To determine the role of LKB1 in mitotic spindle orientation in vivo, we assayed mitotic spindle angles in mice carrying a germline inactivating STK11
. These mice are heterozygous for the mutation (homozygous deletion is lethal during embryogenesis) and phenocopy most features of human Peutz Jeghers syndrome including spontaneous intestinal tumor formation 
. We collected small tumors (polyps) from STK11
mutant mice aged 9–11 months and corresponding gastric/duodenal tissues from wild-type littermates, fixed and embedded them in OCT, and cut sections that were thick enough to encompass entire mitotic spindles within the section. We used immunofluorescence for microtubules and staining for DNA to image spindles, and stained cortical actin to mark the apical brush border as an orientation landmark. We measured spindle angle relative to the apical brush border, defining an angle of 0° as exactly parallel to this surface. We limited the analysis to metaphase and anaphase spindles, for which orientation is fully established in wild-type animals 
Analysis of spindles from wild-type upper gastrointestinal tissues showed planar spindle orientation, as we had found previously in the lower gastrointestinal tract 
. The mean spindle angle in controls was 10°±7° (n
90 spindles from 5 animals), and the highest angle was 30°, seen in only two cells (). In contrast, spindles in cells from STK11
mutant tumors showed a large range of spindle angles. The mean spindle angle was 19°±16° (n
117 spindles from 7 animals), a highly statistically significant difference from controls (P
). Spindle angles greater than 30° were seen in all tumors (). These spindle angle measurements demonstrate that STK11
mutation leads to spindle misorientation in vivo.
STK11 mutation causes spindle misorientation in vivo.
Tumors evolve over several weeks to months in LKB1 mutant animals, during which time additional genetic changes could occur that could be responsible for spindle abnormalities. Thus, we assessed whether a short term LKB1 RNAi approach would also produce spindle misorientation. We first analyzed the effect of LKB1 RNAi on monolayers of MDCK cells, a cell line with well-established spindle orientation regulation, grown on Transwell filters 
. This showed a barely appreciable spindle orientation defect (mean angle of 4° in controls versus 7° upon LKB1 RNAi, p
0.02; data not shown).
Despite the fact that MDCK cells grown on Transwell filters polarize well, growth on filters generates a two-dimensional epithelium, and three-dimensional growth can reveal polarity regulation not seen in two-dimensions 
. We therefore tested spindle orientation in the MDCK cyst system, in which single cells are plated on a bed of Matrigel and allowed to form hollow cysts. These cystic structures are spherical monolayers of polarized cells with the apical cell surfaces facing inward towards a single hollow lumen and the basal cell surfaces facing outward to contact the Matrigel, and they more closely mimic the geometry of the epithelial structures that form in vivo 
In three-dimensional cyst culture, LKB1 RNAi caused a pronounced spindle orientation defect similar to that seen in vivo. Control cysts showed planar spindle orientation, with a mean spindle angle of 10±10° (n
86 spindles from 3 experiments), while cysts with LKB1 RNAi showed spindle misorientation, with a mean spindle angle of 16±18° (18% of spindle angles were >30°, n
84 spindles from 3 experiments; ). Of note, this is a similar magnitude of spindle orientation defect as we saw with RNAi of the tumor suppressor adenomatous polyposis coli (APC) in this same MDCK cyst system (mean angle 18±17°, with 35% of spindles >30°; data not shown). This data supports the conclusion that LKB1 function directly contributes to planar spindle orientation without intervening genetic changes.
LKB1 RNAi causes spindle misorientation in MDCK cell cysts.
To further investigate the mechanism by which LKB1 controls spindle orientation, we assayed for changes in cell polarity markers. In tumors from STK11 mutant mice, cortical actin, ZO-1, and β-catenin localization appeared normal ( and data not shown). Analysis of control and LKB1 RNAi cysts likewise showed that the localization and appearance of cortical actin, as well as the tight junction component ZO-1 and the adherens junction mediator E-cadherin were unaffected by LKB1 reduction. This included the formation of an actin brush border at the cells’ apical/luminal surface, localization of ZO-1 at apical borders and apical cell-cell boundaries, and localization E-cadherin along lateral cell surfaces ( and ). Thus, LKB1 control of spindle orientation appears to be mediated through factors that function downstream of cell-cell junctions and the associated cell polarity machinery rather than by controlling cell-cell junction formation or actin cortex organization.
ZO-1 and E-cadherin localization are unaffected by LKB1 RNAi in MDCK cell cysts.
We previously found that APC mutation caused spindle misorientation without eliminating astral microtubules, suggesting that loss of other components of the spindle orientation machinery besides astral microtubules was responsible for the misorientation 
. Although we were not able to do an extensive quantitative analysis, we also found that astral microtubules were present in cells from STK11
mutant animals (). This suggests that the role of LKB1 in spindle orientation is in control of the orientation machinery rather than in elongation of astral microtubules, although it does not rule out effects of LKB1 on astral microtubule dynamics.
Since cortical markers and astral microtubule length appeared unaffected by LKB1 loss of function, we next looked at LKB1 substrates to identify potential mediators of the role of LKB1 in spindle orientation. We focused on AMPK as one of the best characterized LKB1 substrates; AMPK also has a potential cancer association 
. We assayed the localization of activated (phosphorylated) AMPK, which has been shown to localize to spindle poles in cultured cell lines and tumors 
In cells from wild-type animals, an antibody to AMPK phosphorylated at Thr-172 (the site of phosphorylation by LKB1) showed the expected localization to mitotic spindle poles (). In contrast, this antibody showed a new and dramatically different localization in LKB1 mutant tissues. A majority of mitotic cells from LKB1 mutant animals showed localization of phospho-AMPK to the cell cortex, either combined with spindle pole localization or alone (). For wild-type tissues, 75% of mitotic cells showed localization of phospho-AMPK to the spindle poles, and 6% showed localization to the cell cortex; these localizations were not mutually exclusive (n
134 mitotic cells from 4 animals). The finding of rare cortical localization in controls suggests the possibility that the cell cortex is a physiological site for phospho-AMPK localization that is transient.
LKB1 mutant tumors show mislocalization of activated AMPK.
mutant tissues, 70% of mitotic cells showed localization of phospho-AMPK to spindle poles and 68% showed cortical localization (n
130 cells from 5 animals). Thus, STK11
mutation caused dramatic mislocalization of activated AMPK in addition to spindle misorientation. Of note, we also saw apparent kinetochore localization of phospho-AMPK in 10–15% of mitotic cells from both wild type and STK11
mutant animals (, inset). Thus, LKB1 mutation produced cortical localization of phospho-AMPK without perturbing other sites of activated AMPK localization in mitotic cells.
While STK11 mutation caused mislocalization of activated AMPK in tissues, we did not see this mislocalization in MDCK cell cysts with LKB1 RNAi. Further, depolymerization of astral microtubules with low-dose Nocodazole in MDCK cyst cells did not reproduce this mislocalization (data not shown). Thus, additional features of the tumorigenic process must be involved in this activated AMPK mislocalization.
To test for a role of AMPK in controlling spindle orientation, we attempted to reduce AMPK level or function in MDCK cysts by a variety of specific methods. However, for various technical reasons, these were unsuccessful (see Methods
). Thus, we used the pharmacologic agent Compound C, which is often used as a specific AMPK function-blocking tool, but which in fact inhibits several other important signaling pathways 
. Since AMPK is reported to play a role in tight junction formation in MDCK cells 
, we applied Compound C to pre-formed MDCK cell cysts in which tight junctions have already been formed.
Compound C treatment for 3.5–4 hours resulted in significant spindle misorientation, with a mean spindle angle in controls of 8±8° (n
62 spindles from 3 experiments) and a mean angle of 22±19° in Compound C treated cyst cells (n
70 spindles from 3 experiments; ). This demonstrates a similar degree of spindle misorientation upon AMPK inhibition as with LKB1 RNAi. Of note, Compound C treatment also caused an increased incidence of monopolar spindles (27% of spindles (22 of 83); ) and misattached chromosomes (28% of spindles (17 of 61); ), which did not occur with LKB1 RNAi. These findings suggest that AMPK could be a mediator of the role of LKB1 in spindle orientation and could also have additional mitotic roles.
AMPK inhibition by Compound C causes spindle misorientation and other mitotic defects in MDCK cell cysts.