Death is the endpoint for nearly every keratinocyte leaving the basal layer. Its life will terminate as a corneocyte, a flattened cell with no nucleus, no metabolism, and the remnants of a living cell. That death is programmed, but it isn’t apoptosis. If a keratinocyte died by apoptosis instead, before its differentiation is complete, it would leave a gap in the tissue it was born to build. Terminal differentiation takes weeks, and it begins earlier than you might expect. Earlier than I expected, too. For all that time, the cell has to avoid the wrong death.
Live imaging in adult mouse ear skin. The same patches of skin were revisited every 24 hours, so individual cells could be followed for days. Division was blocked genetically, by switching on p27 in keratin 10-expressing cells, and tape stripping tested what happened when demand rose.[1]
In Katie Cockburn’s study of living adult mouse skin, a fluorescent reporter linked to keratin 10, a protein associated with early differentiation, allowed individual cells to be followed over several days. The reporter usually switched on three to four days before a cell left the basal layer, so differentiation was well under way before the cells began to move. Some of the labelled cells were still dividing while they occupied the basal layer, producing daughters that retained the signal. Most of those daughters subsequently left the basal layer, continuing the differentiation process begun by their mother cell.[1]
A marker isn’t a mechanism. Keratin 10 is one of the most reliable early signs that a keratinocyte has started to differentiate. Yet epidermis made without keratin 10 still stratifies and forms a working barrier. A marker tells you a programme has started, not what the programme needs.[1]
It’s easy to picture a keratinocyte finishing proliferation, then beginning to differentiate, then moving upwards through the epidermis. Cockburn’s observations change that sequence. Some cells had already begun differentiating, then divided again, and so the start of differentiation did not end their capacity to divide.
To find out whether those further divisions were necessary for the cells to progress, the researchers switched on a cell-cycle inhibitor, p27, in cells that had already begun expressing keratin 10. Those cells stopped dividing, but they still left the basal layer and matured as normal. Their undifferentiated neighbours made up the numbers through compensation: they increased the rate of cell division, and more of them switched on keratin 10, so the differentiating pool stayed the same size. Even after tape stripping, the differentiating pool recovered with only a mild delay, despite a deeper initial drop.[1] The divisions help as a buffer, but undifferentiated neighbours can cover for them.
There is a long road between the beginning of this transformation and a finished corneocyte. The living cell must re-organise its keratin network, assemble the proteins of its cornified envelope, and both synthesise and secrete lipids for the surrounding intercellular matrix. Its nucleus and other organelles are eventually dismantled as cornification produces the non-living cell retained at the surface. This preparation begins while the keratinocyte is still dependent on the machinery it will later lose.
So the cell has to be protected from the wrong death while it carries out the differentiation that makes the right one possible. How much of differentiation depends on that protection only becomes apparent when researchers remove it.[2,3]
A failure to differentiate
A poorly formed epidermis leaves us with several possible explanations. Perhaps its cells have lost their instructions or the machinery needed to differentiate. Perhaps they are dying prematurely. If reducing one protein causes both abnormalities, the appearance of the tissue will not tell us how closely the two are connected.
This was the problem exposed by Barry Thrash’s experiments with AKT1. AKT proteins are kinases, enzymes that transfer phosphate groups onto other proteins, thereby changing their activity. They influence multiple processes ranging from metabolism to survival, so an increase in AKT activity during differentiation does not immediately reveal what exactly it is doing there.
There was a reason to suspect a survival signal. Detachment from the basement membrane triggers keratinocyte differentiation. In many other epithelial cells, losing this attachment triggers a rapid form of apoptosis called anoikis. Something has to keep a detaching keratinocyte alive long enough to differentiate, and Thrash’s group asked whether AKT was part of that explanation.[2]
RNA interference in reconstructed human epidermis. Primary human keratinocytes with AKT1 or AKT2 reduced were grown into layered tissue on a fibroblast-containing collagen matrix. z-VAD-fmk was present for the whole culture, with a cathepsin inhibitor as the control treatment. It tests whether premature death explains the defect; it doesn’t restore AKT1.[2]
Both AKT1 and the related AKT2 became activated under the differentiation conditions Thrash tested. To investigate their functions, he reduced production of each protein using RNA interference in primary human keratinocytes, then grew the cells in organotypic cultures. Above a collagen matrix containing fibroblasts, these keratinocytes usually form a layered epidermal tissue. The consequences of changing a signal could therefore be followed through the organisation of the developing layers.[2]
With AKT1 depleted, organisation deteriorated. The layers above the basal cells were thin, no granular layer was apparent, and the stratum corneum was thickened but lacked its normal lamination. Keratin 10 extended up into the cornified region, and filaggrin staining was non-uniform. Reducing AKT2 alone did not reproduce the defect, despite both forms being activated during differentiation.[2]
Within the AKT1-depleted tissue, cells above the basal layer had condensed nuclei, and DNA-break labelling extended from the layer immediately above the basal cells up through the cornified layer. In control cultures, it was confined to the boundary between the granular and cornified layers, where cells are expected to die. Without AKT1, cells were dying as soon as they left the basal layer.[2]
One inhibitor, many caspases. z-VAD-fmk blocks many caspases at once and binds them irreversibly. That makes it a strong test of whether caspase activity matters to the outcome, and a weak one for which caspase is responsible. Janes’s narrower inhibitors were an attempt at the second question.[2,5]
Now the researchers had a way to investigate the connection. They left AKT1 depleted and added z-VAD-fmk, a broad-spectrum inhibitor of caspases. Caspases are proteolytic enzymes, including several key members that drive the apoptotic response. Their programmed activity leads to dismantling of the cell; suppressing them would test whether premature caspase-dependent death was contributing to the abnormal tissue.
The treatment reduced DNA-break labelling to control levels and prevented the abnormal cornified layer from forming. A granular layer reappeared, and keratin 1 expression returned to normal.[2] The missing AKT1 had not been restored, but blocking caspases had let the tissue stratify similar to the controls. The authors concluded that AKT1 acts primarily as a survival factor.
The death itself was unusual. Active caspase-3, the executioner of classical apoptosis, wasn’t detected in the AKT1-depleted cultures. Instead, caspase-14, a family member associated with cornification rather than apoptosis, was deregulated, with extensive staining through all suprabasal layers. The authors described the death as consistent with normal differentiation. Without AKT1, the planned death seems to have arrived too early.[2]
Calling AKT1 necessary for differentiation in this model now tells us something more specific. Part of that necessity was to keep the differentiating cells alive. But how much of an abnormal tissue could be recovered this way? A separate study of another survival protein, Mcl-1, took the rescue further, into cornification itself.
The incomplete rescue
Mcl-1 belongs to the BCL-2 family, whose members modulate the intrinsic, mitochondrial route into apoptosis. Pro-apoptotic family members such as BAK and BAX induce this response through outer mitochondrial membrane permeabilisation (MOMP), which releases proteins such as cytochrome c into the cytoplasm. There, cytochrome c joins APAF-1 to form the apoptosome, which activates caspase-9 and, through it, the executioner caspases. Mcl-1 restrains BAK- and BAX-mediated permeabilisation, preventing this death-inducing cascade.[3]
RNA interference in raft cultures of human keratinocytes. Mcl-1 was reduced to about 45% of control. Cell death, cornified envelopes and marker proteins were measured separately, which is what let the rescue come apart: z-VAD cut TUNEL-positive cells from about 20 per field to 3, but marker expression didn’t recover.[3]
In work published in 2009, Leonid Sitailo, Anita Jerome-Morais and Mitchell Denning reduced Mcl-1 in primary human keratinocytes and grew them as epidermal raft cultures. Premature apoptosis increased in the living layers, accompanied by active caspase-3, absent from the AKT1 cultures, and DNA-break labelling. This was classical apoptosis. The granular and cornified layers developed poorly, and fewer cornified envelopes were produced.[3]
Cornified envelope assembly appears late in differentiation, as proteins are cross-linked into a tough shell beneath the plasma membrane. It remains after cellular life has ended, contributing to the strength of the resulting corneocyte. Measuring the envelopes gave the researchers a view of what the cells had managed to assemble, alongside their measurements of cell death.
They also examined proteins expressed at different stages of differentiation. Keratin 1 and keratin 10 are expressed early, contributing to changes in the cytoskeleton; loricrin, a late stage differentiation marker, becomes part of the cornified envelope. Levels of these proteins, together with filaggrin, were reduced in the Mcl-1-depleted cultures. Processing of caspase-14 was also impaired.[3]
Adding a caspase inhibitor reduced apoptosis and partially restored cornification. So far, the result fitted the idea that premature death had interrupted the programme. Yet when the differentiation proteins were examined again, their expression had not significantly recovered. Neither had caspase-14 processing.[3]
Prior to the rescue, the fall in keratin 1, keratin 10, loricrin and filaggrin could have reflected fewer living cells to synthesise them. The inhibitor weakened that explanation. Apoptosis had declined, yet protein levels in the tissue remained low. Keeping the cells alive had exposed the part of the defect survival couldn’t fix.
The authors concluded that Mcl-1 supported differentiation through a further function, in addition to preventing premature apoptosis. That also explains why cornification only partly recovered: mature envelopes need the differentiation proteins the inhibitor couldn’t restore. Their work did not identify the molecular mechanism of that function, but it gave them a reason to look for one.[3]
Caspases have other work to do
Both rescue experiments depended on inhibiting caspases. We’ve followed the effect of that intervention on dying cells, but there is another question to consider. What else might those enzymes have been doing while a keratinocyte differentiated?
The caspase family had already attracted attention. In 2000, Lippens’s group examined the apoptotic executioner caspases 3, 6 and 7 in normal skin and experimental differentiation systems. In this setting, they did not detect the processing of these enzymes usually seen during apoptosis and concluded they play no role in keratinocyte terminal differentiation. Caspase-14 followed a different pattern, with its production and processing accompanying differentiation. Inducing apoptosis experimentally activated caspase-3 without reproducing the caspase-14 response.[4]
The differentiation response involved a different member of the family from the executioner caspases measured during apoptosis. But what happened earlier, as a cell began to differentiate?
Human skin sections and suspension-induced differentiation. Active Akt and caspase activity were located in sections of human skin. Commitment was then timed in primary keratinocytes held in suspension, using inhibitors to test which signals were needed and flow cytometry to compare committed cells with apoptotic controls.[5]
Reading a caspase probe. The probe is a fluorescent inhibitor that binds active caspases. A bright cell shows caspase activity at that moment, not that the cell is dying.[5]
In 2009, Sam Janes and colleagues, with Fiona Watt as senior author, approached an earlier stage. In sections of human skin, Akt in its phosphorylated, active form was concentrated in the first two or three layers above the basal layer and faded higher up. Caspase-probe labelling was absent from the basal layer but present in every suprabasal layer. Both were detected in cells exiting the basal layer. To follow this cross-over period, the team placed primary human keratinocytes in suspension, removing their attachment to a culture surface. This is the anoikis test again. As expected, detached MDCK cells, a dog kidney line used as the apoptosis control, were apoptotic within one day. Detached keratinocytes were not. Within about four hours these keratinocytes had committed to differentiation, and when returned to a culture surface, they would no longer divide. Caspase-probe labelling appeared before involucrin, a differentiation protein, and inhibiting caspases blocked differentiation. A caspase-8 inhibitor suppressed it as efficiently as the broad inhibitor, and a caspase-9 inhibitor partly did.[5]
Yet commitment wasn’t a classical apoptotic process. Cells of the committed population displayed apoptotic signatures such as loss of mitochondrial membrane potential, signs of cytochrome c release, and they became permeable to 7-AAD, a DNA dye which selectively labels membrane-compromised cells. They didn’t, however, expose phosphatidylserine or produce extra reactive oxygen species, classical signs of apoptosis, and their nuclei also swelled rather than condensed. Here, commitment had used parts of the apoptotic machinery, but it wasn’t apoptosis.[5]
Here, inhibition interfered with differentiation, yet in the previous AKT1 and Mcl-1 cultures, it had helped parts of the tissue recover. These experiments examined different systems and events, and the comparison doesn’t resolve every difference between them. It does show why detecting activity in the caspase family is insufficient evidence to establish that a cell is undergoing apoptosis.[2–5]
The boundary between survival and death machinery cannot be reduced to a simple explanation. Akt and active caspases are both present in the same suprabasal layers. A caspase inhibitor rescues one culture and stalls another. Keratin 10 is a marker of differentiation in a cell that may still divide. In each case, a protein’s function depends on the stage and setting in which it acts.
The corneocyte at the surface is the outcome of this complex balance. By then, the keratinocyte has made material the skin will retain, lost its nucleus and sealed its fate. Apoptosis in the living layers is a different kind of death: earlier, faster, and activated by different machinery. The need for protection persists even as the cell prepares for its own death.[2–4]
There is another source of difficulty, though. Janes’s inhibitor experiments implicated caspase-8 during differentiation without identifying what activated it. Caspase-8 is also the enzyme that death receptors at the cell surface activate, which lets a keratinocyte respond to signals from its surroundings. What determines whether those signals proceed? That is the question cFLIP will bring us to.
Apoptosis. A regulated, caspase-driven form of cell death in which the cell condenses and is cleared without inflammation.
Cornified envelope. A cross-linked protein shell beneath the plasma membrane of a corneocyte that gives it mechanical strength.
Kinase. An enzyme that adds phosphate groups to other molecules, often switching their activity on or off.
Anoikis. Apoptosis triggered when a cell loses its attachment to the extracellular matrix.
RNA interference. A method that reduces production of a chosen protein by destroying the messenger RNA that encodes it.
Organotypic culture. Keratinocytes grown at an air–liquid interface on a dermal-like support, so that they form layered tissue resembling epidermis. Also called raft culture.
Caspase. A protease that cuts proteins after aspartate residues. Some run apoptosis; others, such as caspase-14, work in differentiation.
BCL-2 family. Proteins that control whether mitochondria release the signals that start intrinsic apoptosis. Some promote release; others, such as Mcl-1, prevent it.
Apoptosome. A complex formed when cytochrome c, released from mitochondria, binds APAF-1 in the cytoplasm. It activates caspase-9.
Involucrin. One of the first proteins built into the cornified envelope, used as an early marker of terminal differentiation.
References
1. Cockburn, K., Annusver, K., Gonzalez, D.G., et al. (2022). Gradual differentiation uncoupled from cell cycle exit generates heterogeneity in the epidermal stem cell layer. Nature Cell Biology, 24(12), pp. 1692–1700. https://doi.org/10.1038/s41556-022-01021-8
2. Thrash, B.R., Menges, C.W., Pierce, R.H., et al. (2006). AKT1 provides an essential survival signal required for differentiation and stratification of primary human keratinocytes. Journal of Biological Chemistry, 281(17), pp. 12155–12162. https://doi.org/10.1074/jbc.M512116200
3. Sitailo, L.A., Jerome-Morais, A. and Denning, M.F. (2009). Mcl-1 functions as major epidermal survival protein required for proper keratinocyte differentiation. Journal of Investigative Dermatology, 129(6), pp. 1351–1360. https://doi.org/10.1038/jid.2008.363
4. Lippens, S., Kockx, M., Knaapen, M., et al. (2000). Epidermal differentiation does not involve the pro-apoptotic executioner caspases, but is associated with caspase-14 induction and processing. Cell Death and Differentiation, 7(12), pp. 1218–1224. https://doi.org/10.1038/sj.cdd.4400785
5. Janes, S.M., Ofstad, T.A., Campbell, D.H., et al. (2009). PI3-kinase-dependent activation of apoptotic machinery occurs on commitment of epidermal keratinocytes to terminal differentiation. Cell Research, 19(3), pp. 328–339. https://doi.org/10.1038/cr.2008.281
6. Sayama, K., Yamasaki, K., Hanakawa, Y., et al. (2002). Phosphatidylinositol 3-kinase is a key regulator of early phase differentiation in keratinocytes. Journal of Biological Chemistry, 277(43), pp. 40390–40396. https://doi.org/10.1074/jbc.M112423200
A short note
The AKT1 and Mcl-1 experiments used reconstructed tissue grown from primary human keratinocytes. Their caspase-inhibitor rescues measured cell-death signals, tissue organisation and differentiation outcomes. Neither established complete recovery of permeability barrier function.[2,3] Cockburn’s live-cell observations came from adult mouse epidermis.[1]
The broader PI3K–AKT pathway also resists a single rule about differentiation. PI3K is a lipid kinase involved in signalling upstream of AKT. Sayama found that inhibiting PI3K increased early keratin 1 and keratin 10 expression in human keratinocyte cultures, while Janes found reduced differentiation-marker induction after PI3K inhibition in a suspension system.[5,6] The conditions and readouts differed, and the precise cause of the divergent responses is not resolved by comparing these studies.
Caspase-probe labelling, caspase processing and DNA-break labelling measure different events. Janes’s assignment to caspase-8 rested on inhibitor comparisons, which did not establish that a death receptor had assembled the complex responsible for the response.[5]
Thrash’s and Sitailo’s papers acknowledge NIH support.[2,3] Janes’s publisher acknowledgements include Cancer Research UK, an MRC fellowship, an EU Marie Curie fellowship and Cambridge/Hutchison Whampoa support. One author’s affiliation was Immune System Therapeutics Ltd.[5]
