bioRxiv preprint doi: https://doi.org/10.1101/2020.10.05.326009; this version posted October 5, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

1 distribute randomly to, but not within, human neutrophil nuclear lobes

2

3 Christine R Keenan1,2 *, Michael J Mlodzianoski 1,2 *, Hannah D Coughlan1,2, Naiara G Bediaga1,2,

4 Gaetano Naselli1,2, Erin C Lucas1,2, Qike Wang1,2, Carolyn A de Graaf1,2, Douglas J Hilton1,2, Leonard

5 C Harrison1,2, Gordon K Smyth1,3, Kelly L Rogers1,2, Thomas Boudier1,2, 4, Rhys S Allan1,2 * and

6 Timothy M Johanson1,2 *

7

8 Affiliations: 1The Walter and Eliza Hall Institute of Medical Research, Parkville, Victoria, 3052,

9 Australia. 2Department of Medical Biology, 3School of Mathematics and Statistics, The University of

10 Melbourne, Parkville, Victoria, 3010, Australia. 4 Institute of Biology Paris-Seine, Sorbonne Université,

11 Paris

12

13 Corresponding author: Timothy Johanson ([email protected])

14

15

16

1 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.05.326009; this version posted October 5, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

17 Abstract

18

19 The proximity pattern and radial distribution of territories within spherical nuclei are

20 well understood to be random and non-random, respectively. Whether this distribution pattern is

21 conserved in the partitioned or lobed nuclei of polymorphonuclear cells is unclear. Here we use

22 chromosome paint technology and a novel high-throughput imaging analysis pipeline to examine the

23 chromosome territories of all 46 chromosomes in hundreds of single human neutrophils – an abundant

24 and famously polymorphonuclear immune cell.

25

26 By comparing the distribution of chromosomes to randomly shuffled controls, and validating with

27 orthogonal chromosome conformation capture technology, we show for the first time that all human

28 chromosomes randomly distribute to neutrophil nuclear lobes, while maintaining a non-random radial

29 distribution within these lobes. Furthermore, by leveraging the power of this vast dataset, we are able

30 to reveal characteristics of chromosome territories not detected previously. For example, we

31 demonstrate that chromosome length correlates with three-dimensional volume not only in

32 neutrophils but other human immune cells.

33

34 This work demonstrates that chromosomes are largely passive passengers during the neutrophil lobing

35 process, but are able to maintain their macro-level organisation within lobes. Furthermore, the random

36 distribution of chromosomes to the naturally partitioned nuclear lobes suggests that specific

37 transchromosomal interactions are unimportant in mature neutrophils.

2 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.05.326009; this version posted October 5, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

38 Introduction

39

40 First proposed in 1885 (1), interphase chromosomes maintaining a territorial organisation is now a

41 widely accepted principle of nuclear organisation in most eukaryotes (2). This is unsurprising given

42 the importance of this organisation to functions as fundamental as expression and DNA repair.

43 For example, the radial position of a chromosome within the nucleus is strongly correlated with its

44 transcriptional activity (3, 4). Furthermore, the proximity of chromosomes to one another (both

45 homologous and non-homologous) is thought to be important during DNA repair (5) and potentially

46 even in direct gene regulation (6-10).

47

48 While the radial distribution of chromosomes is well understood to be non-random (2), the position of

49 chromosomes relative to each other, or proximity pattern, is contentious, with reports of both non-

50 random (11-15) and random distributions (16). In the absence of being able to observe interphase

51 chromosome movements in live cells over long periods of time, combined with the absence of

52 physical barriers to restrict chromosome movement, it is possible that these studies simply differ in

53 their detection of transient or infrequent interphase chromosomal interactions or movements.

54

55 Human neutrophils constitute approximately two thirds of the immune cells in human blood. They are

56 readily identifiable by their polymorphonuclear nature with their nuclei being segmented into 2-6

57 lobes joined only by thin filaments of nucleoplasm (17). Here we exploit this natural nuclear

58 segmentation to examine the importance of interphase chromosome distribution. We hypothesise that

59 if interactions between chromosomes are biologically important these chromosomes would

60 preferentially locate together into neutrophil nuclear lobes, enabling continuing interaction.

61

62 Using chromosome paint technology alongside novel high-throughput image analysis pipelines, we

63 have examined all 46 human chromosomes in 240 single neutrophils. We reveal that while the radial

64 distribution of chromosomes within neutrophil nuclear lobes is non-random, the distribution of

3 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.05.326009; this version posted October 5, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

65 chromosomes to lobes is in fact random, suggesting that gene regulatory interactions between specific

66 chromosomes are highly unlikely to occur in human blood neutrophils.

67

68 Results

69

70 Novel analysis pipeline detects the position and characteristics of all human chromosomes in

71 three-dimensions

72

73 Given the complexity of examining all 46 human chromosomes in hundreds of single, segmented

74 neutrophil nuclei, we developed a bespoke image analysis pipeline to detect the position and three-

75 dimensional characteristics of all chromosomes in images generated using chromosome paint.

76

77 In brief, each of the 22 autosome pairs and the X and Y chromosomes within fixed healthy male

78 human blood neutrophil nuclei (Supp Fig 1A) is “painted” with a specific combination of fluorescent

79 oligonucleotides within the chromosome paint mix (Fig 1A, B). The whole nucleus is then imaged

80 and analysed using our analysis pipeline (Fig 1C). First, the intensity of each of the five channels in

81 each individual image is normalised. A malleable grid with lines that flex to incorporate nearby

82 voxels of similar channel intensity patterns is then applied to each image. This flexibility allows the

83 grid to capture the highly variable three-dimensional shapes formed by chromosomes. Adjacent cubes

84 of the grid that share channel intensities are then combined to create “objects”. Based upon the

85 expected spectral combinations for each chromosome (Fig 1 B) these objects are then assigned as

86 chromosomes. To avoid the common pitfall of arbitrary thresholding to define genuine signal from

87 background (2), here every image has an individually determined threshold. This threshold is

88 calculated by automated sequential testing of various thresholds to determine the value which

89 maximises the number of objects with genuine chromosome channel combinations, while minimizing

90 spectrally spurious objects (e.g. detected objects that have a channel combination differing from the

91 24 specific chromosome combinations).

92

4 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.05.326009; this version posted October 5, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

93 Importantly, chromosome paint and our image analysis pipeline detect the majority of the 22 human

94 autosomes and the sex chromosomes at approximately the expected proportions (Fig 1 D, Supp Fig

95 1B). While some chromosomes appear more difficult to define or detect (e.g. chromosome 13, 15, 21

96 and 22), the proportions of chromosomes detected are similar across human immune cell types (CD4+

97 and CD8+ T cells) (Supp Fig 1C, D), suggesting that the variation in detection frequency observed is

98 technical, not biological.

99

100 We next examine the three-dimensional character of all 46 chromosomes, including volume (Fig 1 E)

101 and surface area (Fig 1 F), among others (Supp Fig 1E, F). While the physical characteristics of the

102 chromosome territories varies greatly across neutrophil nuclei, the volume and surface area of the Y

103 chromosomes are, as expected, consistently and significantly (P=0.02 and P= 0.04, respectively)

104 smaller than the X chromosome (Fig 1 E, F).

105

106 While the differences in the three-dimensional character of the chromosomes are subtle, the ability to

107 examine all chromosomes (as opposed to between 2-7 chromosomes (18-20)) in large numbers of

108 cells affords the power to reveal correlations previously missed. For example, contrary to previous

109 studies suggesting that there is no relationship between chromosome linear length and three-

110 dimensional volume (18, 20), our analysis reveals a significant linear relationship between the two (r2

111 = 0.359, P = 0.001), not only in neutrophils (Fig 1G), but other human immune cells (Supp Fig 1G).

112

113 Thus, while chromosome paint data contains variance in both chromosome detection and three-

114 dimensional parameters, the scale of our dataset allows elucidation of biologically important

115 correlations and phenomena that were undetectable using previous methods.

116

117 Human neutrophil chromosomes distribute randomly to nuclear lobes

118

119 There is conflicting evidence as to whether the proximity pattern of chromosomes is random or non-

120 random, even within polymorphonuclear nuclei (18, 20-23).

5 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.05.326009; this version posted October 5, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

121

122 In order to examine the distribution of chromosomes to neutrophil nuclear lobes, we first developed a

123 high throughput pipeline to define the lobes. In brief, the pipeline uses watershed analysis from

124 manually assigned seeds within each lobe (Supp Fig 2A) to rapidly define nuclear lobes (Supp Fig

125 2B). Consistent with previous studies (22), we find that the majority of healthy human blood

126 neutrophils have between 3-4 lobes (Supp Fig 2C) of approximately equal volumes (Supp Fig 2D).

127

128 Next, we performed a lobe colocalization analysis to examine the frequency with which each

129 chromosome is located within a neutrophil lobe with any other (Fig 2A). We find significant non-

130 random distribution of 44 pairs of chromosomes (Supp Table 1). However, this analysis does not

131 consider the character of the chromosomes. For example, a larger chromosome would likely have less

132 lobe partners due simply to occupying more lobe volume. Thus, to avoid the confounding impacts of

133 chromosome character, we repeated our colocalisation analysis on randomly shuffled controls (Fig

134 2B). To generate these shuffled images, we created an imaging framework that randomly places every

135 chromosome (with its individual three-dimensional character and that of the associated nucleus

136 preserved) within the nuclear space of each individual image. Chromosomes were not permitted to

137 overlap one another or the nuclear boundary. This shuffling was performed five independent times for

138 each of the 240 biological images. The colocalisation frequency was then examined in each set of

139 shuffled images to determine the variability of shuffling, but more importantly the impact of

140 chromosome character on lobe sharing frequency. The analysis revealed significant lobe sharing in

141 the randomly shuffled images, suggesting that chromosome character does impact lobe distribution.

142

143 To overcome the influence of chromosome character on lobe distribution, we filtered our significant

144 lobe partner pairs (Supp Table 1) by significance of co-lobe localisation greater than that observed in

145 the randomly shuffled analysis. Eleven chromosome pairings surpassed this threshold (Supp Table 2),

146 suggesting that these chromosomes may preferentially co-localise in neutrophil nuclear lobes.

147 However, the pairs in question share very similar spectral combinations, suggesting that perhaps the

148 significance observed is an artefact of miscalling. For example, if one region of chromosome 1

6 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.05.326009; this version posted October 5, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

149 (painted Cy5) was occasionally miscalled as (painted Cy5 and FITC) or vice versa this

150 could lead to a significant colocalization score.

151

152 To validate our chromosome paint findings, we performed an orthogonal analysis of chromosome

153 proximity using in situ HiC (24). In situ HiC utilises proximity ligation to determine the three-

154 dimensional association of regions of DNA in populations of fixed nuclei (Fig 2C). While in most

155 cells the vast majority of ligation events occur within chromosomes, transchromosomal interactions

156 can be used as a measure of chromosome-chromosome proximity (24). Performing in situ HiC on

157 human neutrophils we find that they exhibit dramatically more transchromosomal interaction than

158 human immune cell types with spherical nuclei (Fig 2D). This is likely due to the increased physical

159 confinement within neutrophil nuclei (25). If, as hinted at by chromosome paint, particular

160 chromosomes are more frequently co-localised to neutrophil nuclear lobes than expected by chance,

161 we would expect to observe an enrichment of transchromosomal interactions between these

162 chromosomes compared to other chromosome pairs. However, we find that the distribution of

163 transchromosomal interactions in human blood neutrophils is predominantly dependent on

164 chromosome length (r2 = 0.968, Fig 2E) with no notable increase in interaction observed between any

165 of the chromosome pairs suggested to co-localise by chromosome paint (Fig 2F).

166

167 Thus, using two independent methods, and by comparing to randomly shuffled controls, we find no

168 consistent evidence of non-random chromosome distribution to human neutrophil nuclear lobes,

169 suggesting that the distribution is in fact random.

170

171 Human neutrophil chromosomes do not position randomly within nuclear lobes

172

173 Consistent with findings in spherical nuclei (16), our distribution analysis finds that the proximity

174 pattern of chromosomes within neutrophil nuclei is random. Alongside random proximity pattern,

175 non-random radial chromosome distribution is also well documented. As such, larger chromosomes

7 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.05.326009; this version posted October 5, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

176 are consistently observed closer to the nuclear periphery than their more diminutive counterparts (16).

177 This has even been observed in a handful of neutrophil chromosomes (19).

178

179 To determine whether all human neutrophil chromosomes exhibit this relationship with the nuclear

180 periphery, we examine the position of chromosomes relative to the lobed nuclear membrane. To do so

181 we first break the radius of each lobe into a continuous scale from 0 to 1, 1 being the three-

182 dimensional centre of the lobe and 0 being the nuclear periphery (Fig 3A). We then give each

183 chromosome within the lobe a value between 0 and 1 based upon the mean position of its total volume

184 (Fig 3B). By applying this method to all chromosomes in all lobes we are able to determine the

185 position of each chromosome relative to the lobe boundary. As such, we find a significant relationship

186 between chromosome linear length and its relative position within a lobe (r2 = 0.749, P = 4.60E-08),

187 with larger chromosomes more likely to be found nearer the nuclear periphery than shorter

188 chromosomes (Fig 3C). Importantly, when we apply this same method to our randomly shuffled

189 control images the association is no longer detected (Fig 3D), strongly suggesting that it is not simply

190 the volume of the larger chromosomes underlying the relationship.

191

192 Thus, as in spherical nuclei, chromosomes within the highly physically restricted neutrophil nuclear

193 lobes are positioned in a non-random and size-influenced manner.

194

195 Discussion

196

197 Here we outline an automated method for the analysis of the three-dimensional character and position

198 all 46 chromosomes in hundreds of single cells. By leveraging the power provided by this vast data

199 set we resolve a number of longstanding questions in the field of chromosome organisation. We

200 demonstrate in numerous immune cell types that chromosome length does in fact have a linear

201 relationship with chromosome volume. Moreover, in human neutrophils we show that chromosome

202 lobe distribution is random, while chromosome radial position within lobes is non-random.

203

8 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.05.326009; this version posted October 5, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

204 The similarities in chromosome position between lobed neutrophil nuclei and spherical nuclei suggest

205 that neutrophil chromosomes simply move and adapt to the more restricted nuclear environment

206 created by lamin-mediated constriction (26). Thus, it appears that most chromosomes are simply

207 passengers in the lobing process. However, there is one chromosome that appears to influence nuclear

208 morphology – the inactive X chromosome. As such, the inactive X chromosome is frequently located

209 within a small appendage protruding from the terminal lobe of female neutrophil nuclei (18, 22, 27).

210 Strong evidence that the inactive X chromosome influences the formation of these appendages comes

211 from the neutrophils of XXX or XXXX individuals which frequently exhibit two or three appendages,

212 respectively (18). How the inactive X chromosome itself, its epigenetic state or its associated factors

213 drive the formation of the nuclear appendages is unclear. However, these unique appendages do

214 suggest that chromosomes are not always simply passive during nuclear morphology changes, but can

215 under certain circumstances can influence the process.

216

217 Compared to their precursors and other immune cells, mature neutrophils are transcriptionally muted

218 (19, 28, 29). It is unclear why. Prior to our study, one possibility was that neutrophil lobing perturbed

219 the macroscale organisation of chromosomes, and thus their transcriptional activity. However,

220 consistent with examinations of fewer (between 4-7) chromosomes (18, 19), we find that chromosome

221 territories remain highly organised within neutrophil nuclear lobes. Thus, overt chromosomal

222 disorganisation does not appear to explain the relative transcriptional inactivity of neutrophils.

223

224 It has been previously reported that regions on one chromosome can influence the expression of a

225 gene on another, presumably via three-dimensional physical proximity (30). These interactions are

226 known as gene regulatory transchromosomal interactions (31). Interestingly, these interactions have

227 been predominantly reported in immune cells (6-10). However, both their function and indeed

228 existence is debated (32). Here, while we reveal relatively large numbers of transchromosomal

229 interactions in neutrophil nuclei via in situ HiC, our data demonstrate little to no preferential

230 distribution of chromosomes to neutrophil nuclear lobes (where chromosomes are physically

9 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.05.326009; this version posted October 5, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

231 constrained and could physically interact only with their lobe mates), which suggests that gene

232 regulatory transchromosomal interactions do not occur in mature human neutrophils.

233

234 The non-random radial distribution of chromosomes we have detected in neutrophil nuclear lobes is

235 common in other cell types (16) and species (33). However, the cause and possible purpose of this

236 organisation is not known. It has been proposed that the organisation could be due to the differential

237 timing of centromere separation (34) or chromosome position during mitosis (16), differential

238 interaction between chromosomes and the nuclear periphery and/or other nuclear bodies (35) or

239 simply the force of transcription from highly transcribed chromosomes acting on those lowly

240 transcribed (36). A number of purposes for the radial organisation have also been proposed, from

241 protecting the genome from viruses and mutagens using a layer of highly repetitive and

242 heterochromatic DNA (37) to providing nuclear stability or rigidity (16, 38). Given our findings in the

243 highly malleable neutrophil nuclei the latter seems unlikely. While the function of peripheral

244 heterochromatin is unclear in neutrophils or other cell types, there is one cell type in which the radial

245 position of heterochromatin has a clear function. As the retinal cells of nocturnal animals develop,

246 they invert the vast majority of their heterochromatin from its peripheral position to a central core

247 (39). This remarkable transformation endows the mature retinal cell enhanced light channelling

248 characteristics thus enhancing low light vision.

10 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.05.326009; this version posted October 5, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

249 Figure Legends

250

251 Figure 1: Novel analysis pipeline detects the position and characteristics of all human

252 chromosomes in three-dimensions. (A-C) Schematic of neutrophil isolation, chromosome paint (A),

253 spectral character of each human chromosome (B) and novel image analysis pipeline (C). (D)

254 Proportion of total chromosomes detected made up by each chromosome in human blood neutrophils.

255 Red line represents expected proportion if all 24 chromosomes were detected equally (0.04 for

256 autosomes, 0.02 for sex chromosomes). (E, F) Box and whisker plot (5th-95th percentile) of the

257 volume (µm3) (E) and surface area (µm2)(F) of each chromosome across all 240 neutrophils. Unpaired

258 T test used to compare X and Y chromosomes. (G) Plot of median chromosome volume across all

11 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.05.326009; this version posted October 5, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

259 neutrophils against chromosome linear length with straight line fitted (y=1.09e-3x+0.649, R2 = 0.359

260 and P = 0.001).

261

262

263

12 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.05.326009; this version posted October 5, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

264

13 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.05.326009; this version posted October 5, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

265 Figure 2. Human neutrophil chromosomes distribute randomly to nuclear lobes. (A) Heatmap of

266 the -log10(p-value) from chromosome lobe colocalization analysis in human blood neutrophil nuclear

267 lobes. The analysis determined if pairs of chromosomes are found colocalized within a lobe at a

268 higher rate than expected by chance (B) Top: Three-dimensional render of chromosomes within a

269 human neutrophil nucleus. Left: Three-dimensional renders of the same neutrophil and chromosomes

270 after five sets of random chromosome shuffling. Right: Heatmap of the -log10(p-value) from the

271 chromosome colocalization analysis of five independent sets of random chromosome shuffling. (C)

272 Schematic of the proximity ligation reaction central to the in situ HiC protocol, (D) Proportion of total

273 DNA-DNA interactions detected by in situ HiC and the diffHiC pipeline that occur between

274 chromosomes (transchromosomal interactions) in human immune cells. (E) Frequency of

275 transchromosomal interactions to the power of 0.25, plotted as a function of summed chromosome

276 length in human neutrophils. Straight line fitted to data with no intercept (y=8.621e-3x, R2 = 0.9683

277 and p-value <2.2e-16). (F) Heatmap of the number of transchromosomal interactions by each

278 chromosome to all others in human neutrophils.

279

14 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.05.326009; this version posted October 5, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

280

281

282

283 Figure 3. Human neutrophil chromosomes do not position randomly within nuclear lobes. (A)

284 Schematic showing how radial position values within a neutrophil nuclear lobe are calculated (B) A

285 slice through a neutrophil nuclear lobe image showing the approximate radial position values of the

286 chromosomes. For clarity most chromosomes are not shown. (C) Scatterplot of the median of the

287 mean chromosome volume radial position within human neutrophil nuclear lobes plotted as a function

288 of chromosome length (Mb). Straight line fitted to data (y=-6.18e-4x+0.735, R2 = 0.7498 and p-value

15 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.05.326009; this version posted October 5, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

289 4.6e-8). (D) Scatterplot of the median of the mean chromosome volume radial position from five

290 independent sets of random chromosome position shufflings plotted as a function of chromosome

291 length (Mb). Straight line fitted to each instance independently (y=-2.38e-5x+0.612, R2 = 0.008 and

292 p-value = not significant (ns), y=-8.47e-5x+0.612, R2 = 0.0473 and p-value ns, y=-1.24e-5x+0.618, R2

293 = 0.0023 and p-value ns, y=-2.12e-5x+0.622, R2 = 0.0066 and p-value ns, y=-7.45e-5x+0.630, R2 =

294 0.0378 and p-value ns).

295

296

297

16 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.05.326009; this version posted October 5, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

298 299

17 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.05.326009; this version posted October 5, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

300 Supp Fig 1. (A) Example purity check of enriched human blood neutrophils. (B) Number of

301 chromosomes defined within a nucleus for all neutrophil images (proportion shown). (C) Proportion

302 of total chromosomes detected made up by each chromosome in human blood neutrophils, CD4+ T

303 cells and CD8+ T cells. Data from 171 CD4+ T cells and 30 CD8+ T cells. (D) Example sort profile of

304 human blood CD4+ and CD8+ T cells. (E, F) Box and whisker plot (5th-95th percentile) showing the

305 elongation (E) and compactness (F) of chromosomes detected in human blood neutrophils. Elongation

306 is the ratio between the largest axis of a fitted 3D ellipsoid and the second largest. Compactness is the

307 normalized ratio between volume and surface. (G) Scatterplot of median chromosome volume in

308 CD8+ T cells (left) and CD4+ T cells (right) against chromosome linear length. Data fitted with a

309 straight line (y=2.94e-3x+0.621, R2 = 0.517 and p-value 7.5e-5 and y=1.41e-3x+0.588, R2 = 0.377 and

310 p-value 0.00141, respectively).

311

312

313

314

18 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.05.326009; this version posted October 5, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

315

316 Supp Fig 2 (A) Slice of a neutrophil image moving through the lobe calling pipeline, from raw image

317 to the final lobed nucleus defined using watershed analysis. (B) Further examples of lobe calling

318 pipeline results. (C) Bar plot showing the proportion of human blood neutrophils containing between

19 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.05.326009; this version posted October 5, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

319 1-7 nuclear lobes. (D) Box and whisker plot (5th-95th percentile) showing the volume of lobes as a

320 proportion of the total nuclear volume in human blood neutrophils containing between 1-6 lobes.

321

322 Methods

323

324 Ethics Statement

325

326 Collection of human blood for research studies was approved by Human Research Ethics Committees

327 of Melbourne Health and Walter and Eliza Hall Institute of Medical Research (application 88/03).

328 Written consent was obtained.

329

330 Cell isolation

331

332 Neutrophils were isolated from 2ml of healthy male donor whole blood following the EasySep

333 Human Neutrophil Enrichment kit manufacturer’s protocol). Purity was ~90%. Human T cells were

334 isolated from the PBMC layer of the remaining whole blood after density separation in Leucosep™

335 tubes containing 15 mL Ficoll-Paque, following the manufacturer’s protocol. These cells were stained

336 with TCRab-PerCP-eFluor 710 (eBioscience Cat.No. 46-9986-42), CD4-APC (BD Pharmingen

337 Cat.No. 555349), CD25-PECy7 (BD Bioscience Cat.No. 557741), CD45RA-FITC (eBioscience

338 Cat.No. 556626), CD14-PE (BioLegend Cat.No. 367104), CD16- APC-Cy7 (BD Bioscience Cat.No.

339 557758), HLA-DR-eFluor450 (eBioscience Cat.No. 48-9952-42) and CD19-BV650 (BioLegend

340 Cat.No. 302238). CD4+ T cells (CD16- CD14- TCRab+ CD4+ CD45RA+ CD25-) and CD8+ T cells

341 (CD16- CD14- TCRab+ CD4- CD45RA+ CD25-) were sorted to a purity >97%.

342

343 Chromosome paint

344

345 Sixty thousand cells were settled on a poly-L-Lysine (Sigma Aldrich Cat no. P4707) coated cover slip

346 at 37oC for 20 minutes, washed with phosphate buffer saline, before fixing with fresh 3:1

20 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.05.326009; this version posted October 5, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

347 methanol:acetic acid (glacial) for 15 minutes at 22oC. The cells were then rinsed twice with water

348 before being incubated in 2x saline-sodium citrate buffer (SSC) for 2 minutes. The cells were then

349 dehydrated in an ethanol series (75%, 85% and 100%) for 2 minutes each. 7uL of Metasystems

350 24XCyte Human multicolour FISH probes (Metasystems Cat no. D-0125-060-DI) was pre-warmed on

351 a glass slide at 37oC for 5 minutes before a further 2 minutes at 75oC with the cells added. The cells

352 were then sealed and incubated at 37oC for 18 hours in a dark, humid chamber. After hybridisation

353 was complete the cells were immersed in 72oC 0.4x SSC for 2 minutes, 2x SSC with 0.05% Tween-20

354 for 30 seconds before being sealed on a glass slide in 85% glycerol.

355

356 Confocal Microscopy

357 Imaging experiments were performed on a Zeiss 880 confocal microscope using a 63x 1.4 NA

358 objective lens. The system was run in lambda mode, recording fluorescence signal in 32 spectral

359 channels over a spectral range of 410 nm to 690 nm (in 8.9 nm increments). Samples were imaged in

360 3D using Nyquist sampling of 70 nm pixel size and z-steps of 200 nm using the 405 nm, 488 nm, 561

361 nm and 633 nm lasers to excite the samples consisting of the fluorescent labels DEAC, FITC,

362 Spectrum Orange, Texas Red and Cy5. Single colour control experiments were performed to

363 determine the spectral signatures using the above microscope settings.

364

365 Tetraspeck beads (ThermoFisher) adhered to a coverslip and mounted to a microscope slide were

366 imaged in 3D. Custom written Matlab (Mathworks, Natick, MA, USA) scripts measured the axial

367 positions of the Tetraspeck beads in each spectral channel and removed the chromatic aberrations

368 from the sample images. The five spectral signatures from the single colour controls were used to

369 linearly unmix the 32-channel spectral fluorescence signal in each voxel into the five constituent

370 fluorophores.

371 Chromosome assignment and measurement

21 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.05.326009; this version posted October 5, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

372 After channel unmixing and normalisation, voxels with similar values in the 5 channels are clustered

373 together using a 3D Simple Linear Iterative Clustering (SLIC) algorithm (40). Thresholding is then

374 applied to the 5 channels and chromosome identity is assigned according to known channel

375 combinations. The threshold is set automatically by examining a range of thresholds and select that

376 which yields the optimal number of chromosome objects (2) and the minimum number of false

377 combination objects. Chromosome measurements are performed on assigned chromosomes using

378 algorithms and tools from the 3D ImageJ suite (41). 3D Eroded Volume Fraction (42) is also

379 performed to compute the position of chromosomes within the nucleus.

380 Nuclear lobe calling

381 Nuclei boundaries are detected by summing all channels and global thresholding. The approximate

382 3D positions of the lobe centres are manually marked before a watershed method separates the

383 nucleus into lobes.

384

385 Chromosome shuffling

386 While considering the chromosomes three-dimensional character, the centre positions of all

387 chromosomes are randomly distributed within the nuclear space until all chromosomes fit and no

388 overlapping is observed between chromosomes. Due to space constraints shuffling was not possible

389 for every nucleus attempted.

390 All images were stored within an OMERO Database (43), processing and analysis were then

391 automated using the TAPAS home system (44).

392 Chromosome measurements analysis

393 Chromosome measurements were analysed with R using packages dplyr and purr. Objects not assigned

394 to a known combination of values were removed for analysis and plotting. Straight lines were fitted to

395 the data with function lm() from R.

396

22 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.05.326009; this version posted October 5, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

397 Chromosome co-localisation in lobes

398 Nuclei with less than ten chromosomes detected or just one lobe were excluded from the analysis. For

399 the remaining nuclei, the volume of each chromosome in each lobe was recorded. The co-localisation

400 score of any two chromosomes in a nucleus was quantified by Sum( pil * pjl ), where pil is the

401 proportion by volume of chromosome i in lobe l , pjl is the proportion by volume of chromosome j in

402 lobe l , and the sum is over all lobes in the nucleus. Co-localization p-values were obtained as follows.

403 Each chromosome was treated in turn as the reference chromosome. The co-localisation scores for the

404 reference chromosome with each other chromosome were ranked within each nucleus. The ranks were

405 summed across nuclei and converted to z-scores assuming uniformly distributed ranks for each

406 nucleus. The p-values were adjusted for multiple testing using the Bonferroni correction. Heatmaps of

407 the -log10(p-value) were generated using the R package gplots with the function heatmap.2.

408

409 In situ HiC

410 In situ HiC was performed as previously described (24) for one biological replicate of primary

411 human neutrophils. The library was sequenced on an Illumina NextSeq 500 to produce 81-bp paired-

412 end reads. HiC libraries for CD4+ T cells, CD8+ T cells and B cells are from GSE105776. The data

413 pre-processing and analysis was performed with the diffHic pipeline (45) in R as previously described

414 with changes in parameters (32). Where biological replicates were available, the libraries were summed

415 after pre-processing.

416

417 Chromosomal looping interactions were detected using a method similar to that described previously

418 by Rao et al. (24) and in (46). Read pairs were counted in bin pairs of 50 kbp anchors for all

419 libraries. For each bin pair, the log-fold change over the average abundance of each of several

420 neighbouring regions was computed. Neighbouring regions in the interaction space included a square

421 quadrant of sides 'x+1' that was closest to the diagonal and contained the target bin pair in its corner; a

422 horizontal stripe of length '2x+1' centred on the target bin pair; a vertical stripe of '2x+1', similarly

423 centred; and a square of sides '2x+1', also containing the target bin pair in the centre. The enrichment

424 value for each bin pair was defined as the minimum of these log-fold changes, i.e., the bin pair had to

23 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.05.326009; this version posted October 5, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

425 have intensities higher than all neighbouring regions to obtain a large enrichment value. The

426 neighbourhood counts for the libraries at 50 kbp bin size were computed with the neighborCounts

427 function from the diffHic package with flank (x) 5 bin sizes (i.e., 250 kbp) and enrichment values

428 were determined with filterPeaks function with get.enrich=TRUE. Looping interactions were then

429 filtered with filterPeaks. Loops were defined as those with enrichment values above 1, were more

430 than 100 kbp from the diagonal and with minimum count greater than 10 for all libraries except the

431 neutrophils which used a minimum count of 5 to accounts for differences in library size. Directly

432 adjacent loops in the interaction space were aggregated into clusters to a maximum cluster size of 500

433 kbp using the clusterPairs function from the csaw package v1.18.0 (47). Blacklisted genomic regions

434 were obtained from ENCODE for hg38 (48). Loops that that had at least one anchor in a blacklisted

435 genomic region were removed.

436

437 Heatmaps of the loops between chromosomes where generated using the R package gplots with the

438 function heatmap.2. The frequency of transchromosomal interactions to the power of 0.25 was plotted

439 as a function of the sum of the chromosome lengths. A linear model was fitted to data with the lm

440 function.

441

442 Data availability

443 Sequence data that support the findings of this study are tabulated in the supplementary tables and are

444 available in the GEO database under accession number GSE157229.

445

446

24 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.05.326009; this version posted October 5, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

447 1. Rabl C. Uber Zelltheilung. Morphol Jahrb. 1885;10:214-330. 448 2. Cremer T, Cremer M. Chromosome territories. Cold Spring Harb Perspect Biol. 449 2010;2(3):a003889. 450 3. Stadler S, Schnapp V, Mayer R, Stein S, Cremer C, Bonifer C, et al. The architecture of 451 chicken chromosome territories changes during differentiation. BMC Cell Biol. 2004;5(1):44. 452 4. Finlan LE, Sproul D, Thomson I, Boyle S, Kerr E, Perry P, et al. Recruitment to the nuclear 453 periphery can alter expression of in human cells. PLoS Genet. 2008;4(3):e1000039. 454 5. Agmon N, Liefshitz B, Zimmer C, Fabre E, Kupiec M. Effect of nuclear architecture on the 455 efficiency of double-strand break repair. Nat Cell Biol. 2013;15(6):694-9. 456 6. Hewitt SL, Farmer D, Marszalek K, Cadera E, Liang HE, Xu Y, et al. Association between 457 the Igk and Igh immunoglobulin loci mediated by the 3' Igk enhancer induces 'decontraction' of the 458 Igh locus in pre-B cells. Nat Immunol. 2008;9(4):396-404. 459 7. Kim LK, Esplugues E, Zorca CE, Parisi F, Kluger Y, Kim TH, et al. Oct-1 regulates IL-17 460 expression by directing interchromosomal associations in conjunction with CTCF in T cells. Mol 461 Cell. 2014;54(1):56-66. 462 8. Ling JQ, Li T, Hu JF, Vu TH, Chen HL, Qiu XW, et al. CTCF mediates interchromosomal 463 colocalization between Igf2/H19 and Wsb1/Nf1. Science. 2006;312(5771):269-72. 464 9. Spilianakis CG, Lalioti MD, Town T, Lee GR, Flavell RA. Interchromosomal associations 465 between alternatively expressed loci. Nature. 2005;435(7042):637-45. 466 10. Zhao Z, Tavoosidana G, Sjolinder M, Gondor A, Mariano P, Wang S, et al. Circular 467 chromosome conformation capture (4C) uncovers extensive networks of epigenetically regulated 468 intra- and interchromosomal interactions. Nat Genet. 2006;38(11):1341-7. 469 11. Brianna Caddle L, Grant JL, Szatkiewicz J, van Hase J, Shirley BJ, Bewersdorf J, et al. 470 Chromosome neighborhood composition determines translocation outcomes after exposure to high- 471 dose radiation in primary cells. Chromosome Res. 2007;15(8):1061-73. 472 12. Parada LA, McQueen PG, Misteli T. Tissue-specific spatial organization of genomes. 473 Genome Biol. 2004;5(7):R44. 474 13. Parada LA, McQueen PG, Munson PJ, Misteli T. Conservation of relative chromosome 475 positioning in normal and cancer cells. Curr Biol. 2002;12(19):1692-7. 476 14. Roix JJ, McQueen PG, Munson PJ, Parada LA, Misteli T. Spatial proximity of translocation- 477 prone gene loci in human lymphomas. Nat Genet. 2003;34(3):287-91. 478 15. Khalil A, Grant JL, Caddle LB, Atzema E, Mills KD, Arneodo A. Chromosome territories 479 have a highly nonspherical morphology and nonrandom positioning. Chromosome Research. 480 2007;15(7):899-916. 481 16. Bolzer A, Kreth G, Solovei I, Koehler D, Saracoglu K, Fauth C, et al. Three-dimensional 482 maps of all chromosomes in human male fibroblast nuclei and prometaphase rosettes. PLoS Biol. 483 2005;3(5):e157. 484 17. P Ehrlich AL. Die Anæmie, Nothnagel's Specielle Pathologie und Therapie, 1900. 485 18. Aquiles Sanchez J, Karni RJ, Wangh LJ. Fluorescent in situ hybridization (FISH) analysis of 486 the relationship between chromosome location and nuclear morphology in human neutrophils. 487 Chromosoma. 1997;106(3):168-77. 488 19. Hubner B, Lomiento M, Mammoli F, Illner D, Markaki Y, Ferrari S, et al. Remodeling of 489 nuclear landscapes during human myelopoietic cell differentiation maintains co-aligned active and 490 inactive nuclear compartments. Epigenetics Chromatin. 2015;8:47. 491 20. Yerle-Bouissou M, Mompart F, Iannuccelli E, Robelin D, Jauneau A, Lahbib-Mansais Y, et 492 al. Nuclear architecture of resting and LPS-stimulated porcine neutrophils by 3D FISH. Chromosome 493 Res. 2009;17(7):847-62. 494 21. Hepperger C, Mayer A, Merz J, Vanderwall DK, Dietzel S. Parental genomes mix in mule 495 and human cell nuclei. Chromosoma. 2009;118(3):335-47. 496 22. Karni RJ, Wangh LJ, Sanchez JA. Nonrandom location and orientation of the inactive X 497 chromosome in human neutrophil nuclei. Chromosoma. 2001;110(4):267-74. 498 23. Bartova E, Kozubek S, Jirsova P, Kozubek M, Lukasova E, Skalnikova M, et al. Higher-order 499 chromatin structure of human granulocytes. Chromosoma. 2001;110(5):360-70.

25 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.05.326009; this version posted October 5, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

500 24. Rao SS, Huntley MH, Durand NC, Stamenova EK, Bochkov ID, Robinson JT, et al. A 3D 501 map of the at kilobase resolution reveals principles of chromatin looping. Cell. 502 2014;159(7):1665-80. 503 25. Zhu Y, Gong K, Denholtz M, Chandra V, Kamps MP, Alber F, et al. Comprehensive 504 characterization of neutrophil genome topology. Genes & Development. 2017;31(2):141-53. 505 26. Skinner BM, Johnson EE. Nuclear morphologies: their diversity and functional relevance. 506 Chromosoma. 2017;126(2):195-212. 507 27. Hoffmann K, Sperling K, Olins AL, Olins DE. The granulocyte nucleus and lamin B receptor: 508 avoiding the ovoid. Chromosoma. 2007;116(3):227-35. 509 28. Gemelli C, Montanari M, Tenedini E, Zanocco Marani T, Vignudelli T, Siena M, et al. 510 Virally mediated MafB transduction induces the monocyte commitment of human CD34+ 511 hematopoietic stem/progenitor cells. Cell Death Differ. 2006;13(10):1686-96. 512 29. Newburger PE, Subrahmanyam YV, Weissman SM. Global analysis of neutrophil gene 513 expression. Curr Opin Hematol. 2000;7(1):16-20. 514 30. Williams A, Spilianakis CG, Flavell RA. Interchromosomal association and gene regulation 515 in trans. Trends Genet. 2010;26(4):188-97. 516 31. Kioussis D. Gene regulation: kissing chromosomes. Nature. 2005;435(7042):579-80. 517 32. Johanson TM, Coughlan HD, Lun ATL, Bediaga NG, Naselli G, Garnham AL, et al. 518 Genome-wide analysis reveals no evidence of trans chromosomal regulation of mammalian immune 519 development. PLoS Genet. 2018;14(6):e1007431. 520 33. Tanabe H, Muller S, Neusser M, von Hase J, Calcagno E, Cremer M, et al. Evolutionary 521 conservation of chromosome territory arrangements in cell nuclei from higher primates. Proc Natl 522 Acad Sci U S A. 2002;99(7):4424-9. 523 34. Gerlich D, Beaudouin J, Kalbfuss B, Daigle N, Eils R, Ellenberg J. Global chromosome 524 positions are transmitted through mitosis in mammalian cells. Cell. 2003;112(6):751-64. 525 35. Cremer T, Cremer C. Chromosome territories, nuclear architecture and gene regulation in 526 mammalian cells. Nat Rev Genet. 2001;2(4):292-301. 527 36. Yin H, Wang MD, Svoboda K, Landick R, Block SM, Gelles J. Transcription against an 528 applied force. Science. 1995;270(5242):1653-7. 529 37. Hsu TC. A possible function of constitutive heterochromatin: the bodyguard hypothesis. 530 Genetics. 1975;79 Suppl:137-50. 531 38. Nava MM, Miroshnikova YA, Biggs LC, Whitefield DB, Metge F, Boucas J, et al. 532 Heterochromatin-Driven Nuclear Softening Protects the Genome against Mechanical Stress-Induced 533 Damage. Cell. 2020;181(4):800-17 e22. 534 39. Solovei I, Kreysing M, Lanctot C, Kosem S, Peichl L, Cremer T, et al. Nuclear architecture of 535 rod photoreceptor cells adapts to vision in mammalian evolution. Cell. 2009;137(2):356-68. 536 40. Tran Thi Nhu H, Arrojo EDR, Berggren PO, Boudier T. A novel toolbox to investigate tissue 537 spatial organization applied to the study of the islets of Langerhans. Sci Rep. 2017;7:44261. 538 41. Ollion J, Cochennec J, Loll F, Escude C, Boudier T. TANGO: a generic tool for high- 539 throughput 3D image analysis for studying nuclear organization. Bioinformatics. 2013;29(14):1840-1. 540 42. Ballester M, Kress C, Hue-Beauvais C, Kieu K, Lehmann G, Adenot P, et al. The nuclear 541 localization of WAP and CSN genes is modified by lactogenic hormones in HC11 cells. J Cell 542 Biochem. 2008;105(1):262-70. 543 43. Allan C, Burel JM, Moore J, Blackburn C, Linkert M, Loynton S, et al. OMERO: flexible, 544 model-driven data management for experimental biology. Nat Methods. 2012;9(3):245-53. 545 44. Whitehead L, Wimmer, V., Lafouresse, F., Ratnayake, D., Currie, P., Groom, J., Rogers, K. 546 and Boudier, T. Towards an Automated Processing and Analysis System for multi-dimensional light- 547 sheet microscopy big data using ImageJ and OMERO. International Microscopy Congress IMC 548 192018. 549 45. Lun AT, Smyth GK. diffHic: a Bioconductor package to detect differential genomic 550 interactions in Hi-C data. BMC Bioinformatics. 2015;16:258. 551 46. Johanson TM, Lun ATL, Coughlan HD, Tan T, Smyth GK, Nutt SL, et al. Transcription- 552 factor-mediated supervision of global genome architecture maintains B cell identity. Nat Immunol. 553 2018;19(11):1257-64.

26 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.05.326009; this version posted October 5, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

554 47. Lun AT, Smyth GK. csaw: a Bioconductor package for differential binding analysis of ChIP- 555 seq data using sliding windows. Nucleic Acids Res. 2016;44(5):e45. 556 48. Consortium EP. An integrated encyclopedia of DNA elements in the human genome. Nature. 557 2012;489(7414):57-74. 558

559

27 bioRxiv preprint doi: https://doi.org/10.1101/2020.10.05.326009; this version posted October 5, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY 4.0 International license.

560 Financial Disclosure Statement 561

562 This work was supported by grants and fellowships from the National Health and Medical Research

563 Council of Australia (T.J. #1124081, R.A. and T.J. #1049307, #1100451, G.S. #1058892, R.A. and

564 G.S. #1158531, C.K. #1125436, D.H #1113577, L.H. #1080887 and #1037321). H.C. was supported

565 by a H.C. Marian and E.H. Flack Fellowship. This study was made possible through Victorian State

566 Government Operational Infrastructure Support and Australian Government NHMRC Independent

567 Research Institute Infrastructure Support scheme. The funders had no role in study design, data

568 collection and analysis, decision to publish, or preparation of the manuscript.

569

570 Acknowledgments

571

572 The authors thank S. Johanson for constructive discussions, and the volunteers at the Walter and Eliza

573 Hall Volunteer Blood Donor Registry for their donation.

574

28