(bipy)3]2+ at varying concentrations between 0.334 mM – 3.34 mM and increasing EOM gain at 880 nm excitation. The small graph represents the full range of lifetimes measured and the red box represents the zoom in of the full size graph. The error bars represent the standard deviation of lifetime fitting. The dashed line dictates χr2 ≤1.17 and the values above the data points are the count rate percentage. ................................................... 137 Figure 78 – Box and whiskers plot of minimal pile-up influence over lifetime variability below the cut-off limit χr2 ≤ 1.17 at an excitation of 780 nm (black box) and 880 nm (red box). The whiskers represent the maximum and minimum values the box size is the standard deviation, the open square is the mean. ...................................................................... 137 Figure 79 – Lifetime decay of [Ru(bipy)3]2+ and deionised water solution at a concentration of 0.167 mM in atmosphere (Atm: black square and line) and with sodium sulfite (Na2SO3: red cicrle and line). Solutions were maintained at 37ºC and approximately 1x106 photons were collected in 120 sec. ........................................................................................... 139 Figure 80 – Weighted residual plot of mono-exponential fit with lifetime decay data for [Ru(bipy)3]2+ and deionised water solution at a concentration of 0.167 mM in atmosphere (Atm: black square) and with sodium sulfite (Na2SO3: red circle), χr2 is 1.101 and 2.175, respectively. .................................................................................................................................. 139 Figure 81 – Lifetime decay of [Ru(bipy)3]2+ in deionized water at a concentration of 0.412 mM in atmosphere and room temperature, at three laser excitation rates 1 MHz (black), 500 kHz (red), and 200 kHz (blue) using an ADC 1024..................................................... 143 Figure 82 – Average lifetime and standard error of varied repetition rate 200 kHz (black square), 400 kHz (red circle), 500 kHz (blue triangle) with incrementing collection times (n=3) plotted on a semi-log plot....................................................................................................... 144 Figure 83 – Adjusted R2-sqaure of mono-exponential fitted lifetime measurements at repetition rates of 200 (black square), 400 (red circles), and 200 (blue triangle) kHz against total number of photons counted per decay. The inset graph represents a zoomed in section represented by the red box.......................................................................................... 145 Figure 84 – The effect on the total and peak count of a lifetime decay through the combined effects of collection time and repetition rate 200 kHz (black squares), 400 kHz (red circles), and 500 kHz (blue triangles) displayed on a log plot. A: represents the total counts and B: represents the measured peak counts. ......................................................................... 146

xvi

Figure 85 – Two photon cross-section (σ2P) of a set of biological molecules and the excitation and emission peaks of [Ru(bipy)3]2+. A: TP absorption cross-section (σ2P) of intrinsic intracellular fluorophores (1 GM=1x10-50) all measured in saline solutions. B: Emission spectra of the intrinsic fluorophores shown on the left. Adapted from Zipfel et al. [327]. .................................................................................................................................. 150 Figure 86 – Emission spectra with TP excitation at varying concentrations of [Ru(bipy)3]2+. Four concentrations of [Ru(bipy)3]2+ dissolved in water at 3.34 mM (black squares), 1.67 mM (red circle), 0.845 (blue up pointing triangle), and 0.418 mM (green down pointing triangle) were excited at 780 nm TPE and intensity emission spectra captured between 507 nm and 722 nm. .................................................................................................. 154 Figure 87 – The effect of incrementing dye concentration on emission intensity. Increasing concentrations of [Ru(bipy)3]+2 dissolved in water at 37°C shows a positive linear relationship with increasing fluorescence intensity, adjusted R2>0.99. ........................ 155 Figure 88 –The lifetime of [Ru(bipy)3]2+ in water has been measured at varying dye concentration and plotted on a semi-log graph. Incrementing dye concentration measured using the current system TP confocal TCSPC (red circles) compared with a wide-field time-gated system (black squares), used by Sud[277], at a temperature of 25°C. Each point represents the mean and standard deviation of 3 replicates. ........................................ 156 Figure 89 – Semi-log plot showing decay of photon counts with time and the effect induced by temperature. Inset: represents the central section of the decay between 500-1500 sec displaying the shift in gradient when [Ru(bipy)3]+2 is exposed to 32°C (red line), 34°C (green line), and 39°C (dark blue line). ....................................................................... 157 Figure 90 – Response of [Ru(bipy)3]2+ in deionized water to incremental changes in temperature. Lifetime measurements (red squares) are shown with error bars in x-axis representing temperature drift and error bars y-axis representing standard deviation of fit. Measured data compared with Morris [212] (black line) shown with upper (UCL – purple dot dash line) and lower confidence limits (LCL – blue dot dash line)....................................... 157 Figure 91 – Measured lifetime of [Ru(bipy)3]+2 in phosphate buffer with incrementing concentration of pH. [Ru(bipy)3]+2 dissolved in varying concentrations of phosphate buffer between 57.5 pH at room temperature with a linear fit............................................................... 158 Figure 92 – Autofluorescent lifetime decays of DMEM and the components that make up cell culture media excited with a 8MHz repetition rate at 780 nm. Abbreviation: FBS – Fetal Bovine Serum, DMEM – Dulbecco's Modified Eagle Medium, H&A – HEPES and LAscorbic acid. ............................................................................................................ 159 Figure 93 – Lifetime decay of cell culture media and [Ru(bipy)3]2++media plotted on a semi-log graph with a 500 kHz repetition rate. Autofluorescence can be seen in the media decay and the [Ru(bipy)3]2+ + media decay. ......................................................................... 160 Figure 94 – Emission wavelength of [Ru(bipy)3]2+ with incrementing concentrations of FBS. Normalised intensity measurements at an excitation of 780 nm and repetition rate of 500 kHz across an emission range of 540-700 nm.............................................................. 161 Figure 95 – Mean intensity (n=4) and lifetime decays (n=5) of [Ru(bipy)3]2+ with increment concentrations of FBS . Left y-axis displays the mean lifetime (black squares) and right y-

xvii

axis refers to mean intensity (red circles). Data points are fitted with a dose response curve described by the solid red and black line (adjusted R2>0.96). The percentage of FBS in the cell culture media is represented by the purple dashed line, which corresponds to a lifetime of 388 ns. The maximum response is represented by the dotted green line with a lifetime of 389 ns. ............................................................................................ 161 Figure 96 – A Jablonski schematic diagram representing the electronic energy states of [Ru(bipy)3]2+ and the thermally activated non-luminescent energy state kdd. The electron is excited from ground state (S0) by light energy (hυA) to the singlet metal-to-ligand charge transfer (1MLCT) where the electron dissipates to the triplet metal-to-ligand charge transfer (3MLCT). At this point it can take optional pathways: non-radiative (kd+kq+k2[O2]), radiative (kr) or thermal activation (kdd). Abbreviation: energy change ∆E[212]........... 163 Figure 97 – Effect of temperature on concentration of dissolved oxygen. Experimental data of Benson (black squares) [33] and Rettich (red circles) [249] plotted against a theoretical model (blue line - Eq. 51) for determining the oxygen molality based on atmospheric mole fraction of oxygen (0.20939). Inset figure shows an expanded section of the response between temperature of 300 and 315K. ........................................................ 171 Figure 98 – Estimate oxygen molality in solutions at 37°C with incrementing oxygen mole fractions Theoretically calculated oxygen concentration in pure water (cO2 black squares), media solution ((cO2)I red square). Experimentally measured results from Gertz et al. [96] of pure serum ((cO2)Serum green triangles) are plotted with linear fits. ....................................... 174 Figure 99 – The effect of oxygen quenching on [Ru(bipy)3]2+ in solutions of deionized water (black squares) and media (red circles) at 37°C. The average water and media data points (n=10-20) are fitted with linear fits and have error bars in the x-axis relating to 0.1% for the accuracy of the Xvivo system and error bars in the y-axis representing the standard deviation between the results...................................................................................... 177 Figure 100 – Determining KSV with linear fits of water (black squares) and media (red circles) measurements of τ0/τ against [O2] concentration......................................................... 178 Figure 101 - Brightfield images were used to identify the two exposed cells seeded in agarose within a single chamber to aid relocation during viability tests. A: Two perpendicular lines appearing as dark colouration on the left and top of the image were drawn on the underlying coverslip and used for positioning markers; the image was taken using x20 objective lens and 2048x2048 image format (scale bar = 150 µm). B: The large box shown in A is represented by the zoomed in image of B, where the same two cells have been identified as cell 1 and cell 2 using a x63 objective lens with an image format of 1024x1024 (scale bar = 50 µm)................................................................................... 188 Figure 102 – Process of selecting and identification of cells 1&2 prior and after exposure to asses viability. A: Brightfield image of identified cells before exposure, scale bar = 5 µm. B: Lifetime image cells exposed to the imaging protocol and then chamber returned to the culture incubator for 24 h. C: Same cells relocated using brightfield (Figure 101A,B) and imaged using 488 nm and 564 nm excitation to reveal cells labelled with calcien AM (live-green) and EthD-2 (dead-red), respectively, scale bar = 50 µm. ........................... 189

xviii

Figure 103 – Image adjustment of 64x64 image format of a cell-agarose construct 40x106 cells/ml. A: Leica intensity image, B: Lifetime image pixel time 90 µsec C: Lifetime time pixel time 6.6 µsec, D: Lifetime image pixel time 11 µsec. Yellow circles represent matching cells. Scale bar =20 µm. ............................................................................................. 193 Figure 104 – Estimated collection times for various image formats with a minimum of 20,000 counts per pixel necessary to give a adjusted R2>0.9. Colour contour represents the number of hours for the total collection time and the red boxes are the standard image formats available on the Leica system including the minimum image format. ......................... 195 Figure 105 – Example of square binning selection for a single central pixel. The number inside the boxes represents the bin number and the total pixel shows how many pixels the bin number is equivalent to, adapted from Becker [27]. .................................................... 196 Figure 106 – Semi-log plot of the variation of pixel binning and the effect on collection time for varying image format sizes based on 20,000 counts per pixel and a collection rate of 5000 counts/sec. ................................................................................................................ 197 Figure 107 – Box and whiskers plot of the counts recorded per sec in the CFD and TAC of a noncellular agarose sample. A comparison between max and min CFD (black and red) and TAC (green and blue), respectively, described as a percentage difference. Box and whiskers represent maximum and minimum values (n=8) and standard deviation, respectively. ............................................................................................................... 200 Figure 108 – Box and whiskers plot of the peak and average counts/pixel in the non-agarose cellular sample. Compared with the estimated peak and minimum count of the max ADC and min ADC, based on the collection rate. Box and whiskers represent standard deviation and maximum and minimum values. ......................................................................... 201 Figure 109 – Diagrammatic example of the photon counting loss from CFD to acquired image. The components are shown on the left with the CFD, ADC, and example image. Based on the maximum collection rate and outcome from non-cellular agarose images an example estimates the effective collection rate of the setup. ...................................................... 202 Figure 110 – Example of test cell in SPCI data analysis program used for determining the lifetime components in the measured data set. A – Original intensity image, latter overlaid by lifetime data, B – lifetime plot for pixel selected by cross hairs. This is automatically selected as the pixel maximum intensity but this can be adjusted by moving the cross hair on the intensity/lifetime image. Highlighted red boxes identify where the binning, tau, threshold ‘thld’, total number of counts and reduced chi-square are displayed corresponding to this lifetime. C – Histogram of number of pixels at each lifetime values and the rainbow lookup table which can be adjusted by moving the max and min markers. The lifetime of each pixel are then displayed on the intensity image (A) based on this rainbow lookup table. ..................................................................................... 204 Figure 111 – Lifetime maps of test cell in agarose with varying binning regimes and colour map selections. Left: Same colour map cursor selection, but varying bin setting from 2, 3, and 4. Right: Adjusted colour map cursor selection cutting off wings of histogram, but using the same bin settings as in left side images. ................................................................. 205

xix

Figure 112 – Example of test cell in TRI2 time resolved data analysis software for extraction of lifetime information. A – Intensity image, B- Lifetime plot for pixel selected in intensity image, C- Residual corresponding to the lifetime plots above. Highlighted red boxes identify where the binning, tau, total number of counts and reduced chi-square are located. ...................................................................................................................... 206 Figure 113 – Lifetime decay of selected pixel in the intensity image of the test cell at different square binning sizes 2, 3, and 4. The increasing area of the bin is also shown by the yellow box in intensity images, with corresponding lifetime decay and residual plot below........... 207 Figure 114 – Comparison of lifetime maps from the test cell using TRI2 and SPCImage software with binning of 2,3, and 4. The TRI2 software lifetime colours blue to red as 322 to 368 ns and the SPCI the reversed colour red to blue, but the lifetime range changes as shown in figure 13 right. ....................................................................................................... 208 Figure 115 – Diagrammatic representation of a user defined mask applied to extract lifetime decay information. The mask is used on an intensity image of an unresolved test cell where only the selected pixels in the mask are transferred to the time resolved analysis section and binned to produce a lifetime decay. Separate masks are created along the entire image to produce a single data set of lifetime values versus distance from the cell periphery. .................................................................................................................. 210 Figure 116 – Extracted lifetime measurements from chondrocyte with incrementing distance from cell periphery with assumed linear fit (red line) in close proximity to the cell. Test cell 1 shows a significant (p