Biochemical insights into the mechanisms central to the response of mammalian cells to cold stress and subsequent rewarming


C. M. Smales, Protein Science Group, Department of Biosciences, University of Kent, Canterbury, Kent CT2 7NJ, UK
Fax: +44 01227 763912
Tel: +44 01227 823746


Mammalian cells cultured in vitro are able to recover from cold stress. However, the mechanisms activated during cold stress and recovery are still being determined. We here report the effects of hypothermia on cellular architecture, cell cycle progression, mRNA stability, protein synthesis and degradation in three mammalian cell lines. The cellular structures examined were, in general, well maintained during mild hypothermia (27–32 °C) but became increasingly disrupted at low temperatures (4–10 °C). The degradation rates of all mRNAs and proteins examined were much reduced at 27 °C, and overall protein synthesis rates were gradually reduced with temperature down to 20 °C. Proteins involved in a range of cellular activities were either upregulated or downregulated at 32 and 27 °C during cold stress and recovery. Many of these proteins were molecular chaperones, but they did not include the inducible heat shock protein Hsp72. Further detailed investigation of specific proteins revealed that the responses to cold stress and recovery are at least partially controlled by modulation of p53, Grp75 and eIF3i levels. Furthermore, under conditions of severe cold stress (4 °C), lipid-containing structures were observed that appeared to be in the process of being secreted from the cell that were not observed at less severe cold stress temperatures. Our findings shed light on the mechanisms involved and activated in mammalian cells upon cold stress and recovery.


chaperonin containing T-complex polypeptide1


cold-inducible RNA-binding protein


endoplasmic reticulum


heat shock factor


non-equilibrium pH gradient gel electrophoresis


quantitative real-time PCR


RNA-binding motif protein 3

The heat shock response has been extensively studied in a variety of systems and organisms, and generally involves the conserved and coordinated upregulation of heat shock proteins that act to alleviate the cellular stresses imposed by hyperthermic stress. Our current understanding of the cellular responses to subphysiological temperatures (hypothermia) is less extensive. This is somewhat surprising, because of their relevance in medicine for the storage of cells, organs, and tissues, and the treatment of brain damage; as well as in the biopharmaceutical sector, where reduced culture temperature can sometimes improve recombinant protein yields from mammalian cells cultured in vitro [1]. What is clear is that the general response to hypothermia appears to include the global attenuation of transcription and translation, whereas a small group of proteins, termed the cold shock proteins, are selectively induced [2]. However, unlike their heat shock counterparts, these cold shock proteins do not appear to be particularly well conserved between prokaryotic and eukaryotic systems, and their functions, such as have been defined, have to date been described in terms of their RNA rather than their protein biology. Exposure to subphysiological temperature is also known to generally lead to changes in the lipid make-up of membranes, resulting in increased membrane rigidity, compromised membrane-associated cell functions, and alterations in lipid synthesis and disposition [3].

The most well-characterized cold shock responses to date are those in plants and bacterial systems [3], there being much less information on the molecular mechanisms underpinning the cold shock response in mammalian cells. Our current understanding is that the cold shock response in mammalian cells involves the coordination of transcription, translation, the cell cycle, metabolism, and cell cytoskeleton organization, but the exact mechanisms by which these are modulated remain to be elucidated in most cases. Furthermore, mammalian cells respond to mild hypothermia (25–35 °C) in a different manner to more severely reduced temperatures (0–10 °C). This may largely reflect the fact that at more moderate temperatures, cells can still proliferate and grow, whereas at the severe temperatures, growth is fully arrested. Studies on hibernating animals and various in vitro cultured mammalian cell systems have also reported that mammalian cells in general respond to cold shock by disassembly of the cell cytoskeleton, delayed apoptosis, reduced metabolism with reduction of ATP expenditure, reduced protease activity, a reduction in free radical oxygen species, and attenuation of transcription and translation [3]. As a result of such reports, it has been suggested that there are five general mechanisms by which mammalian cells respond to cold shock. These are: (a) a general reduction in transcription/translation; (b) reduction of RNA degradation; (c) increased expression of specific target genes; (d) the generation of alternative mRNAs via presplicing events; and (e) use of internal ribosome entry segments for preferential cap-independent translation of specific mRNAs under cold shock conditions [4].

There have been only two well-characterized mammalian cold shock proteins reported to date, RNA-binding motif protein 3 (Rbm3) [5] and cold-inducible RNA-binding protein (Cirp) [6]. Both of these are induced in response to mild hypothermia (maximal expression around 32 °C) but not severe hypothermia, and are probably general stress response proteins, as they are also induced by a number of other stresses. It is generally thought that Rbm3 and Cirp are involved in the modulation of transcription and translation upon cold stress and function as RNA chaperones, although the exact function of these proteins remains to be elucidated [7]. Cirp and Rbm3 are highly similar proteins that consist of an N-terminal RNA-binding domain and a C-terminal glycine-rich domain, but show no homology to the cold shock proteins found in bacterial systems [3]. Our current understanding of the cold shock response, and the mechanisms involved in coordinating that response, in mammalian cells is therefore extremely limited.

It is interesting to note that although cold stress appears to generally reduce protein synthesis, in recent years recombinant protein production from in vitro cultured mammalian cells has been improved by reducing operating temperatures from 37 °C to mildly hypothermic levels (28–34 °C) towards the end of the logarithmic increase in cell number [8]. This strategy has been adopted because although cell division and protein synthesis rates are appreciably slowed, cells show prolonged viability and increased cell-specific productivity under these mildly hypothermic conditions. With regard to the cultivation of mammalian cells at subphysiological temperatures, the prolonged cell viability, delayed apoptosis, decreased glucose and glutamine consumption, decreased waste product release and increased tolerance to shear stress during cultivation under mildly hypothermic conditions are all features likely to extend the productive life of cells in terms of recombinant protein production [1]. However, as the cell division cycle slows and even arrests in G1 at the lower end of the mildly hypothermic range, and both transcription and translation rates are reduced at subphysiological temperatures, this may offset any potential positive effects of subphysiological culturing.

To date, characterization of the mammalian cold stress response has largely derived from microarray analyses of transcriptional changes and single snapshot proteomic analyses of changes in protein levels upon exposure to subphysiological temperatures [3]. These have been most useful in defining the overall adaptations to growth at subphysiological temperatures and in highlighting areas for further, more detailed investigations into the mechanisms of cold adaptation. Here, we have investigated changes in protein synthesis rates upon cold stress, and show that an examination of changes in the synthesis rates of specific proteins during cold stress, and during recovery, identifies subtle adaptations to growth at subphysiological temperatures, not all of which have been previously uncovered by proteomic analysis of overall protein levels. The proteins identified in this study encompass a wide range of cellular activities, including cell cycle regulation, translation initiation, cytoskeleton organization and, most particularly, molecular chaperone activity. Specific investigation of the roles of a number of the proteins identified leads us to the conclusion that the regulation of p53, Grp75 and eIF3i protein levels may play a key role in the response to, and recovery from, cold stress in mammalian cells. The implications of these findings in terms of the cold stress response in mammalian cells are further discussed.


The cell lines chosen for this study were the commercially relevant CHOK1 cell line and two mouse cell lines, P19 embryonal carcinoma cells and NIH 3T3 fibroblasts. P19 cells were chosen because of their particular sensitivity to cold stress. The temperature ranges investigated covered both severe cold stress (4 and 10 °C) and more mild cold stress (27 and 32 °C), and were considered to be relevant to organ/tissue storage and the subphysiological in vitro culturing of mammalian cells for recombinant protein production respectively. The periods of time for which cells were exposed to cold stress (6–30 h) were those sufficient to elicit detectable cold stress responses while still allowing the majority of cells, if not all, to recover upon rewarming.

Severe, but not mild, cold stress results in marked changes in the cellular architecture of in vitro cultured mammalian cells

Immunofluorescence studies highlighted changes in the structural architecture of all cell lines investigated upon their exposure to more severe cold stress (Figs 1 and S1). Cells cold-stressed at 27 or 32 °C for 6 h were indistinguishable from those maintained at 37 °C when investigated by immunofluorescence, with the exception that although the microtubule content was unchanged in P19 cells at 27 °C, the microtubule organization appeared to be compromised. Cells cold-stressed at 27 or 32 °C, or maintained at 37 °C, were well spread, and the organelle distribution appeared to be unchanged as exemplified by mitochondrial staining (Fig. 1). Furthermore, when we monitored poly(ADP-ribose) polymerase levels by western blot analysis, there was no evidence of poly(ADP-ribose) polymerase cleavage (Fig. S2A), which is activated upon apoptosis, and therefore we conclude that the cells are not apoptotic at the times and temperatures investigated. At 27 and 32 °C, reduced cell proliferation was observed, and previous reports have suggested that the cold shock protein Cirp may be responsible for cell cycle arrest of mammalian cells at subphysiological temperatures [9]. However, our results suggest that cell cycle arrest at 27 °C is the result of an increase in the overall level (Fig. 2A) and changes to the post-translational modification pattern (Fig. 2B) of p53, which persisted in cells maintained at 27 °C for 6 h to 6 days. The expression of p21, a general inhibitor of cyclin-dependent protein kinases and a downstream effector of elevated p53 levels, was also induced at 27 °C. During recovery from cold stress at 37 °C, the overall amounts of these two proteins, and the isovariants of p53, returned towards their normothermic values and appearance (Fig. 2). The change in isovariant levels of p53 was not due to increased levels of acetylation, as shown by western blot analysis with an antibody specific for acetylated p53 (Fig. S2B). Across the mild cold stress conditions and time periods examined, there was 100% viability and recovery upon returning the cells to 37 °C.

Figure 1.

 Structural changes in CHOK1, P19 and NIH 3T3 cells during mild and severe cold stress and after exposure to nocodazole. Cells were maintained at 37 °C, or then transferred to 27 or 4 °C for 6 h, or exposed to growth medium containing 3 μg·mL−1 nocodazole, prior to fixation for immunofluorescence microscopy. In each panel of four: top left, phase contrast; top right, α-tubulin detection; bottom left, F-actin detection; bottom right, mitochondrial Hsp60 detection. In CHOK1 cells at 4 °C, the lower section of the upper right panel shows the presence of multiple microtubule asters. In each panel of two (nocodazole staining):left, phase contrast; right, α-tubulin detection. Bar: 10 μm. Arrows in the 4 °C phase contrast image point to vesicle-like structures that are also highlighted in Fig. S1.

Figure 2.

 p53 is elevated and undergoes changes in post-translational modification in CHOK1 cells during maintenance at 27 °C. (A) Twenty micrograms of proteins extracted from CHOK1 cells, maintained at 27 °C or then rewarmed at 37 °C for the indicated times, were resolved by SDS/PAGE and detected by probing immunoblot for p53 or p21. (B) One hundred micrograms of proteins extracted from CHOK1 cells as in (A) were resolved by NEPHGE-SDS/PAGE and immunoblots were probed for p53. Arrows highlight changes in isoform distribution between the 37 and 27 °C samples. (C) As in (A) at 27 °C for up to 5 days (d, day) and upon rewarming (csr, cold shock recovery) for 5 h.

At near-freezing temperatures, the cellular appearance and architecture were profoundly affected, in a cell line-specific manner, as compared with cells maintained at 37 °C. Under conditions of severe cold stress, cells were much more rounded and less well spread, and both nuclear and cellular diameter were decreased. Most cells cold-stressed at 4 °C had translucent, vesicle-like structures that were easily detected by phase contrast viewing, on the surface of the cell (arrowed in Figs 1 and S1). These vesicle-like structures appeared within 2 h of exposure to the severe cell stress but disappeared very rapidly (< 0.5 h) upon rewarming. The content of these vesicles was not revealed by general stains for DNA, RNA or protein, although this last stained the perimeter of the vesicles, but was instead strongly stained by the dye Oil Red O, which preferentially binds to uncharged lipids (Fig. S1). The proportion of cells able to recover from severe cold stress was much smaller than observed for recovery from mild cold stress, and typically 20–30% of cells did not survive upon rewarming from extreme cold (4 °C). Electron microscopy of cold-stressed cells revealed the presence of lipid-containing structures (as stained by osmium tetroxide) corresponding to these vesicles that appeared to be in the process of being extruded from the cell (Fig. 3).

Figure 3.

 Electron micrographs of P19 and CHO cells at 37 and 4 °C. P19 cells (A, C) and CHO cells (B, D) at 37 °C or after 6 h at 4 °C. At 4 °C, vesicle-like structures that contain lipids were observed (arrowed) that appeared to be in the process of being extruded from the cell. Bar: 1 μm.

Of the cell lines investigated, CHOK1 cells proved to be the most resilient to subphysiological temperatures, possibly because they contain a sizeable population of cold-stable microtubules and therefore maintain a relatively ordered intracellular organization at low temperatures (Fig. 1). Even at 4 °C, a sizeable population of microtubules persisted in CHO cells, although these appeared to be less ordered than those in cells maintained at 37 °C. Mitotic cells appeared to be particularly vulnerable to cold stress, and below 20 °C cells were observed that contained multiple microtubule asters (Fig. 1; CHOK1 cells at 4 °C, lower section of the upper right panel) that were very similar in appearance to those formed in mitotic cells exposed to the microtubule-stabilizing drug taxol [10]. Such structures are formed when the nuclear envelope breaks down during mitosis and releases proteins (e.g. NuMA) that stabilize microtubule minus-ends and that would normally fulfil this function within the mitotic spindle poles. In the presence of taxol, these proteins mis-localize to the cytoplasm, where they stabilize microtubule aster formation [11]. The structures reported here in cold-stressed cells may originate from similar mis-localization of nucleating proteins.

P19 cells had no detectable cold-stable microtubules, and organelle disposition was severely disrupted by temperatures below 20 °C (Fig. 1). The difference in microtubule stability between CHOK1, NIH3T3 and P19 cells was reflected by a relative abundance of STOP proteins (data not shown), splice variants of which stabilize microtubules to cold exposure [12–14]. However, the apparent reduction in cell size and vesicle release observed in cells exposed to severe cold shock is not simply a consequence of a loss of microtubules. Exposure of CHOK1, NIH3T3 and P19 cells to the antimicrotubule drug nocodazole at 37 °C completely depolymerized the microtubule networks in all three cell types and caused retraction of the cytoplasm, but, in contrast to cold-stressed cells, the nuclear size appeared to be unchanged (Fig. 1). Furthermore, no extracellular vesicular structures were observed after nocodazole treatment alone, suggesting that microtubule depolymerization is not the key signal in the formation of the vesicle structures.

Mammalian cells regulate the synthesis rates of specific proteins in response to cold stress and upon recovery at 37 °C

Although there have now been a number of proteomic studies of the cold shock response in various systems that have yielded valuable information, these have been static-based measurement approaches and so do not account for any variation in protein synthesis and turnover rates upon exposure to any given cold-related stress. In order to determine the protein synthesis capacity of in vitro cultured mammalian cells at subphysiological temperatures, newly synthesized proteins were radiolabelled with [35S]methionine/cysteine mix for 1 h at the cold stress temperatures described above. Scintillation counting was undertaken of samples from cells stressed at different temperatures, to ensure that any difference in the radioactive amino acid uptake did not account for differences in label incorporation into polypeptides (Fig. 4A). The acid-soluble material extracted from CHOK1 cells changed little between 20 and 37 °C, and at 20 °C the acid-soluble/insoluble ratio was actually increased (Fig. 4A). We therefore concluded that methionine/cysteine uptake was not limiting for changes in label incorporated into polypeptide/protein over the temperature range investigated.

Figure 4.

 Amino acid uptake and protein synthesis in CHOK1 cells exposed to, and recovering from, cold stress. (A) Cells maintained at 37 °C, or then exposed to the indicated temperatures for 6 h, were radiolabelled in methionine/cysteine-deficient growth medium supplemented with 1770 kBq·mL−1 [35S]methionine/cysteine cell labelling mix for 10 min at the indicated temperature prior to extraction into ice-cold 0.6 m trichloracetic acid. (B) Cells maintained at 37 °C, or then exposed to the indicated temperatures for the indicated times, were radiolabelled as in (A) but for 1 h at the indicated temperatures. Thirty micrograms of extracted proteins were resolved by SDS/PAGE and detected by autoradiography. Molecular mass markers were 205, 150, 100, 75, 50, 37, 25, 20 and 15 kDa, and are indicated by lines on the left-hand side of the figure.

SDS/PAGE analysis followed by autoradiography revealed that overall protein synthesis capacity was generally reduced at subphysiological temperatures, although cells at 32 °C, and even at 27 °C, still displayed an appreciable amount of protein synthesis (Fig. 4B), and at this level of resolution the range of proteins being synthesized was similar at 37, 32 and 27 °C. Below 27 °C, protein synthesis was much more severely attenuated, being most affected in the most temperature-sensitive cell line, P19 (Fig. S3), the cell line whose cellular architecture was also most compromised upon severe cold stress. Protein synthesis rates were fully restored upon rewarming to 37 °C, and close examination of the SDS/PAGE analyses revealed that changes in protein synthesis rates were discernible between control cells maintained at 37 °C and those subjected to cold stress and then rewarming, particularly in the 50–75 kDa range (Fig. 4B).

To examine changes in the synthesis rates of individual proteins at subphysiological temperatures and upon recovery more closely, proteins were resolved by 2D non-equilibrium pH gradient gel electrophoresis (NEPHGE)-SDS/PAGE (pI > 4.5 and size 20–150 kDa; Fig. 5A). Owing to this size range limitation, we did not detect the two well-characterized mammalian cold shock-inducible proteins Cirp and Rbm3 (< 20 kDa). However, changes in the synthesis rate (up or down) for 25 newly synthesized polypeptides at 32 °C (Fig. 5B) and for 16 at 27 °C (data not shown) relative to those at 37 °C was observed. The synthesis rates of rather more (31) polypeptides changed upon rewarming after cold stress relative to their rates during continuous growth at 37 °C (Fig. 5C). The synthesis rates of a similar number of polypeptides were observed to change for the other rewarming conditions examined: 4–37 °C (27); 10–37 °C (33); and 20–37 °C (23). The range of polypeptides showing altered synthesis rates during cold shock recovery was similar regardless of whether the recovery was from severe or mild cold stress. Most of the changes in synthesis rates upon cold stress or recovery observed in CHOK1 proteins were observed in similar experiments with P19 and 3T3 cells (Fig. S3).

Figure 5.

 Changes in protein synthesis rate in CHOK1 cells during exposure to, and recovery from, cold stress. CHOK1 cells were radiolabelled for 1 h under the indicated conditions, and then 100 μg of extracted proteins were resolved by two-dimensional NEPHGE followed by SDS/PAGE. (A) Proteins extracted from cells maintained at 37 °C, detected by Coomassie staining (left panel) or by autoradiography (right panel). (B) Proteins extracted from cells held at 32 °C for 6 h. (C) Proteins extracted from cells held at 32 °C for 6 h and then transferred to 37 °C for 5 h. In (B) and (C), white arrows identify increased synthesis relative to 37 °C, black arrows identify decreased synthesis relative to 37 °C, and numbers identify polypeptides referred to in Table 2.

The more abundant CHOK1 polypeptides (i.e. those readily visible by Coomassie staining) showing altered synthesis rates during cold stress or during recovery were excised and subjected to in-gel tryptic digestion followed by MS analysis by MALDI-TOF MS for their identification. In a number of cases, the identity was confirmed by immunoblot, and in a few cases [subunits of the cytoplasmic molecular chaperone chaperonin containing T-complex polypeptide 1 (CCT)], by a combination of immunoblot and previously identified positions on the NEPHGE-SDS/PAGE system used, and with this approach, 17 CHOK1 polypeptide spots were identified (Table 1). All identified proteins are relatively abundant proteins, but they cover a broad spectrum of functional activities in cells, including energy metabolism, cytoskeleton organization, protein synthesis, protein secretion and purine biosynthesis. The majority, however (9/17), were molecular chaperones deriving from at least three subcellular compartments, the cytoplasm (CCT subunits, Hsc73, and HOP p60), the mitochondrion (Grp75, Hsp60), and the endoplasmic reticulum (ER) (ERp57).

Table 1.   Identification of CHOK1 proteins showing above-average temperature-dependent changes in synthesis rate. NC, no change; ↑, increased abundance; ↓, decreased abundance.
No.Common nameMethodIdentifier, Swissprotm (Da)pICold stress (°C) Recovery at 37 °C from (°C):
 1NEM-sensitive fusion proteinMSX1565283 8116.38NCNC
 2Grp75MS/blotU9231373 9705.87NCNCNC
 3Hsc73MS/blotM3456170 9895.24
 4HOPMSNM_13891163 1586.40NCNCNCNCNC
 5IMP cyclohydrolaseMSD8951464 7056.72NCNCNCNC
 6CCTθBlotZ3716459 6005.43NCNCNC
 7CCTαBlotM3466560 3395.71NCNCNCNCNC
 8CCTδBlotZ3155458 1008.24NCNCNCNC
 9Hsp60MS/blotM2238361 1225.83NCNCNC
10ERp57MSQ91Z8157 2175.98NCNC
11CCTβMS/blotZ3155357 7535.91NCNCNC
12β5-TubulinMS/blotNM_01165550 0954.78NCNC
13ActinMSAB01309842 0875.30NC
14eIF3iMSU3906736 8785.38NCNCNCNC
15Lactate dehydrogenase AMSDQ91266136 7817.01NC
16β-Tubulin fragmentMSAJ71732028 8744.86NCNCNCNC
17Tropomyosin 3MSXM_86068724 9184.88NCNCNCNC

Protein degradation is generally attenuated upon cold stress in mammalian cells

The overall abundance of a polypeptide, and any change in it, depends not only on its rate of synthesis but also its degradation, and so protein degradation rates were also examined at subphysiological temperatures. Examples of the protein half-life determinations for some of the specific CHO proteins investigated are shown in Fig. 6. Interestingly, at subphysiological temperatures relevant to bioprocessing (32 °C), protein degradation was severely curtailed at a global level, being undetectable for all proteins examined at 27 °C over the time period (12 h) investigated. These, remarkably, included the normally very short-lived cell cycle regulator p53. It is notable that the half-life measured for CHO p53 at 37 °C (5.2 h) was longer than reported for this protein in many other cell lines (20–60 min), a fact attributed to the CHOK1 p53 gene having a point mutation in exon 6 sufficient to compromise the normal function; that is, CHOK1 cells fail to arrest in G1 after radiation-induced DNA damage, and the mutant protein is present at high spontaneous levels in these cells [15].

Figure 6.

 The half-life of proteins is increased upon exposure to mild cold stress. CHOK1 cells maintained at 37 or 27 °C were then exposed to growth medium containing 50 μg·mL−1 cycloheximide. At the indicated times, cells were extracted, and 20 μg of protein was resolved by SDS/PAGE, and then detected either by Coomassie stain (upper panel) or by probing immunoblots for the indicated proteins (lower panel). Molecular mass markers as in Fig. 4.

When the effect of cold stress and recovery on overall protein levels was examined by immunoblot, the changes detected were, with a few exceptions, much more subtle than might have been anticipated from the observed changes in their synthetic rates, particularly as their degradation rates were negligible at 27 °C. This is most likely because the polypeptides chosen for identification were relatively abundant proteins, so that a small change in amount due to increased synthesis rate might be difficult to detect by immunoblot against a background of the total polypeptide. Furthermore, the time periods investigated were short (6 h of cold shock, 5 h of recovery), which would make detection of small changes in the total amount of an abundant protein difficult by this method. An exception to this was the θ-subunit of the cytoplasmic chaperonin CCT. The level of this particular subunit was particularly sensitive to both hypothermic and hyperthermic stress (Fig. 7). Comparatively, significant changes in the levels of other subunits of this molecular chaperone were not detectable by immunoblot (data not shown), even though the synthesis rates of several changed during cold shock and/or recovery from cold stress (Table 1).

Figure 7.

 Marked changes in synthesis rate do not correlate with large changes in overall amounts of relatively abundant proteins. CHOK1 cells maintained at 37 °C, or exposed to the indicated temperature changes, were extracted, and 20 μg of protein was resolved by SDS/PAGE and then detected by Coomassie stain (A) or by probing immunoblots for the indicated proteins (B).

mRNA degradation is also attenuated during cold stress in mammalian cells

Quantitative real-time PCR (qRT-PCR) was used to ascertain the levels of mRNAs encoding selected proteins for which synthesis rates changed in response to temperature variation. As it was unclear what would be a suitable mRNA to standardize the data to, the data shown in Fig. 8 have been standardized to the respective values at 37 or 27 °C. The mRNAs monitored showed appreciable degradation rates at 37 °C but were found to be more stable at 27 °C. This agrees with a previous study [8] showing increased mRNA levels at reduced temperature, although the study did not analyse synthesis rates. We note further that the more labile a specific mRNA is at 37 °C, the greater are the changes observed in its levels during exposure to, and recovery from, cold stress.

Figure 8.

 Specific mRNAs are longer-lived at 27 °C than at 37 °C. (A) CHOK1 cells maintained at 37 °C (squares) or 27 °C (diamonds) were then exposed to growth medium containing 2 μg·mL−1 actinomycin D at the same temperatures. (B) CHOK1 cells maintained at 37 °C or held at 27 °C for 6 h without or with a recovery period (crs, cold shock recovery) at 37 °C for 1 h or 5 h. At the indicated times, total RNA was extracted from the cells and the indicated mRNAs were quantified by qRT-PCR. Data are normalized to the initial mRNA content at 37 or 27 °C.

Recovery of cold-stressed cells at 37 °C does not induce a full classical heat shock response

It has been reported that rewarming cold-stressed cells of human origin at 37 °C initiates a heat shock response [16]. Although we also observed that the synthesis rates of several constitutively expressed heat shock proteins increased during recovery from hypothermia, these changes were rather modest when compared with the changes in synthesis rates of these same proteins during recovery from a classical heat shock (hyperthermia; Fig. 9A). Furthermore, upon recovery from heat shock, an increase in the total amount of heat shock proteins could be detected by immunoblot (Fig. 7). Interestingly, Hsp72, which is strictly inducible in rodent cells [17], was not detectable during cold stress or recovery, even though it was clearly induced in the same cells during their recovery from hyperthermic heat shock (Fig. 7, arrowed, Fig. 9A). These observations suggested that the increased expression of constitutive heat shock proteins during recovery from cold stress might be regulated by a different mechanism from that involved during a ‘classical’ recovery from heat shock, during which the inducible form of Hsp70 is robustly expressed.

Figure 9.

 A classical heat shock response is not initiated upon recovery of cold-stressed cells at 37 °C. (A) Proteins extracted from CHOK1 cells that had been maintained at 37 °C, or held at 27 °C for 6 h and then transferred to 37 °C for 5 h (cold shock recovery), or held at 43 °C for 1 h and then transferred to 37 °C for 5 h (heat shock recovery), and then radiolabelled for a further 1 h at 37 °C, were resolved and detected as in Fig. 5. Only the area including Grp75 (spot 2) to actin (spot 14) is shown; the spot numbers refer to the proteins listed in Table 2. Hsp72 is arrowed. (B) An immunoblot of proteins extracted from CHOK1 cells maintained at 37 °C, or held at 4 or 27 °C for 12 h, or held at 43 °C for 1 h, with or without subsequent recovery at 37 °C for 0.5 h or 5 h, probed for HSF1. (C) Immunoblots of SDS/PAGE resolutions (upper panels) or of nondenaturing gel resolutions (lower panels) of proteins extracted from CHOK1 and P19 cells maintained at 37 °C, or held at 4 °C for 6 h, or held at 43 °C for 1 h, with or without subsequent recovery at 37 °C for 0.5 h or 5 h, probed for HSF1. Trimerization of HSF1 upon recovery from cold stress is indicated by an asterisk. CS, cold stress; HS, heat stress.

Transcription of inducible heat shock genes is activated by the binding of heat shock factors (HSFs) to heat shock elements in their promoter-proximal regions [18,19], the best understood being that of HSF1. In unstressed cells, HSF1 exists as a constitutively phosphorylated monomer in the cytoplasm, but during heat stress, HSF1 undergoes trimerization [20] and becomes hyperphosphorylated [21]. It is this hyperphosphorylated, trimeric form that accumulates in the nucleus and binds to heat shock elements, thereby activating transcription [21]. Figure 9B shows the basal level of constitutive phosphorylation of HSF1 determined using immunoblots of HSF1 in cell extracts prepared in the presence of protein phosphatase inhibitors (Fig. 9B, as a cluster of bands ∼ 85–90 kDa). The hyperphosphorylation occurring during heat shock could also be readily demonstrated (Fig. 9B). Cold shock produced a much more subtle change in the HSF1 banding pattern, evident immediately after cold shock and then slowly returning to the constitutive pattern during a subsequent 5 h recovery. This cold shock-induced phosphorylation change in HSF1 was more pronounced with increasing hypothermia, and was most evident in the very cold-sensitive P19 cells.

Trimerization was assessed by chemical cross-linking analysis, using ethylene glycol bis(succinimidylsuccinate), to stabilize the trimer for SDS/PAGE resolution prior to immunoblot detection of HSF1. Heat shock-induced trimerization of HSF1, i.e. the hyperthermic response, was extensive, so that immediately after heat shock, almost all HSF1 was in the hyperphosphorylated, trimeric form (Fig. 9C, lower panel). In contrast, little trimeric HSF1 was evident immediately after cold (hypothermic) shock, and only modest amounts were present during recovery from this cold stress, even though the synthesis rates of constitutive heat shock proteins were increased at this time. Once again, although this response was stronger in the most cold-sensitive cell line, P19, it was still weak in comparison to that observed upon heat stress. These findings collectively suggest that the recovery from cold stress, at least in rodent cells, does not initiate a classical heat shock response, and that any response initiated through HSF1 is comparatively weak or restricted in comparison to a classical heat shock response.


Here we report changes in the cellular architecture, and the synthesis and degradation rates, of specific proteins in mammalian cells subjected to both mild and severe cold stress, and during recovery from hypothermic shock. Collectively, they help to define the specific cellular responses and protein players during cold stress and recovery. The changes identified here in the synthesis and turnover rates reveal that adaptations are easy to miss when comparing total protein levels monitored either by densitometry-based studies (typically, global proteomic ‘snapshot’ studies) or by immunoblot. Our studies have shown that subphysiological temperatures induce specific changes in synthesis rates for proteins involved in a wide spectrum of cellular activities, including energy metabolism, cytoskeletal organization, protein synthesis, purine biosynthesis, secretion and, most particularly, molecular chaperone function.

Representative molecular chaperones from three intracellular compartments, the cytoplasm, the mitochondrion, and the ER, were all detected as part of the adaptive changes of cells exposed to mild hypothermia and, more especially, in cells recovering from this state. It is of particular interest that the synthesis rates of the cytoplasmic molecular chaperones Hsc73 and HOP/p60, and of the ER chaperone ERp57, were increased upon cold stress at 27 °C but not at 32 °C. The strength of hydrophobic interactions decreases with decreasing temperature, and so higher orders of protein structure become less stable at subphysiological temperatures [22]. Thus, at 27 °C, this must become problematic and generate unfolding of existing proteins and/or compromise the folding of newly synthesized proteins, as appreciable protein synthesis is still taking place at this temperature. Furthermore, as protein degradation becomes undetectable at 27 °C, the cell responds to the unfolded protein load by increasing the synthesis of selected molecular chaperones to sequester unfolded proteins until more favourable conditions, including revival of turnover apparatus, are restored.

Rapid recovery of protein synthesis capacity upon rewarming after cold stress would be expected to increase the requirement for molecular chaperones involved in protein folding, particularly in the cytoplasm. However, we also observed increases in the synthesis rates of chaperones in the mitochondrion and the ER after restoration to normothermic conditions. This will undoubtedly be, in part, a response to the overall increase in protein synthesis activity, but the fact that two of these chaperones, Grp75 and ERp57, are redox-sensitive chaperones indicates that the resumption of metabolic activity upon rewarming increases the free radical load on the cell, as might be expected. As the synthesis of the mitochondrial chaperones did not increase during cold shock at 27 °C, this further supports the idea that it is a change in the redox state upon rewarming that is the main stimulus for the increased synthesis rate of the mitochondrial chaperones during recovery from cold stress.

During recovery from cold stress, we also detected increased synthesis of several constitutive heat shock proteins but not of the classical heat shock protein, inducible Hsp72. Kaneko et al. [23] also reported no increase in Hsp72 mRNA upon rewarming NIH 3T3 cells from 32 to 37 °C. Earlier studies using human cell lines did detect increased amounts of Hsp72 upon rewarming after cold shock [16]. An explanation for this discrepancy is that in human cells, Hsp72 is constitutively expressed, whereas in rodent cell lines it is strictly inducible [17]. It would appear, then, that the heat shock proteins induced during recovery from cold stress are the constitutive heat shock proteins, not the strictly inducible ones. Specifically with regard to heat shock protein induction, our findings show that the HSF1 activation process during recovery from cold stress is different from that induced during the classical heat shock response. The degree of HSF1 hyperphosphorylation varies from robust in the normal heat shock response to only a partial response as reported here for cold shock recovery, but also following exposure to certain antimicrotubule drugs used in cancer chemotherapy [24]. Under these latter circumstances, not only HSF1 hyperphosphorylation but also HSF1 trimerization occurred at a reduced level, and only induction of the constitutive heat shock proteins Grp75 and Hsp60, not of inducible Hsp72, was detected.

It is generally accepted that cold stress results in the attenuation of mRNA translation, although we show here that at mildly hypothermic temperatures (27 and 37 °C), protein synthesis is active, although reduced, and that both the banding pattern and relative intensity of polypeptides synthesized at these lower temperatures remain very similar to those observed at 37 °C. Translation is a tightly controlled process, modulated greatly by the (de)phosphorylation of key initiation and elongation factors. Previous studies have shown that mutant initiation factors can elevate the effects of such a slowdown in mRNA translation upon cold stress [25]. Here, we observed that cold stress at 32 °C results in reduced levels of newly synthesized eIF3i, a subunit of initiation factor 3. Upon recovery, this is reversed and eIF3i levels are increased. Although eIF3i is essential for mRNA translation in vivo [26–28], it is not essential for the reconstruction of initiation complexes that can scan and find the AUG start codon [29]. Therefore, its role in vivo is likely to be related to regulation of initiation. Furthermore, overexpression of eIF3i has been shown to be associated with increased cell proliferation, an accelerated cell cycle, and an increase in cell size, whereas the knockdown (by RNA interference) of eIF3i resulted in the reverse of these effects [30,31]. These opposing consequences of eIF3i knockdown or overexpression are mirrored in the observations here of the cellular responses to cold stress at 32 °C and recovery, respectively. It is therefore likely that eIF3i plays a pivotal role in directing cell growth and proliferation upon cold stress and subsequent recovery.

It has been reported elsewhere that cells cultivated under mildly hypothermic conditions undergo cell cycle arrest, predominantly in G1, but also in G2/M [32], and it has recently been suggested that this is in part due to expression of the RNA-binding cold shock proteins Cirp and Rbm3, as their overexpression under normothermic conditions can lead to cell cycle arrest [9]. Previous reports, however, have shown that p53-deficient mammalian cells do not show cell cycle arrest at mildly hypothermic temperatures [33,34], and that at 4–20 °C, p53 induces p21 (WAF1) expression [34]. Our results support this mechanism of p53-mediated cell cycle arrest. p53 in CHOK1 cells has a point mutation that confers unusual stability on this protein and prevents these cells undergoing a normal response to DNA damage, i.e. induction of p21 and consequent cell cycle arrest in G1 [15]. Nevertheless, during mild cold stress, we observed an increase in the level of p53 in CHOK1 cells, a change in p53 isoform pattern due to post-translational modification, and induction of p21 expression. Furthermore, re-entry of cells into the cell cycle upon return to normothermic conditions could be mediated by the increased synthesis of Grp75 that we observed under these conditions. Expression of Grp75 has a two-fold positive effect on cell cycle progression. When present in the cytoplasm, it sequesters p53 [35], thereby preventing entry into the nucleus and subsequent activation of p21 transcription. Furthermore, p53 binding to the centrosome [36,37], which is inhibitory to centrosome duplication, is antagonized by Grp75 [38]. Additionally, Grp75 itself binds to the centrosome, thereby activating Mps1 protein kinase, the activity of which is essential for the initiation of centrosome duplication [39]. Under this model, cell cycle arrest upon cold stress and then re-entry upon recovery is modulated and controlled via the balance of p53 and Grp75 levels.

Finally, our electron microscopy studies and Oil Red O staining show the presence of lipid-containing vesicle-type structures under conditions of severe cold stress. These vesicle-like structures may be the result of lipid material being secreted from the cell, or alternatively, these vesicles may only be observed under severe cold stress because the membrane rigidity and/or membrane-associated cell functions are so severely compromised at very low temperatures that this results in the arrest of the vesicles before secretion, whereas at higher temperatures these are secreted efficiently and hence not observed. It is well known that cold stress results in membrane rearrangements [40], and changes in cellular lipids have been linked to the heat shock response in yeast [41]. More recent research has shown that changes in the lipid composition of the cell membrane induce the phosphorylation of p53 by the ataxia–telangiectasia and Rad-3 related kinase [42], and we are now investigating whether cold stress-induced cell cycle arrest is due to this ataxia–telangiectasia and Rad-3 related kinase activation of the p53–p21 signalling pathway.

In conclusion, we have here identified a number of mechanisms involved in the response of in vitro cultured mammalian cells to mild and severe cold stress, and in recovery from such stress. In addition to a global decrease in mRNA and protein turnover, the synthesis of specific proteins involved in regulating cell growth, proliferation and mRNA translation are upregulated or downregulated during cold stress and recovery. Furthermore, changes in the lipid composition of the cell may underpin these responses, especially upon severe cold stress. On the basis of the results presented here, we suggest that the cytoskeleton, and the balance in the levels of p53, Grp75 and eIF3i, are likely to be of particular importance during the response to, and recovery from, cold stress that allows mammalian cells to survive and recover from low-temperature stress.

Experimental procedures

Cell lines, routine culture conditions, and treatment conditions

CHOK1 cells were sourced from the European Collection of Cell Cultures and P19 cells from P. Andrews, University of Sheffield, UK. Cells were routinely cultured in DMEM/F12 (Invitrogen, Paisley, UK) supplemented with 200 mm l-glutamine, 500 μm glutamic acid, 500 μm asparagine, 30 μm adenosine, 30 μm guanosine, 30 μm cytidine, 30 μm uridine, 10 μm thymidine, 1% nonessential amino acids (Invitrogen, Paisley, UK), and 10% (v/v) heat-inactivated fetal bovine serum (PAA Laboratories Ltd, Yeovil, UK) at 37 °C in a 5% CO2 atmosphere. NIH 3T3 cells were also sourced from the European Collection of Cell Cultures and maintained as above, except that DMEM was used in place of DMEM/F12. For radiolabelling, the routine maintenance media were replaced with cysteine/methionine-deficient DMEM (Sigma-Aldrich, Poole, UK) supplemented with 10% (v/v) dialysed, heat-inactivated fetal bovine serum, 2 mm glutamine and 1770 kBq·mL−1 Pro-Mix l-[35S] cell labelling mix (GE Healthcare, Chalfont St Giles, UK), and then incubated for 1 h at the indicated temperature. Uptake and incorporation of the 35S-labelled amino acids was as previously described [43]. Cold shock was undertaken in routine medium for 6 h or 30 h at 4, 10, 20, 27 and 32 °C in appropriately regulated incubators. Heat shock was also undertaken in the routine culture medium for 1 h by flotation in a water bath at 43 °C. Treatment of cells with the antimicrotubule drug nocodazole was performed in routine medium at 1–3 μg·mL−1 for 2 h at 37 °C. Recovery incubations were undertaken in routine culture medium at 37 °C for 0.5, 1.5 and 5 h. For the determination of mRNA half-lives, cells were incubated in routine culture medium containing 2 μg·mL−1 actinomycin D. For protein half-life determinations, cells were incubated in routine culture medium containing 50 μg·mL−1 cycloheximide.

Extraction of RNA and protein from cell pellets

Total RNA was prepared from intact cells using the commercially available RNeasy kit (Qiagen). Cell extracts for protein analyses were prepared by lysing cells into ice-cold extraction buffer [20 mm Hepes/NaOH, pH 7.2, containing 100 mm NaCl, 1% (w/v) Triton X-100, protease inhibitors (10 μL·mL−1 leupeptin, 2 μg·mL−1 pepstatin, 0.2 mm phenylmethanesulfonyl fluoride) and protein phosphatase inhibitors (50 mm NaF, 1 mm activated Na3VO4)]. Cell lysates were then centrifuged at 16 000 g for 2 min at 4 °C, and the resulting supernatants were retained for further analysis. For the determination and detection of HSF1 trimer formation levels, cell extracts were cross-linked with ethylene glycol bis(succinimidylsuccinate) (Sigma) at room temperature for 30 min, and then blocked with 50 mm Tris/HCl (pH 7.5) at room temperature for 15 min.

Gel electrophoresis analysis of protein extracts

For SDS/PAGE analysis, 10% separation gels were utilized according to the procedure of Laemmli [44]. Prior to NEPHGE-SDS/PAGE, the proteins in cell extracts were precipitated overnight with four volumes of acetone at −20 °C. Following NEPHGE-SDS/PAGE, resolved proteins were detected by Coomassie staining and/or autoradiography using Hyperfilm MP (GE). Gel images were analysed using the commercially available progenesis PG200 software package (Nonlinear Dynamics, Newcastle-upon-Tyne, UK) to determine spots that had changed in abundance. Spot detection was undertaken using the spot detection wizard with the parameters set as follows: minimum spot area, 16; split factor, 7; peak location, use centre of mass as peak. Manual splitting of nonsplit spots and deletion of noise were then undertaken. Following spot detection, background subtraction was achieved using the mode of nonspot option with a margin of 45. In-gel tryptic digestion of excised spots and protein identification by MALDI-TOF MS were undertaken according to Smales et al. [45]. Analysis was undertaken on triplicate biological samples, and only spots whose abundance was changed at the 95% confidence level (P < 0.05) relative to the 37 °C control were considered to show significant changes in polypeptide synthesis rates.

Determination of mRNA levels by qRT-PCR

qRT-PCR was used to determine the relative mRNA levels of target genes using the commercially available BioRad iScript qRT-PCR kit according to the manufacturer’s instructions, and the appropriate primers to amplify CHO sequences as listed in Table 2. All reactions were performed using a BioRad DNA Engine Chromo4 Continuous Fluorescence Detector thermocycler (BioRad, Hemel Hempstead, UK). Cycling conditions included a reverse transcription step by incubation at 50 °C for 20 min, followed by heating at 95 °C for 15 min. Sequentially, the target templates were amplified using 39 cycles (30 s at 98 °C, 15 s at 55 °C). The fluorescence threshold value (Ct) was calculated using opticon monitor software (version 3.1; BioRad). For normalization purposes, all levels were normalized to control levels at 37 °C.

Table 2.   Primers used for qRT-PCR experiments described in this article.
NameSequence (5′- to -3′)

Immunoblot analysis

PAGE-resolved polypeptides were transferred to nitrocellulose using standard procedures, and then blocked with 5% (w/v) nonfat milk in NaCl/Pi or for phosphorylation-dependent epitopes in 0.2% (w/v) Tween-20. Antibody probes against α-tubulin (TAT) and β-tubulin (KMX) [46] were gifts from K. Gull (University of Oxford, UK), and the antibody 23c against STOP [13] was a gift from C. Bosc and D. Job (Commisariat A L’Energie Atomique, Grenoble, France). Affinity-purified rabbit polyclonal antibodies against the C-termini of CCT subunits and against Hsc70 were as described elsewhere [47]. Commercial antibodies against Grp75 (clone 30A5), Hsp60 (clone LK-2) and HSF1 (rabbit polyclonal) were from Stressgen, antibody against p53 (clone DO-7) was from Dako, and antibody against p21 was from Santa Cruz Biotechnology (Santa Cruz, CA, USA). Peroxidase-conjugated secondary antibodies were detected by enhanced chemiluminescence using Hyperfilm ECL (GE). Images were analysed using kodak gel logic 100 imaging system software. The linearity of antibody response over the concentration range of target protein used is shown in Fig. S4.

Immunofluorescence microscopy

For immunofluorescence microscopy studies, cells were grown on 13 mm glass coverslips and then fixed with methanol at −20 °C for 5 min, or with 4% (w/v) paraformaldehyde in NaCl/Pi; this was followed by permeabilization with 0.1% (w/v) Triton X-100 in NaCl/Pi. Cells were rehydrated after methanol fixation for 5 min in NaCl/Pi. All coverslips were then blocked for 15–30 min in 3% (w/v) BSA in NaCl/Pi. Incubation with primary antibodies diluted in blocking solution (TAT, 1 : 100; anti-hsp60, 1 : 100) was performed overnight at 4 °C. The appropriate secondary antibodies (anti-mouse tetramethyl rhodamine iso-thiocyanate; Sigma) were diluted 1 : 100 before use. F-actin staining with rhodamine–phalloidin (Molecular Probes, Invitrogen, Paisley, UK) was achieved according to the manufacturer’s instructions. Cells were counterstained with 4′,6-diamidino-2-phenylindole, and coverslips were then mounted in Mowiol containing p-phenylenediamine as antifade. Cells were then examined under a Leica DMR fluorescence microscope, and images were captured with a Leica DC300F digital camera.

Electron microscopy

For electron microscopy cells, were grown at 37 °C or in the cold as described, and then fixed with 2.5% glutaraldehyde in NaCl/Pi, postfixed with 1% osmium tetroxide, and dehydrated with a graded series of alcohols. After two changes of 100% ethanol, they were detached from the flasks by agitation in ethoxypropane, and then embedded in Agar Low Viscosity Resin. Sections were cut at 60–90 nm, stained with uranyl acetate and lead citrate, and examined in a Jeol 1230 transmission electron microscope (Jeol UK, Welwyn Garden City, UK) operating at 80 kV. Images were recorded with a Gatan Multiscan 600CW camera (Gatan UK, Oxford, UK).


This work was partially supported by Grants BB/C006569/1 and BB/F018908/1 from the Biotechnology and Biological Sciences Research Council (BBSRC), UK.