Increased grp78 transcription is correlated to reduced tlr4 transcription in patients surviving sepsis

Summary

Regulated transcriptional readthrough during stress maintains genome structure and ensures access to genes that are necessary for cellular recovery. A broad number of genes, including of the bacterial sensor Toll‐like receptor 4 (TLR‐4), are markedly transcribed on initiating the systemic inflammatory response. Here we study the transcriptional patterns of tlr4 and of its modulator grp78 during human sepsis, and establish their correlations with the outcome of patients. We measured the daily tlr4 and grp78 RNA expression levels in peripheral blood of septic patients, immediately after admission to intensive care, and modeled these RNA values with a sine damping function. We obtained negative correlations between the transcription of tlr4 and grp78 RNA in the survivor group. In contrast, such relation is lost in the deceased patients. Loss of transcriptional homeostasis predicted by our model within the initial 4 days of hospitalization was confirmed by death of those patients up to 28 days later.

Introduction

Adequate regulation of gene expression is a crucial step for long‐term adaptive response to stress, and constitutes the basis for the recovery response 1. Transcriptional readthrough maintains overall genome structure, keeps areas of chromatin open and warrants access to genes that are necessary for adaptation and recovery of the organism during and after severe stress, such as in sepsis 2. Sepsis syndrome is a major public health concern that encompasses the systemic physiological, pathological and biochemical responses of the organism to invading pathogens and their toxins 3. Sepsis is defined as a life‐threatening organ dysfunction caused by a dysregulated host response to infection. Sepsis can develop into septic shock, a subset of sepsis characterized by circulatory and cellular/metabolic abnormalities that substantially increases mortality 4. The recently described ‘genomic storm’ occurring during systemic inflammatory response reported a marked regulation of ~80% of the genome 5. Among the modulated gene targets, tlr4 (Toll‐like receptor 4) is centrally responsible for sensing lipopolysaccharides (LPS) from Gram‐negative bacteria 6, 7, with cross‐reactions to some Gram‐positive bacteria 8. Continued LPS stimulation ultimately induces up‐regulation of the anti‐apoptotic glucose‐regulated protein 78 kDa (GRP78) 9. The early involvement of surface TLR‐4 in mediating systemic responses to both invading pathogens and endogenous ligands and the requirement for GRP78 to ensure proper folding of immunity‐related proteins, i.e. immunoglobulins and cytokines, make these elements essential for sepsis pathogenesis 10.

Our objective is to characterize the changes in the expression levels of tlr4 and grp78 among patients with sepsis, indicative of transcriptional readthrough, as a response to intense physiological stress. Secondly, we seek to determine whether their expression pattern is suggestive of co‐regulation, and whether data are amenable to mathematical modeling to serve as sepsis evaluation tools.

Methods

Patient recruitment

The study involved data collection from clinical and surgical patients, aged 18 years and over (n = 10), referred to the intensive care unit (ICU) of the University Hospital at the University of São Paulo. Patients were diagnosed with sepsis according to Sepsis‐3 definition, including two points in SOFA (Sequential Organ Failure Assessment) and APACHE (Acute Physiology And Chronic Health Evaluation) score within the first 24 h upon admission. Patients were included in this study in the order of their ICU admittance, in accordance with our inclusion and exclusion criteria. Treatment with corticosteroids and/or antibiotics in the days previous to admission to the ICU was an exclusion criterion for recruitment. Additional information was collected, including co‐morbidities, smoking and alcoholism history. All patients who presented comorbidities were excluded from this study, despite the drastic reduction in the cohort sample size.

Transcripts quantification

Two‐and‐a‐half ml of peripheral blood were collected into PAXgene RNA tubes (PreAnalytiX BD; Becton Dickinson, Franklin Lakes, NJ, USA) every 24 h from time of admission (unless otherwise stated), up to 4 days (or less, in cases where the patient died or was discharged). Tubes were kept at –20°C for up to 4 months until RNA isolation. Reaction efficiencies ranged between 95 and 97%. Total RNA was isolated using PAXgene Blood RNA kits (Qiagen, São Paulo, Brazil), and synthesis of cDNA was performed in the presence of RNase inhibitor using random primers and the high‐capacity cDNA reverse transcription kit (Thermo Fisher Scientific, São Paulo, Brazil), according to the manufacturer's instructions. Real‐time quantitative PCR (qPCR) were performed using specific primers and Fast SYBR^® Green Master Mix (Thermo Fisher Scientific) in an Agilent Mx3005P qPCR System (Agilent Technologies, Santa Clara, CA, USA). Triplicate reactions were run for each sample. Oligonucleotides used for transcripts quantification: GRP78 forward (NM_005347.4, exon 1) 5ʹ‐CGAGGAGGAGGACAAGAAGG, GRP78 reverse (NM_005347.4, exon 3) 5ʹ‐CACCTTGAACGGCAAGAACT, glyceraldehyde 3‐phosphate dehydrogenase (GAPDH) forward (NM_001289746.1, exon 7) 5ʹ‐CGACCACTTTGTCAAGCTCA, GAPDH reverse (NM_001289746.1, exon 8) 5ʹ‐CTGTGAGGAGGGGAGATTCA, TLR‐4 forward (NM_138557.2, exon 1) 5ʹ‐CGCTTTCACTTCCTCTCACC, TLR‐4 reverse (NM_138557.2, exon 1‐2) 5ʹ‐ATTAGGAACCACCTCCACGC. Cycling conditions were: 95°C for 20 s, followed by 40 cycles of 95°C for 3 s and 60°C for 30 s. Melting curves were acquired at 95°C. Serial qPCR data on the gene targets normalized to gapdh mRNA are shown in Table 1, as are the protocol, day and times of collection. The mean relative abundance values were calculated using the 2‐ΔΔCt method 11.

Table 1.

qPCR data and sample collection times from sepsis patients.

Patient no.	1	2	3	4	5	6	7	8	9	10
Day 0 TLR‐4 (GRP78)	1·0 (1·0) [10:00 h]	1·0 (1·0) [11:25 h]	1·0 (1·0) [11:00 h]	1·0 (1·0) [11:30 h]	1·0 (1·0) [10:00 h]	1·0 (1·0)  [16:00 h]	1·0 (1·0) [10:00 h]	1·0 (1·0) [00:00 h]	1·0 (1·0) [04:00 h]	1·0 (1·0) [23:00 h]
Day 0.5 TLR‐4 (GRP78)				0·6 (0·64) [23:30 h]		1·97 (1·7) [04:00 h]	4·3 (0·66) [22:00 h]	0·56 (3·6) [12:00 h]	1·97 (0·57) [16:00 h]
Day 1 TLR‐4 (GRP78)	0·3 (1·23) [10:00 h]	0·2 (0·14) [11·:30 h]	2·5 (1·7) [11:00 h]	0·13 (1·5) [11:30 h]	0·89 (1·05) [10:00 h]	1·7 (1·6) [04:00 h]		1·4 (3·07) [00:00 h]	1·71 (0·5) [16:00 h]	1·63 (1·18) [11:00 h]
Day 2 TLR‐4 (GRP78)	0·4 (2·0) [10:00 h]	1·3 (0·74) [23:30 h]	1·5 (1·4) [11:00 h]	0·11 (0·9) [11:30 h]				1·01 (2·8) [10:00 h]	1·0 (0·44) [16:00 h]	1·43 (0·8) [13:00 h]
Day 3 TLR‐4 (GRP78)	0·7 (1·24) [14:00 h]		0·63 (1·4) [11:00 h]	0·25 (0·28) [11:00 h]		0·98 (3·4) [04:00 h]		1·7 (4·0) [10:00 h]	0·45 (0·45) [16:00 h]
Day 4 TLR‐4 (GRP78)	1·0 (0·88) [10:00 h]		0·9 (16·4) [11:00 h]			0·45 (0·4) [04:00 h]		1·6 (4·0) [10:00 h]	0·44 (0·64) [16:00 h]
Outcome	Survival	Death	Death	Survival	Survival	Death	Death	Survival	Survival	Death

Day 0 refers to admission to intensive care unit (ICU) and diagnosis of sepsis. Subsequent sampling was performed every 12 (day 0·5) or 24 h. Times of sampling are indicated in brackets. Quantitative polymerase chain reaction (qPCR) values expressed as relative quantification units, tlr4 in the first line, grp78 within round parentheses in the second line. TLR = Toll‐like receptor; GRP = glucose‐regulated protein.

Clinical and laboratory investigations

Peripheral blood samples as well as clinical and laboratorial data were collected within 6 h after patient ICU admission, and every 24 h afterwards for 4 days. While whole peripheral blood provides a mixed leukocyte population, which is insufficiently characterized and may have varied significantly among patients and across time, we opted for such samples because: (i) their collection and processing reflect real working conditions of intensive care providers at an ICU setting without a cell separation facility, and (ii) every sample's RNA content was normalized against its own housekeeping gene expression, excluding artifacts of significant variation in RNA content intra‐ and interindividuals. The Institutional Research Ethics Committees (CEP‐HU/USP 940/09 SISNEP CAAE 0048.0.019.198‐09, CEPSH‐ICB 975/10) approved of this study and patients signed a written informed consent prior to enrollment in the study.

Sites of suspected or documented infection, blood gas analysis, complete blood count, C‐reactive protein and microbiological cultures were documented, as shown in Table 2) and Supporting information, Fig. S1. During the ICU stay, all patients received medical treatment according to accepted guidelines 12.

Table 2.

Clinical and laboratory data of sepsis patients.

Patient no.	1	2	3	4	5	6	7	8	9	10
Sex	F	M	F	F	M	F	M	M	M	F
Age (years)	32	79	32	89	48	80	43	19	26	81
Infection focus	Uterus post‐ abortion	Bronchopneumonia	Small bowel colitis, peritonitis	Colecistitis	Peritonitis	Hip prothestic infection, bronchopneumonia	Kidney	Pericardium and skin abcesses	Pleural cavity, pneumonia	Septicemia, pneumonia
Bacterial culture	N	C. albicans + S. maltophilia	S. pyogenes + E. coli	E. coli	E. coli + fecal enterococcus	E. coli	S. pyogenes	S. aureus	n.d.	S. aureus + S. hominis
Comorbidities	N	N	N	Hypertension, previous strokes	N	N	Apendicectomy, one month before sepsis	N	N	N
C‐reactive protein at admission	179	123	251	172	499	238	352	372	221	181
Leukocytes at admission	23 400	13 800	9800	3900	2600	20 200	11 200	9500	n.d.	11 700
Neutrophils at admission	0	0	1000	1000	1000	6000	5000	1000	n.d.	n.d.
Lactate at admission	15·9	48·1	8·3	26·2	20·6	12·9	37·4	46·9	46·9	28·1
SAPS 3	58	83	102	96	53	94	82	37	68	82
Mortality (SAPS 3)	31	78	92	89	22	88	76	68	53	76
Causa mortis	n.a.	Sepsis and bronchopneumonia	Sepsis shock refractory to hemodynamic support	n.a.	n.a.	n.d.	Septic shock, multiple organs failure	n.a.	n.a.	Septic shock, bronchopeumonia, acute kidney failure
Outcome	Survival	Death [9]	Death [28]	Survival	Survival	Death [6]	Death [2]	Survival	Survival	Death [3]

N = pathology absent or absence of bacterial growth in vitro; BD = below detection; n.d. = not done; n.a. = not applicable. Normal ranges: C‐reactive protein (CRP) < 1 mg/ml; leukocytes: 4000–10 000 cells/mm³; neutrophils: 1800–7500 cells/mm³; lactate < 20 mg/dl. Numbers in brackets in outcome row indicate the day of hospital death. Survival indicates hospital discharge, and alive at 1‐year follow‐up. SAPS = simplified acute physiology score.

Data analysis

Patient clinical data and outcomes were recorded by medical and nursing staff, but only revealed to those involved in data analysis and modeling at the completion of analyses. Tlr4 and grp78 transcripts data were averaged with OriginPro 9.1 (OriginLab, Northampton, MA, USA), using a linear interpolation algorithm. The resulting averaged curves were fitted to a damped sine wave equation by varying the fitting parameters until 0·9 < R ² < 1. Data for grp78 RNA from patients who succumbed were fitted only to the first 48 h.

Statistical analyses

Data were tested for normality at the 5% decision level using the Kolmogorov–Smirnov normality test. Comparisons among the two groups were performed using Fisher's exact test for pairwise correlations between the averaged values. Effect size was calculated using Cohen's d equation. All calculations were performed in Origin 9.1 (OriginLab).

Results

In our cohort, 80% of the patients (n = 10) had a confirmed bacterial source of infection at the time of death or discharge (Table 2). Our statistical analyses detected significant differences between values from survivors and deceased patients (Supporting information, Fig. S2). Comparisons among the two groups were performed using Fisher's exact test for pairwise correlations between the averaged values of the survivor (n = 5) and the deceased patient groups, within either category of gene transcripts, yielding P‐values of 0·01 (tlr4) and 0·02 (grp78). Between the deceased and the survivors, power was 0·70 (tlr4) and 0·67 (grp78), at 95% confidence level, with α = 0·05. Effect size was calculated using Cohen's d equation, survivors versus deceased, yielding values of 1·868 (tlr4) and 1·208 (grp78). While this analysis is useful for parsing the differences between the expression levels of each gene in patients, it does not yield information on their dynamical transcriptional changes during sepsis, nor does it have predictive value for the ICU clinician. To this end, we studied the relation between the transcriptional profiles of grp78 and tlr4 by using a damped harmonic oscillation model. An energy dissipating oscillator is a system that is able to accommodate exogenous (environmental) and endogenous (i.e. fluctuations in RNA or protein levels) inputs without changing the nature or intensity of its output 13. Transcription of nuclear factor kappa B (NF‐κB) is a known example of a damped oscillator whereby, given a constant source of tumor necrosis factor (TNF‐α (input), the amplitude of the NF‐κB response decreases gradually until cessation 14. Gene transcription in a human osteosarcoma cell line was also modeled to a damped oscillator, with transcription occurring in bursts of decreasing amplitude and frequency 15. Such initial amplitude is evident in our data by plotting the first data points obtained with our cohort, before initiation of treatment (Fig. 1a). This suggests an exacerbated tlr4 expression coupled to a diminished grp78 transcription in patients who succumbed (Fig. 1b,c), while the opposite is present for the survivor group.

Figure 1.

Quantification of transcripts of tlr4 and grp78 from sepsis patients, before treatment initiation. (a) Data for each gene without separation into patient groups. Mean values with error bars of the deceased and survivor groups for tlr4 (b) and grp78 (c). P‐values of less than 0·05 indicated by one star (*), whereas P‐values of less than 0·01 flagged with two stars (**).

In patients who survived sepsis, as shown below in Fig. 2a and Supporting information, Fig. S2c,e, the accumulation of transcripts for both tlr4 and grp78 presented an initial synchronized increase during the initial 6 h (see Fig. 2a, insert 1).

Figure 2.

Averaged mRNA time–courses during sepsis progression. Experimental data and fitted curves for the survivor group (a) and deceased group (b). Squares represent tlr4, triangles represent grp78. Fitted curves for transcripts of the two genes are depicted as lines. For grp78 transcripts in deceased patients (triangles, b), only the sinusoid part of the curve until the switch point at approximately 48 h was considered (refer to text for further details). The insert (1) shows the first 24 h of averaged data in the survivor group. Inserts (2) and (3) show the relative quantification data with standard deviations bars (no interpolation).

This was followed by a marked down‐regulation of tlr4 (square symbols) until 48 h, and by an increase of grp78 mRNA (triangle symbols) that peaked at approximately 24 h, with subsequent continuous decrease. In deceased patients (Fig. 2b and Supporting information, Fig. S2d,f), tlr4 mRNA values (square symbols) increased exponentially, while grp78 (triangle symbols) mRNA from deceased patients oscillated until an exponential increase at a switch point after day 2.

Fluctuations in the transcript levels of tlr4 (survivor group only) and grp78 (both groups) were modeled using an equation describing an energy dissipating harmonic oscillator:

$$\mathit{y}=\mathit{yo}+{\mathit{Ae}}^{\frac{\mathit{x}}{\mathit{t}0}}[sin\mathit{\pi }\left(\mathit{x}-\mathrm{\varnothing }\right)/\mathit{w}]$$ (1)

Data were fitted to a damped sine wave by varying the following parameters: y _o = offset, A = initial amplitude, t ₀ = decay constant, Æ = phase shift and w = period, as shown in Table 3.

Table 3.

Values of the parameters used for curve‐fitting.

	Transcript	Ø (radian)	A (fold changes)	t 0 (fold changes/day)	w (radian/day)
Survivors	TLR‐4	–0·67 ± 0·1	0·12 ± 0·02	3·1 ± 0·9	1·2 ± 0·05
	GRP78	0·6 ± 0·06	0·45 ± 0·07	4·5 ± 0·8	1·9 ± 0·12
Deceased	GRP78	–0·6 ± 0·1	0·08 ± 0·02	0·97 ± 0·4	0·76 ± 0·07

TLR = Toll‐like receptor; GRP = glucose‐regulated protein.

The average tlr4 mRNA values of deceased patients increase exponentially from day 2 and were fitted to the following asymptotic exponential equation:

$$\mathit{y}=\mathit{a}-{\mathit{bc}}^{\mathit{x}}$$ (2)

with fitting parameters: a = asymptote (9·2 ± 0·8), b = response range (8·1 ± 0·7) and c = rate of change (0·84 ± 0·03).

The increase in the levels of tlr4 transcripts on the second day upon ICU admission in survivors (Fig. 2a) may signal the presence of a robust regulatory mechanism that limits the initial amount of available TLR‐4, in order to dampen the subsequent inflammation stage. In contrast, in the deceased group (Fig. 2b), loss of regulation leads to an exponential increase in the production of tlr4 mRNA, exacerbating the inflammatory response with lethal effects for the patients. As endotoxin concentration increases the initial production and the subsequent half‐life of tlr4 mRNA 16, we hypothesize that a robust and physiological response to the septic challenge requires a prompt modulation in TLR‐4 expression and activity, achieved through increased transcription of TLR‐4's different modulators, including GRP78. We measured an opposite transcriptional trend for grp78, reflecting its role as an important TLR‐4 regulator in deceased patients (Fig. 2b. Within the first 48 h to which data could be fitted, the amplitude and decay constants (Table 3) were lower than those found in the survival group.

Discussion

The exponential increase in grp78 transcripts past the 48‐h point in the deceased group (Fig. 2b) may relate to the higher levels of extracellular isoforms of the heat shock protein 70‐kDa (Hsp70) family, including GRP78 17, that are implicated in the progression to septic shock, in the subsequent immunosuppression phase of sepsis 18 and in the serum of septic patients 19. It is possible that the oxidative stress caused by engagement of TLR‐4 and cytokine receptors may contribute to the modulation of serum Hsp70 levels by causing its active secretion 20 or release through necrosis 21. Cell‐membrane localized GRP78 assist in recruiting oxidized phospholipids that promote both endothelial barrier integrity against refractory hypotension 22 or rescue from TLR‐4‐induced acute inflammation 23. While extracellular Hsp70 acts as an agonist to TLR‐4, intracellular members of the Hsp70 family markedly inhibit not only downstream signaling of TLR‐4, but also modulate its ubiquitination and proteasomal degradation 24.

Within the survivor group (Fig. 2a), we observed an inverse relation between the higher amplitude of grp78 transcripts accumulation with respect to tlr4, and comparatively similar decay rates (Table 3). Furthermore, the phase delay for tlr4 transcription succeeded the corresponding transcription of grp78 in survivors, with the inverse trend observed in the group of deceased patients (Table 3). It is important to note that other key TLR‐4 modulators, including myeloid differentiation protein (MD2) and myeloid differentiation primary response gene 88 (MyD88), are also up‐regulated after LPS stimulation of human monocytes and macrophages 14. Such initial high peaks in tlr4/grp78 transcription followed by dampened oscillations may be important for maintaining a balance between the necessity to continuously respond to presence of cytokines (e.g. TNF‐α) while preventing their constitutive activation and transcription that may be deleterious to cells 25. We note that early and intense increase in interleukin‐10 (IL‐10) transcription was associated with complicated recovery of the patients 5. IL‐10 signaling in unstressed cells blocks grp78 transcription by inhibiting nuclear translocation of transcription factor ATF‐6 and its binding to grp78 promoter 26, up‐regulating the expression of TLR‐4, MyD88 and CD14 27. Upon LPS treatment, il10 expression has a very slow onset and is not detected in culture before a 6‐h mark, increasing steadily to 24 h after stimulation 28. Both IL‐10 mRNA and protein responses are slow to occur in human monocytes 29 and microglial cells 30. It is possible that this delayed response may contribute to the initial grp78 up‐regulation that we observed in the survivors group (Fig 2a). As IL‐10 progressively accumulates up to 48 h28, the tlr4 expression may become dominant. However, increasing IL‐10 levels upon TLR‐4 signaling also inhibits the induction of the microRNA miR‐155, that results in higher concentrations of Src homology 2 (SH2) domain‐containing inositol‐5ʹ‐phosphatase 1 (SHIP1), culminating with further inhibition of TLR‐4 own signaling 31. IL‐10 is an anti‐inflammatory modulator responsible for the dampening of the initial proinflammatory response through autocrine/paracrine feedback mechanisms. As such, TLR‐4‐mediated induction of IL‐10 may be subject to intrinsic temporal regulation that delays IL‐10 production to allow time for the initiation of the immune response upon LPS stimulation 32. The combination of positive and negative regulatory loops involving IL‐10 may be responsible for the oscillatory expression patterns we measured in our patients.

All deaths occurred within a period of 28 days upon ICU admission, while all survivors were alive at follow‐up after 1 year, confirming predictions made by our model using data from the first 4 days of hospitalization. Constructing gene expression profiles for key modulators of the sepsis syndrome thus represents a novel and non‐invasive means to assess its evolution. Such gene expression profiles of mediators have previously been introduced for other pathological conditions, such as coronary artery disease 33, or using TLR‐7 and TLR‐8 for acute ischemic stroke 34. Taken together, our observations, along with retrospective studies on the times of sepsis deaths 35 may reflect the requirement for TLR‐4 modulators such as GRP78 to be expressed promptly before TLR‐4 up‐regulation, in order to prevent subsequent excessive proinflammatory signaling. As sepsis is a rapidly changing process, obtaining first transcription points for tlr4 and grp78 is important early on hospitalization. Our findings suggest that sepsis evolution may be inferred if, during the first 36 h of hospitalization, at least three blood samples are collected in order to establish important oscillatory parameters of tlr4 and grp78 RNA accumulation, such as the amplitude and phase shift.

Conclusions

Our research indicates that higher levels of grp78 mRNA, as opposed to lower levels of tlr4 mRNA, anticipated a positive evolution of sepsis, leading to patient survival. In contrast, intense accumulation of grp78 transcripts accompanied by very low levels of tlr4 mRNA in the first 2 days after sepsis onset reflected a poor prognosis. Loss of the damped oscillator profile during transcriptional readthrough may lie at the basis of sepsis pathogenesis, of which TLR‐4 and GRP78 are two key regulators. We envision that model refining through multi‐centric collection of additional ex‐vivo data will shed light on the mechanisms regulating gene transcription during intense systemic inflammatory conditions, and will provide the intensive care personnel with an additional compass to guide therapy.

Disclosures

No conflicts of interest to declare.

Author contributions

R. C. S. performed data and statistical analysis, model building, and wrote the manuscript. C. P. B. performed experiments and statistical analysis, and wrote the manuscript. R. P. recruited patients, collected samples, gathered clinical and laboratory data and wrote the manuscript. F. G. S. recruited patients, gathered clinical and laboratorial data and wrote the manuscript. M. M. D. C. designed the study, warranted funds, performed experiments and data analysis and wrote the manuscript.

Supporting information

Fig. S1. Averaged CRP and lactate time‐course values with standard deviations, measured from days 1‐4 in sepsis patients.

Fig. S2. Complete mRNA dataset for tlr4 (a) and grp78 (b) without categorization into patient groups. Tlr4 mRNA data for each day of measurements (days 1‐4) for the survivor group (c) and deceased group (d). grp78 mRNA data for each day of measurements (days 1‐4) for the survivors (e) and deceased patients (f).