Mitochondria in human acute myeloid leukemia cell lines have ultrastructural alterations linked to deregulation of their respiratory proles

Mitochondria are not only essential for cell metabolism and energy supply but they are also engaged in calcium homeostasis, reactive oxygen species generation and play a key role in apoptosis. As a consequence, functional mitochondria disorders are involved in many human cancers including acute myeloid leukemia (AML). However, very little data are available about the deregulation of their number and/or shape in leukemic cells, despite the evident link between ultrastructure and function. In this context, we analyzed the ultrastructural mitochondrial parameters (number per cell, mitochondria area, number of cristae/mitochondria, cristae thickness) in ve leukemia cell lines (HEL, HL60, K562, KG1 and OCI-AML3) together with the functional assay of their respiratory prole. First of all, we show signicant differences within basal respiration, maximal respiration, ATP production and spare respiratory capacity between our cell lines, conrming the various respiratory proles between leukemia subtypes. Second, we highlight that these variations were obviously associated with signicant inter-leukemia heterogeneity of the number and/or shape of mitochondria. For instance, KG1 characterized by the lowest number of mitochondria together with reduced cristae diameter displayed a very particularly decient respiratory prole. In comparison, HEL and K562, both cell lines with high respiratory proles, harbored the highest number of mitochondria/cells with high cristae diameters. We show the leukemia lines present ultrastructural alterations of their mitochondria likely to impact the regulatory pathways of cell mortality, such as the process of mitophagy or calcium homeostasis. Indeed, a signicant disparity in the presence of Mitochondrial-derived vesicles (MDVs) precursors among AML cell lines, suggesting that leukemic cells displayed alteration of mitophagy, is also shown. For instance, few MDV precursors were observed in K562, carrying ASXL1mutation. Moreover, HL60 carried high levels of matrix granules and Mitochondria-associated Endoplasmic Reticulum membranes (MAMs) both implicated in calcium-dependent


Background
Acute myeloid leukemia (AML) is characterized by a clonal proliferation of myeloid precursors which fail to differentiate into more mature hematopoietic cells. Compared to normal cells, leukemic cells are shown to have a lower respiratory chain complex activity and a lower spare reserve capacity in the respiratory chain balanced by an increase in mitochondrial biogenesis 1 . Moreover, some disparities exist in mitochondria functionalities between subtypes of AML. For instance, oncogenic mutations IDH1 can modify the mitochondrial metabolism in a speci c way linked to the production of 2-hydroxyglutarate, an onco-metabolite 2 . Parallel to these functional deregulations, mitochondrial defects such as mitochondrial DNA mutations, enzymes deregulations and morphological alterations have been reported in cancer 3 .
From a morphological point of view, the mitochondrion is an organelle constituted by an outer and inner membrane. The internal membrane forms some folds towards the interior space, which are called crista.
The area enclosed by the internal membrane is named the mitochondrial matrix. However, the mitochondria, depending on tissue location and the constant communication between the cellular host and the mitochondria, have considerable variations in their size and morphology. Indeed, this organelle is essential for cell metabolism and energy supply through tricarboxylic acid cycle as well as in calcium homeostasis, reactive oxygen species generation and apoptosis. Interestingly, mitochondrial shape highly re ect respiratory e ciency 4 and cristae remodeling also occurs during apoptosis 5 . In solid cancers, excessive ssion and reduced fusion appears as a hallmark of many tumors 6 . Unless not speci c to any neoplasm, the electron features of mitochondria reveal intrinsic properties of the cancer cells and depends on microenvironment and nutrition support 7 . Furthermore, the mitochondria network may determine the cell sensitivity to doxorubicin though a 3D quantitative and ultrastructural analysis of colon carcinoma cells 8 . In hematological malignancies, mitochondrial dynamics contribute to AML stemness and differentiation throught its regulation by FIS1. The FIS1 knockdown decreases the mitochondrial ssion and mitophagy, leading to decreased leukemic stemness and increased AML differentiation, but this does not happen in normal hematopoietic stem cells 9 .
Mitochondria-Associated ER Membranes (MAM) are specialized areas of the rough endoplasmic reticulum in physical contact with the outer mitochondrial membrane that support the communication process and regulate the mitochondria function. They play a role in the modulation of the Ca2 + homeostasis in the cancer cells 10 . In leukemia, targeting this endoplasmic reticulum-mitochondria interface potentializes the cytostatics effect by creating an endoplasmic reticulum stress, releasing calcium, decreasing mitochondrial potential and leading to apoptosis 11 . Altogether, decrypting mitochondrial network appears as an essential prerequisite to develop therapies for AML.
In this context, we carried both functional and ultrastructural analysis of the mitochondria from ve leukemia cell lines (HEL, HL60, K562, KG1, OCI-AML3). Mitochondrial respiration, reactive oxygen species (ROS) production and ultrastructural parameters appears complement in de ning mitochondrial tness.
Altogether, this functional and ultrastructural analysis provided new original data concerning mitochondria in AMLs, which will help to understand their role in the disease.

Results
2.1 Mitochondrial function in AML cell lines. Figure 1A represents the respiratory pro les and the key parameters of mitochondrial respiration in the ve leukemia cell lines. The basal respiration (BR) differs between all the cell lines. KG1 has signi cantly lower BR than HEL and K562 (respectively BR = 80.1 pmol/min for KG1, BR = 66.9 pmol/min for HEL and BR = 233.5 pmol/min for K562, Kruskal-Wallis test, Fig. 1B). Further, signi cant differences were observed concerning the maximal respiration (MR) reached by cells when they were stimulated by the uncoupling agent FCCP: KG1 have a lower MR than K562, HL60 and HEL (Fig. 1C). Concerning the spare respiration capacity (SRC) reserve which indicates the ability of cells to respond to an increased energy demand, we have shown that K562 has a higher SRC % than OCI-AML3 and KG1 (Fig. 1D). Contrarily, KG1 had the lowest SRC%. Finally, concerning mitochondrial ATP synthesis, KG1 also displayed a lower ATP production than HEL, K562 and OCI-AML3 (Fig. 1E). As mitochondria are one of the main producers of reactive oxygen species (together with NADPH oxidases), we measure the production of superoxide ion through the uorimetric quantity of DHE by ow cytometry. In basal condition, KG1 signi cantly produces less ROS than K562, HL60 and HEL ( Fig. 2A). Interestingly, under stressful conditions with rotenone and antimycin A (AMA/rot), respectively complex 1 and 3 inhibitors, K562 presented a higher level of ROS emission (Fig. 2B).

Quanti cation of the mitochondrial ultrastructural parameters
Since the important mitochondrial pleiomorphism could generate misinterpretation 7 , the number of mitochondria and their morphological features (the mitochondrial area, the cristae number per mitochondrion and the cristae diameter) were quanti ed as explain in supplementary data Fig. 1. First, the number of mitochondria in KG1 was signi cantly lower than in HEL and HL60. On the opposite, the number of mitochondria was signi cantly higher in K562 compared to KG1 and OCI-AML3 (Fig. 3A). The measurement of the mitochondrial areas showed that K562 displayed signi cantly smaller mitochondria than all the other leukemia cell lines (Fig. 3B). In the same way, K562 carried a lower number of cristae per mitochondrion compared to HEL, HL60 and OCI-AML3 (Fig. 3C). Concerning cristae diameters, HEL showed the largest ones whereas KG1 displayed the smallest ones. K562 and OCI-AML3 presented larger cristae than HL60 ones (Fig. 3D). Morphologically, despite being few, K562s'cristae appeared messy or curved to the inner membrane (supplementary data Fig. 2).

Mitophagy activities in leukemia cell lines through mitochondrial-derived vesicle precursors
The mitochondria network is a dynamic compartment and mitochondrial morphology is continuously remodeled by fusion and ssion events 6 with a synergic action of mitophagy in order to maintain a functional pool of mitochondria. In addition to mitochondrial parameters, ultrastructural studies reveal mitophagy activity via the mitochondrial-derived vesicle (MDV) formation. The MDV formation is a process that involves the packing of damaged mitochondria in vesicles which then fuse with the lysosomes for a degradation purpose. We quanti ed the occurrence of buds formation from the outer mitochondrial membrane being a sign of mitophagy and which corresponds to the rst step of MDVs formation 12 . Figure 3 shows the mitochondria forming MDVs precursors in HL60 (Fig. 4A, 4C) and OCI-AML3 (Fig. 4B, 4D). HL60 and OCI-AML3 present signi cantly higher percentage of mitochondria forming MDV than KG1, HEL and K562 (Fig. 4E, Chi² test, p < 0.001).

Different expression of mitochondrial matrix granules in leukemia cell lines
The matrix granules found in the intercristae space are more or less spherical, electron-dense inclusions not surrounded by a membrane. Their diameters are between 25 and 50 nm; they have been observed in several tissues (liver, kidney, plant…) 13 . We quantify the presence of matrix granules according to leukemic cell lines (Fig. 5). Even though all the cell lines were cultivated in the same medium and simultaneously xed, the presence (or not) and the number of mitochondrial matrix granules differed between the cell lines. In fact, 63% of HL60's mitochondria carried one or more mitochondrial matrix granules (Fig. 5B, 5C), whereas these matrix granules were less frequent in HEL, KG1 and OCI-AML3 and were totally absent in K562's mitochondria (Fig. 5D, 5E). To exclude the hypotheses that these structures were due to a staining artefact, a second coloration was done which con rmed this result.

Presence of the Mitochondria-Associated endoplasmic reticulum Membranes (MAMs)
The ultrastructural analysis of our leukemic cell lines indicated their presence in all cell lines (Fig. 6). However, electronic microscopy is not the best way to quantify MAMs, it appears that contacts between mitochondria and endoplasmic reticulum are more numerous in HL60 than in K562 cells (supplementary data Fig. 3).

Discussion
Mitochondrial structural features can not only modify the ability to generate ATP by mitochondrial oxidative phosphorylation but can also commit apoptosis, which could be compromised leading to chemoresistance. Knowing the critical role of respiratory capacity in AML 1 , it would not be surprising that ultrastructural disorders of mitochondria could be implicated in AML pathogenesis 3,14 . In this context, we investigated the mitochondrial ultrastructure in ve leukemic cell lines that represent the different subtypes of AML together with their functional capacity.
Firstly, we showed that there were different respiratory pro les in basal and under stressful conditions according to leukemia cell lines subtypes. On one hand, we identi ed a subgroup of "low respiration" leukemia cell lines, including KG1 and OCI-AML3. Their percentage of spare respiratory capacity and maximal respiration was signi cantly reduced compared to the other subgroups of cell lines that could be considered as a "high respiration" leukemia cell line, including K562, HEL, HL60. KG1 showed the lowest basal respiration, the lowest maximal respiration, the lowest ATP production and the lowest spare respiratory capacity % compared to K562, HEL, HL60. This result was concordant with previous OCR studies concerning KG1 as low OXPHO cell lines 15 . Furthermore, we observed alterations of ROS production that were correlated to the respiratory pro les of leukemic cells. KG1 was less able to produce ROS in basal conditions and conditions stimulating mitochondrial. This rst set of results emphasizes that leukemic cells display various alterations of their mitochondrial function which are probably linked to their molecular status. Indeed, NGS analysis of the ve cell lines revealed that molecular alterations are signi cantly different between cell lines.
Our ultrastructural analysis of mitochondria rstly highlighted the prominent importance of the number of mitochondria on the respiratory pro les. Indeed, we showed that KG1 presented the lowest number of mitochondria/cells which could explain its bioenergetics characteristics. On the contrary, K562, which has the highest basal respiration, the highest maximal respiration and the spare respiratory capacity compared to the other leukemia cell lines, displayed the higher number of mitochondria per cell. Similarly, both HL60 and HEL presented a higher number of mitochondria per cells and higher maximal respiration than KG1. K562 and HL60 also presented the highest ROS emission regardless of the experimental conditions. Interestingly, K562 harbors an ASXL1 mutation that has been shown to enhance mitochondrial activity and to elevate ROS levels in a mouse model 16 . Therefore, we could hypothesize that ASXL1 mutation could, at least partially, induce as "high respiration" pro le through alteration of mitochondrial number.
Since the enzymes involved in oxidative phosphorylation are located on the cristae and the cristae morphology adapts to energy requirements in normal cells 17,18 , we focus on the number of cristae and cristae diameter in leukemia cell lines. K562, HEL and OCI-AML3 presented the higher OCR consumption during ATP production compared to KG1, and as expected, HEL and OCI-AML3 harbored a high number of cristae per mitochondrion. However, K562 presented few cristae per mitochondrion even its high ATP production level. Taking into account its high number of mitochondria, it suggests that leukemic cells could regulate their number of mitochondria to maintain high respiratory levels despite structural defects.
Moreover, we showed that HEL, OCI-AML3 and K562 carry high cristae diameter compared to KG1, to a less level to HL60 with the highest levels of ATP production. That is concordant with previous reports showing that the cristae diameter also gives information about mitochondria's capacity to generate ATP 19 . In synthesis, the mitochondria content measured by the number of mitochondria per cell seems to drive the basal and maximal respiration of leukemia cells and, by consequence, the spare respiratory capacity (%), whereas the cristae diameter seems to better re ect the ATP production. On the opposite, the number of cristae appears as a versatile indicator of ATP production.
Analysis of mitochondria by electron microscopy also depicted disparity in the presence of mitochondrialderived vesicle (MDV) precursors. Originally, MDVs were described in the mitophagy process and enabled the mitochondria to repair damages and restore their normal function 9 . The mitophagy is essential for the regulation of mitochondria degradation and e ciency. Autophagy and mitophagy are mostly described in myelodysplastic disorders. The budding formation of mitochondria's outer membrane -the rst step of MDVs production -was less frequent in K562 than in the other cell lines. Mitophagy de ciency of K562 could lead to the accumulation of damaged mitochondria, explaining the high number of mitochondria/cells in K562 and the "messy" appearance of K562s' cristae. Recently, the induction of mitophagy in AML appears interesting as a novel cell death mechanism especially in FLT3-ITD subgroups, that also arbored mitophagy de ciency 20 . Knowing that the ASXL1 mutant leukemia cell line is also mitophagy-de ciency, it could be interesting to test the induction-mitophagy strategy not only in FLT3-ITD subgroups but also in ASXL1 mutant.
Interestingly, the ultrastructural analysis also gives information about calcium homeostasis through analysis of matrix granules, whose role in calcium sequestration was suggested, as their size and density increased when a high concentration of calcium was present in the uid 21 . Interestingly, these matrix granules could also regulate the internal ionic mitochondrial environment 13 and, thus, impact mitochondrial respiration. In our study, the presence of matrix granules differed according to the cell lines, although they were simultaneously cultured, xed or colored. The HL60 cell line -an in vitro model of AML2, carrying the ampli cation of the proto-oncogene MYC -displayed many matrix granules. On the contrary, none of them were present in the K562 cells that express the BCR-ABL transcript. This observation could be concordant with an indirect role of the matrix granules in the calcium homeostasis.
Indeed, the bcr-abl oncogene was shown to reduce the endoplasmic reticulum releasable calcium levels 22 , while MYC expression -a calcium-dependent proto-oncogene-increased intracellular calcium concentration 23 . Then, we described the presence of the close association of the endoplasmic reticulum with the mitochondria outer membrane known as "Mitochondria-outer mitochondrial membrane that also plays a role in the modulation of the Ca2 + homeostasis in the cancer cells 10 . In fact, the proximity of the mitochondrial and rough endoplasmic reticulum makes selective Ca2 + uptake possible by the mitochondria 24 . In our study, these structures could be observed in all the leukemia cell lines, con rming that this modi cation exists in AML. Interestingly, they were more abundant in HL60 than in K562 cells, where RER was less extensive. Recently, targeting the endoplasmic reticulum-mitochondria interface appears as a novel strategy to make leukemia more sensitive to cytostatics 11 by releasing calcium from the endoplasmic reticulum and inducing apoptosis. Our work suggests that this strategy must be personalized to calcium-dependent leukemia having, for instance, a huge amount of matrix granules.

Conclusions
Our study shows new and interesting data on the shape and quantitative alterations in AMLs mitochondria together with bioenergetics variations. These data suggest that leukemic cells could modulate their energetic metabolism through modi cation of mitochondria shape and/or a number and, thereby, regulate or adapt their proliferative potential to their metabolic environment.

Cell culture and genetics characterization
The cells used for the experiment were purchased from the ATCC (American Type Culture Collection) society. The human leukemic cells of myeloid lineages HEL, HL60, K562, KG1 and OCI-AML3 were routinely cultured at a density of 5 × 10 5 cells/mL for exponential growth in a medium containing RPMI 1640, fetal bovine serum 10%, penicillin/streptomycin 1% and L-glutamine 1% in an autoclave at 37 °C and 5% CO 2 . Genetics characterization of cell lines combined next generation sequencing analysis and ATCC bibliography. Next-generation sequencing methods and results are detailed in supplementary data table 1. Brie y, HEL was established from a secondary leukemia (AML M6 after treatment for Hodgkin lymphoma) and carries a complex karyotype with the JAK2V617F mutation and TP53 mutation. HL60 was established from M2 AML and carries ampli ed MYC gene and NRAS mutation. K562 derived from the pleural effusion of chronic myeloid leukemia in blast crisis and carries the BCR-ABL1 transcript. Next generation sequencing revealed ASXL1 and TP53 mutations in K562. KG1 established from a secondary AML carries TP53 mutation. OCI-AML3 was established from an M4 AML (myelomonoblastic) at diagnosis and carries NPM1 mutation (type A), DNMT3A R882C and NRAS mutation.

Mitochondrial respiration assays
The bioenergetics assays were performed using the XFe96 extracellular ux analyzer and the Seahorse Mito stress test kit (Agilent Technologies). The leukemic cells were suspended after count in an assay medium at pH 7.4. The cells were seeded in a speci c tissue culture plate, previously coated during one night at 4 °C with Corning® Cell-Tak™ (Corning, USA) according to the supplier recommendation. The cells were equilibrated in an unbuffered medium for 45 minutes at 37 °C in a CO 2 -free incubator before being transferred to the XF96 analyzer. The basal oxygen consumption rate (OCR) was measured at baseline and after sequentially injecting the following mitochondrial inhibitors: the ATP synthase inhibitor oligomycin (1.5 µM), the uncoupling agent FCCP (carbonyl cyanide p-tri uoromethoxylphenylhydrazone) in two times (at 2 µM then at 0,5 µm) and the combination of complex III inhibitor antimycin A and the complex I inhibitor rotenone (0,5 µm). Different parameters of respiration, i.e. basal respiration (BR), maximal respiration (MR), ATP-linked respiration and spare respiratory capacity (SRC), were calculated.
All the experiments were performed in triplicate and repeated at least in three independent experiments.

Transmission electron microscopy analysis
Approximately ve million cells were rapidly centrifuged at 1300 rpm for ve minutes. The cell pellets were xed in phosphate buffer containing 2% of glutaraldehyde and 2% of paraformaldehyde overnight.
The xed cells were washed twice in a phosphate buffer 0.1 M at pH 7.2. Then, 50 mM of glycin was added to neutralize aldehyde groups in a third wash. The pellets were resuspended with a pre-warmed gelatin, centrifuged at 10000 rpm for ve minutes and incubated at 4 °C for one hour. The embedding cells in gelatin were sectioned in small fragments (≈ 1 mm 3 ), post-xed in 1% osmium tetroxide (0.1 M cacodylate, pH 7.2) for one hour at 4 °C and stained with 1% uranyl acetate. The samples were dehydrated through graded ethanol concentrations (30, 60, 90, and 100%) before being progressively embedded in Epoxy resin (EPON®). The resin-embedded cells were placed in tubes which were incubated at 60 °C for 48 hours to polymerize the resin. Then, the polymerized blocks were cut in ultra-thin sections (60 nm) which were stained with 5% uranyle acetate and 0.4% lead citrate before being observed under a transmission electron microscope at 80 kV (Jeol 1200 EX). The acquisitions and analysis were done with a digital camera (Veleta, Olympus) and iTEM software. For each cell line, the cellular areas and the number of mitochondria per cell were measured in about 20 cells [19 to 21]. Depending on the cell lines, 74 to 94 mitochondria were examined. For each examined mitochondrion, the mitochondrial area (nm²), the number of cristae and the cristae thickness (nm) were quanti ed (Supplementary data Fig. 1). The cristae thickness was considered at the largest spot of cristae. The matrix granules were considered as 25 to 50 nm electron-dense granules, more or less spherical, located in mitochondria.

Statistical analysis
The Graphpad Prism® software was used for all data analysis. For statistical analysis, non-parametric tests were applied: a Mann-Whitney test for the comparison of two groups, a Kruskal-Wallis test for the comparison of three or more groups and a Chi square test for the qualitative variables. A p value < 0.05 was considered to be statistically signi cant.