Comparison of ultracentrifugation and a commercial kit for isolation of exosomes derived from glioblastoma and breast cancer cells

Exosomes are small extracellular vesicles around 30-100 nm in diameter that are secreted from cells and can be found in most body fluids. Exosomes can be a vital source of biomarkers as they contain various substances (e.g. lipids, RNAs, metabolites and proteins) that can reflect the cell of origin (e.g. cancer cells). For isolation of exosomes present in biological matrices, ultracentrifugation (UC)-based procedures are most common. Other approaches exist, including commercial kits developed for easy and low sample volume isolation. In this study, differential UC and an isolation kit from a major vendor (Total Exosome Isolation Reagent from Thermo Fisher Scientific) were compared. Exosomes were isolated from cell culture media of two different cell sources (patient derived cells from glioblastoma multiforme and the breast cancer cell line MDA-MB-231). For both isolation methods, transmission electron microscopy, dynamic light scattering and western blotting indicated the presence of exosomes. The kit- and UC isolates contained similar amounts of protein measured by the bicinchoninic acid (BCA) assay with absorbance at 562 nm. Using western blot, positive exosome markers were identified in all isolates, and additional exosome markers were identified using MS-based proteomics. For the glioblastoma exosome isolates, the number of proteins identified with liquid chromatography tandem MS (LC-MS/MS) was higher for the UC isolates than the kit isolates when injecting equal protein amounts, contrary to that for the breast cancer exosome isolates. However, negative exosome markers were also found in glioblastoma isolates using LC-MS/MS. Thus, we would not use the term “exosome isolation” as impurities may be present with both isolation methods. Notably, potential biomarkers for both diseases were identified in the isolates using LS-MS/MS. In our opinion, the two isolation methods had rather similar performance, although with some minor differences based on cell of origin.


0
The protein amount per million cells (hereafter referred to as protein amount) in the BC-( Figure  4 0 1 1A) and GBM ( Figure 1B) isolates was measured using UV-Vis spectrophotometry (with 4 0 2 absorbance at 562 nm). The total protein amount measured for kit isolates was 15-28 times higher 4 0 3 than for UC isolates. A higher protein amount in exosomes isolated by the kit compared to that 4 0 4 by UC were also observed in a study by Van Deun et al., who compared UC to the same isolation 4 0 5 kit used in the present study for MCF7 derived exosomes (57). However, the measured 4 0 6 absorbance in the kit blanks (i.e. cell culture medium grown without cells and isolated by kit) was 4 0 7 high in comparison to UC blanks (i.e. cell culture medium grown without cells and isolated by 4 0 8 UC), where the absorbance was below the limit of quantification. The high absorbance in the kit 4 0 9 blanks could indicate protein contaminations. When correcting for the blank (subtracting the 4 1 0 protein amount measured in blank samples from the protein amount in exosome isolates), the 4 1 1 measured protein content for exosomes isolated by the kit and UC was similar. 4 1 2 4 1 3

TEM and DLS detected vesicles in the expected size range for exosomes 4 1 4
Morphological analysis of the exosome samples was performed using TEM and immunogold 4 1 5 labelling of CD9. In addition, the hydrodynamic particle size distribution was measured using 4 1 6 DLS analysis. Clusters of vesicles were observed in the micrographs of the samples isolated with 4 1 7 both kit and UC (Figure 2). Vesicle structures similar to that described in literature were 4 1 8 observed (6,58,59). The DLS experiments disclosed the coexistence of two populations of 4 1 9 moieties, single entities and clusters, both with a narrow size distribution. 4 2 0 4 2 1

GBM exosomes 4 2 2
No CD9-labelling was observed for the vesicle structures observed in the GBM isolates ( Figure  4 2 3 2AI and 2AIII) and the presence of a membrane enclosing the vesicles could not be confirmed. 4 2 4 Compared to the kit isolates, the UC isolates presented more distinct double membranes in the 4 2 5 expected size range for exosomes. The blank samples for both isolation methods did not display 4 2 6 membrane structures (Figure 2AII and 2AIV). The absence of vesicles was further confirmed by 4 2 7 DLS analysis of the UC blank ( Figure 2B). The DLS-analysis of the GBM isolates exhibited 4 2 8 particles of similar sizes of 51 and 73 nm (mean) with both isolation methods ( Figure 2B). Thus, 4 2 9 both isolation methods gave rise to comparable exosome populations. 4 3 0 4 3 1

BC exosomes 4 3 2
Several of the BC vesicle structures were CD9-labelled ( Figure 2CI and 2CIII). CD9-labelled 4 3 3 vesicles have also been observed in a previous study of the same cell line (60). Notably, the blank 4 3 4 isolates displayed contamination ( Figure 2CII and 2CIV), e.g. exosome-resembling vesicles 4 3 5 were found in the UC blank (red dashed circles). However, no contaminations were found in the 4 3 6 UC blank using DLS, while the kit blank displayed 67 nm (mean) contaminations ( Figure 2D). 4 3 7 The DLS analysis also presented two distinct particle diameters in kit isolates (28 and 95 nm, 4 3 8 mean values) while only one particle diameter was present in UC isolates (137 nm, mean value), 4 3 9 indicating differences in the particle sizes isolated with the two isolation methods. The sizes observed with DLS correlates well with that found in other studies (30-250 nm) (13, 4 4 2 57, 61-65). In conclusion, the isolates showed structures resembling those of EVs, but some 4 4 3 blank were not entirely devoid of vesicles or particles. Observations made with TEM are not 4 4 4 necessarily detectable with DLS because TEM analyses dry material, whereas DLS measures on 4 4 5 solutions or suspensions of particles. In addition, the micrographs taken with TEM display a 4 4 6 narrow section of the grid, which again represents only a small part of the isolate. According to ISEV, for characterization of exosomes at least three exosome markers should be 4 5 1 included; transmembrane proteins (e.g. tetraspanins), cytosolic proteins (e.g. TSG101 or 4 5 2 annexins) and negative markers (e.g. calnexin) (45). In the present study, WB was performed 4 5 3 using antibodies for a selection of positive exosome markers (the tetraspanins CD81, CD9 and 4 5 4 CD63, TSG101 and flotillin-1). Calnexin was selected as a negative marker for purity evaluation 4 5 5 as recommended by ISEV. This protein is located at the endoplasmic reticulum (ER) and has 4 5 6 been absent in exosome samples in some studies (45,61). Hence, the presence of calnexin is 4 5 7 assumed to signalize ER-contamination. Thus, contaminations from other cellular organelles 4 5 8 cannot be excluded. For the GBM cells and exosomes, positive and negative exosome markers were detected in 4 6 2 isolates from both kit-and UC. The positive marker CD81 was only found in the UC isolate from 4 6 3 the first batch (Figure 3). The WB-bands were also more apparent for most positive markers for 4 6 4 exosomes isolated by UC (lower protein amount loaded than for the kit isolates), and thus is in 4 6 5 accordance with the study of Van Deun et al. (57). The kit isolate bands were also circular, which 4 6 6 implies higher detection uncertainty. 4 6 7 4 6 8 3.3.2 BC exosomes 4 6 9 For the BC cells and exosomes, inconsistency on the presence of several positive exosome 4 7 0 markers were observed between the kit and UC isolates (Figure 3) The reason for the variation in tetraspanin appearance in the BC kit-and UC isolates could be due 4 7 8 to protein concentrations below detection limits or poor antibody quality (see Figure 3). Several 4 7 9 antibodies for CD63 and CD81 (different batch number/catalog number) were tested for the BC 4 8 0 isolates before a signal was obtained (signal obtained for CD81 using catalog number 10630D), 4 8 1 and this could indicate poor antibody quality. On the other hand, the WB was performed under 4 8 2 reducing conditions. When the epitope binds to cysteine-conserved protein domains (i.e. 4 8 3 tetraspanins), performing WB under non-reducing conditions is more commonly selected. The 4 8 4 stronger signals for the kit isolates from BC could be due to the higher loaded protein amount. 4 8 5 Nevertheless, the presence of positive markers indicates the presence of exosomes in the isolates 4 8 6 obtained using both methods. The absence of calnexin in BC exosomes from both isolation 4 8 7 9 exclusively from cell impurities, or if they occur naturally in EVs. To summarize, from our point 5 4 0 of view, complete information about exosome purity cannot be obtained by any of the common 5 4 1 characterization techniques used today, and one can argue that the term "exosome isolation" can 5 4 2 be misleading. The total number of proteins identified in the GBM and BC isolates using LC-MS/MS is 5 4 7 presented in the Venn diagrams in Figure 6 (see Supplemental Proteins for a list of all 5 4 8 identified proteins). For the GBM isolates, the number of identified proteins reflects the findings 5 4 9 in both WB and LC-MS/MS exosome marker investigations. UC isolates provided more unique 5 5 0 proteins than the kit isolates (75 % higher number of identified proteins). An increased number of 5 5 1 potential biomarkers for GBM (e.g. heat shock proteins 70 kDa and 90 kDa (71-73), chondroitin 5 5 2 sulfate proteoglycan 4 (71, 74), CD44 (71,74,75) and CD276 (76)) were also identified in the 5 5 3 UC isolates compared to the kit isolates using LC-MS/MS. The identification of relevant 5 5 4 biomarkers is of great interest for further studies on exosomes. However, the identified 5 5 5 biomarkers cannot exclusively be related to exosomes due to the presence of negative exosomes 5 5 6 markers indicating cellular contaminations with both isolation methods. 5 5 7 5 5 8 For the BC exosomes, the opposite was observed; kit isolates provided 12 % higher number of 5 5 9 identified proteins than UC isolates. However, there was no correlation between the injected 5 6 0 protein amount or the starting volume used for isolation, and the number of identified proteins 5 6 1 with kit or UC for the BC exosomes (result not shown). Thus, the reason for the variation in the 5 6 2 number of identified proteins between the two cell sources and isolation methods is unknown.

6 3
The identification of biomarkers related to triple negative breast cancer (e.g. histone H4 (77), heat 5 6 4 shock 90 kDa α and β protein (78), calmodulin and epithermal growth factor receptor (79)) was 5 6 5 similar for both isolation methods (see Supplemental Proteins). 5 6 6 5 6 7 When comparing cell sources, the number of identified proteins was lower in GBM isolates than 5 6 8 BC isolates, but the number of identified proteins for GBM isolates is comparable to another LC-5 6 9 MS/MS study on GBM exosomes (80). 5 7 0 5 7 1 3.7 Choosing the proper exosome isolation method is not straight forward 5 7 2 A complete comparison of the characteristics of the two exosome isolation methods is given in 5 7 3 Table 2. For all isolates, the kit and UC isolates displayed similarities and differences. For the GBM exosomes, one of the positive markers detected in the UC isolates (CD81) was not 5 7 7 found in kit isolates by WB. In TEM, double membrane structures were more defined in the UC 5 7 8 isolates, but the existence of double membranes cannot be excluded by looking at the 5 7 9 micrographs from the kit isolates. The largest differences between the two isolation methods for 5 8 0 the GBM exosomes were found by the LC-MS/MS studies (positive markers and number of 5 8 1 identified proteins). All tetraspanins investigated were identified in the UC isolates in several 5 8 2 replicates. In the kit isolates, CD81 was not found, and the detected tetraspanins (CD63 and CD9) 5 8 3 were only found in one replicate each. A larger number of proteins and biomarker candidates 5 8 4 were also identified in the UC isolates compared to kit isolates. However, the negative marker 5 8 5 calnexin was detected in more replicates for the UC than the kit using LC-MS/MS. In total, from 5 8 6 For the BC exosomes, there was a slight difference in favor of the kit method regarding the 5 9 0 number of positive markers found by WB and the number of identified proteins (LC-MS/MS).

9 1
However, using LC-MS/MS, more positive protein markers were found in the UC isolates in 5 9 2 contrary to what was found by WB. For the UC isolates, TEM presented double membrane 5 9 3 structures with more CD9-labelling. However, the micrograph displays an extremely small part 5 9 4 of the whole sample. The isolation methods also performed similarly regarding biomarker 5 9 5 identifications. Thus, for BC exosome isolation there is no obvious reason for choosing one 5 9 6 method over the other, even though there were some differences in the characteristics (i.e. the 5 9 7 identified protein content-and amount, CD9-labelled vesicles, particle sizes) of the isolated 5 9 8 exosomes by kit and UC. 5 9 9 6 0 0 The sample volume (e.g. of cell culture medium) and number of samples should also be taken 6 0 1 into consideration when choosing the proper isolation method. For the UC isolation, higher 6 0 2 starting volumes can be used compared to isolation with kit, while the kit are more compatible 6 0 3 with lower starting volumes (81). The high cost of ultracentrifuges has larger impact when a 6 0 4 smaller number of samples are to be isolated with UC. On the other hand, larger sample numbers 6 0 5 increase the cost for kit isolations due to reagent consumption. The observations made in our study (summarized in Table 2) support the view that exosome 6 0 9 isolation depends on the isolation protocol used, differences in the behavior of exosomes between 6 1 0 cell sources, characterization methods and the conditions applied (82). Hence, we suggest that the 6 1 1 application area (e.g. determine exosome purity or for biomarker discovery) and sample volumes 6 1 2 available for the exosome isolation should be strong determining factors when selecting the 6 1 3 proper isolation method. The characterization methods used in this study are not able to 6 1 4 distinguish exosomes from cellular contaminations and other vesicles, but the untargeted 6 1 5 proteome analyses using LC-MS/MS provided more extensive and versatile information on the 6 1 6 protein content of the samples than targeted WB of a few proteins. Consequently, we suggest that 6 1 7 LC-MS/MS should be implemented to a higher extent regarding exosome characterization. 6 1 8 Considering our findings, it is important to state that the term "exosome enrichment" is more 6 1 9 appropriate than "exosome isolation". This work was supported by the Department of Chemistry, University of Oslo, Norway. We 6 2 3 would like to acknowledge DIATECH@UiO, since parts of this work have been carried out 6 2 4 within this strategic research initiative at the Faculty of Mathematics and Natural Sciences, 6 2 5 University of Oslo. This work has also been supported by the UiO:Life Science funded 6 2 6 convergence environment "Organ on a chip and nano-devices".      K  a  l  r  a  ,  H  .  ,  S  i  m  p  s  o  n  ,  R  .  J  .  ,  J  i  ,  H  .  ,  A  i  k  a  w  a  ,  E  .  ,  A  l  t  e  v  o  g  t  ,  P  .  ,  A  s  k  e  n  a  s  e  ,  P  .  ,  B  o  n  d  ,  V  .  C  .  ,  B  o  r  r  a  s  ,  7  8  4  F  .  E  .  ,  B  r  e  a  k  e  f  i  e  l  d  ,  X  .  ,  B  u  d  n  i  k  ,  V  .  ,  B  u  z  a  s  ,  E  .  ,  C  a  m  u  s  s  i  ,  G  .  ,  C  l  a  y  t  o  n  ,  A  .  ,  C  o  c  u  c  c  i  ,  E  .  ,  F  a  l  c  o  n  -P  e  r  e  z  ,  J  .  7  8  5  M  .  ,  G  a  b  r  i  e  l  s  s  o  n  ,  S  .  ,  G  h  o  ,  Y  .  S  .  ,  G  u  p  t  a  ,  D  .  ,  H  a  r  s  h  a  ,  H  .  C  .  ,  H  e  n  d  r  i  x  ,  A  .  ,  H  i  l  l  ,  A  .  F  .  ,  I  n  a  l  ,  J  .  M  .  ,  J  e  n  s  t  e  r  ,  7  8  6  G  .  ,  K  r  a  m  e  r  -A  l  b  e  r  s  ,  E  .  M  .  ,  L  i  m  ,  S  .  K  .  ,  L  l  o  r  e  n  t  e  ,  A  .  ,  L  o  t  v  a  l  l  ,  J  .  ,  M  a  r  c  i  l  l  a  ,  A  .  ,  M  i  n  c  h  e  v  a  -N  i  l  s  s  o  n  ,  L  .  ,  7  8  7  N  a  z  a  r  e  n  k  o  ,  I  , , a n d 7  e  x  p  r  e  s  s  i  o  n  o  f  D  N  A  m  e  t  h  y  l  t  r  a  n  s  f  e  r  a  s  e  1  ,  S  u  v  4  -2  0  h  2  h  i  s  t  o  n  e  m  e  t  h  y  l  t  r  a  n  s  f  e  r  a  s  e  a  n  d  m  e  t  h  y  l  -8  7  0  b  i  n  d  i  n  g  p  r  o  t  e  i  n  s  .   C  a  n  c  e  r  B  i  o  l  o  g  y  &  T  h  e  r  a  p  y   5  ,  6  5  -7  0  8  7  1  7  8  .  B  e  l  i  a  k  o  f  f  ,  J  .  ,  a  n  d  W  h  i  t  e  s  e  l  l  ,  L  .  (  2  0  0  4  )  H  s  p  9  0  :  a  n  e  m  e  r  g  i  n  g  t  a  r  g  e  t  f  o  r  b  r  e  a  s  t  c  a  n  c  e  r  8  7  2 t h e r a p y .
A n t i -C a n c e r D r u g s