Biological Features of Herpes Simplex Virus Type 1 Latency in Mice According to Experimental Conditions and Type of Neurones

Abstract

Purpose.: To compare the biological patterns of viral transcripts during herpes simplex virus type 1 (HSV1) latent infection according to experimental conditions.

Methods.: Two types of murine models of HSV1 infection were used: the corneal scarification model, often used in studies of HSV1 latency, and the oro-ocular murine model. Two strains of HSV1 were used for the inoculation: SC16, a wild type strain considered as highly neuroinvasive, and KOS, previously described as poorly neurovirulent. The amounts of viral genomes, and those of four types of viral transcripts (immediate-early, early and late, together with latency-associated transcripts [LATs]), were measured by quantitative PCR and RT-PCR in the main sites of HSV1 latent infection at 6 days, 1 and 3 months post inoculation, and the number of LAT-expressing neurons was assessed by in-situ hybridization on histological sections of trigeminal ganglia (TG).

Results.: Using the SC16 strain of HSV1 in the oro-ocular model, immediate-early transcripts were still present at 1 month post inoculation (early stage of latent infection), but were not detected at 3 months (late stage of latent infection). In both cases, early and late viral genes transcripts were not detected, demonstrating the latent nature of the infection with this combination of experimental conditions. In contrast, such progress in the viral gene expression was not observed in the corneal scarification model, particularly when the KOS strain of HSV1 was used.

Conclusions.: These results highlight that the behavior of the virus in the nervous system depends on the method inoculation, and the viral strain. All these parameters are likely to influence the establishment of latent infection.

Introduction

The herpes simplex virus type 1 (HSV1) is a common human pathogen that affects the face and the upper body with recurrent infections, mostly in tissues innervated by sensory neurons (i.e., lips and cornea), and is a major cause of blindness in humans.¹ After a primary infection that is almost always located into the oral mucosa,^2,3 the virus moves by axonal transport to neuronal cell bodies.⁴ Within the neuron nuclei, HSV1 establishes a latent infection. The virus can reactivate and migrate to peripheral tissues, leading to clinical or subclinical herpetic recurrences.^5–7 

During the lytic phase of infection, the synthesis of viral proteins progresses in a regulated cascade, successively immediate-early (IE), early (E), and late (L) proteins, resulting in the formation of new infectious particles.⁵ In contrast, during latency, no viral protein is produced. Latent viral DNA is circularized, and tightly associated with histones.^8,9 During latent infection, HSV1 is transcriptionally repressed with the exception of the repeated domains that encode the latency-associated transcripts (LATs), which classically are considered as the only viral transcripts that strongly accumulate during HSV1 latency.^9–13 

A mouse model of oro-ocular (OO) HSV1 infection that mimics the natural history of HSV1 infection in humans was previously developed.^14,15 In this model, mice develop an acute infection in the iris and the cornea (6 to 8 days post inoculation [dpi]), following viral inoculation in the lip that spontaneously resolves and which is followed by a latent infection in the nervous system. Latency was demonstrated by the expression of LATs in absence of production of infectious particles at 28 dpi in the trigeminal ganglia (TG), the superior cervical ganglia (SCG), the spinal cord, and the hypothalamus (respectively peripheral sensory or sympathetic, and central neurons).^14,15 Using the same OO model, we showed that the TG was characterized by a significant accumulation of both LATs and nonspliced copies of transcripts encoding ICP0, an IE proteins of HSV1 that plays a key role in viral reactivation,^16–18 suggesting that the balance between LATs and ICP0 transcripts may be involved in the maintenance of HSV1 latency in TG neurons. In the same tissues, neither transcripts for thymidine kinase (TK, an early gene) nor UL18 transcripts (a late gene, encoding the capsid protein VP23) were detected, supporting that HSV1 productive cycle was blocked.¹⁹ 

The presence of both ICP0 transcripts had also been observed at 28 dpi in the corneal scarification (CS) murine model, which is used in the vast majority of studies on HSV1 latency in the literature. However, in these studies, transcripts encoding ICP4 (IE protein), TK (E protein), and even UL18 (L protein) were also evidenced,^20–22 which raised the question of the complete extinction of HSV1 replication cycle in tested tissues. Besides the differences in the method of inoculation (CS versus lip injection), these experiments also differed by the HSV1 strain used (KOS versus SC16, respectively). 

To assess the differences in herpetic pathogenesis induced by various conditions of infection, we compared both routes of inoculation (i.e., the mouse OO and CS models) using the two strains of HSV1 (SC16 and KOS). 

Materials and Methods

Virus Strains, Viral Inoculation, and Euthanasia

Wild-type HSV1 strain SC16 and KOS were used.^23–25 Viruses were propagated and titrated as previously described.²⁶ 

Six-week-old female BALB/c mice (Janvier Breeding, Le Genest Saint Ile, France) were used for the experiments. Animals received unrestricted access to food and water. All experimental procedures involving animals were performed under general anesthesia, conformed to ethical standards such as Association for Research in Vision and Ophthalmology (ARVO) Statement for the Use of Animals in Research, and were approved by the local Ethical Committee of UPR-3296-CNRS. 

For the OO model, 1 μL suspension of either SC16 or KOS strain (10² to 10⁶ Plaque Forming Units [PFU]) was inoculated into the left upper lip. For the CS model, similar viral suspensions were inoculated on the scarified cornea of the left eye. Mice were observed daily for clinical signs of ocular infection from days 0 to 100 dpi, as previously described.^14,15 Mice showing clinical signs of ocular disease were selected for the study. In these mice, all clinical signs resolved rapidly after 10 dpi. They were randomly killed at 28 dpi, a time for which latent infection has been evidenced in previous studies,^15,19 and at 100 dpi (late stages of latent infection). As negative controls, mice were inoculated with 1 μL of sterile saline solution and were killed at 6, 28, and 100 dpi. 

Before tissue dissection, intracardiac perfusion of the mice was performed with PBS only for RNA or DNA extraction, or successively with PBS, 4% paraformaldehyde in PBS, and 20% saccharose in PBS for tissue cryosection.^19,26 

Cryostat Sections

The spinal cord and the whole head (containing the SCG and the TG) were prepared and sectioned as previously described.¹⁵ Briefly, after removal of skin and teeth, the complete skull was decalcified for 7 days at 4°C in 0.1 M EDTA dissolved in PBS, stored in PBS containing 20% sucrose for 24 hours at 4°C, and finally frozen before being cut on a cryostat (Bright Instrument, Cambridge, England). The complete spine was treated with the same protocol. Frontal cryosections (10 μm) of the entire brain (decalcified skull) and spinal cord (decalcified spine) were collected into three parallel series. The first and the second series were used for in situ hybridization (ISH) or detection of HSV1 antigens. The third series was kept in reserve. 

In Situ Hybridization

A digoxigenin-labeled riboprobe was produced after HindIII digestion of the pSLAT2 plasmid, provided by S. Efstathiou and in vitro transcription with T7 RNA polymerase, using the RNA Dig transcription kit (Roche, Meylan, France), according to manufacturer's instructions. In situ hybridization was performed as previously described.^15,19,27 Briefly, after rehydration, overnight hybridization was carried out at 72°C. After several stringent washes, sections were incubated with a polyclonal sheep antibody anti-digoxigenin (Roche) overnight at 4°C. After washes, sections were incubated overnight at 4°C with rabbit antibody anti-sheep coupled to peroxidase (Dako, Les Ulis, France). The hybridization signals were revealed with a solution of diamino-benzidine in 0.05% (Metal Enhanced DAB Kit; Thermo Scientific Pierce, Courtaboeuf, France). 

Detection of HSV1 Antigens

For the study on the propagation of KOS strain in the CS model, 12 mice were killed at 2, 4, and 6 dpi. After dissection and preparation of the tissues, HSV1-positive cells were detected as previously described.²⁶ 

DNA and RNA Extraction

For RT-PCR and PCR studies, 39 mice were used. For each of the three combinations of viral strain and method of inoculation, five animals were killed at 6, 28, and 100 dpi. Fifteen additional mice were used as control at the same times. 

Spinal cord, TG, SCG, and hypothalamus were quickly dissected, incubated overnight at 4°C in RNA later (Qiagen, Courtaboeuf, France), and stored at −80°C. Frozen tissues were homogenized using the QIAshredder columns (Qiagen) and ¹/₅ of each homogenized tissue was used for DNA extraction by phenol-chloroform protocol, whereas the remaining ⁴/₅ were used for total RNA extraction using the Rneasy Mini extraction kit (Qiagen) as previously described.¹⁹ 

Production of DNA and RNA Standards for Quantification

The gG gene of HSV1 and the cellular GADPH gene were amplified by PCR, and the 2-kb LATs, ICP4, TK, UL18, and GADPH transcripts were amplified by RT-PCR. Products of amplification were then cloned into pCRII-TOPO vector (Invitrogen, Cergy Pontoise, France). The RNA standards were produced following digestion of the vector with EcoRV (Ozyme, Saint-Quentin-en-Yvelines, France) and in vitro transcription using SP6 polymerase (Roche). 

Quantitative PCR and RT-PCR

Quantification of DNA copies and transcripts was performed using a Light Cycler 2.0 machine (Roche) with Taqman technology, with the exception of GADPH gene and GADPH transcripts that were quantified using SYBR Green technology, as previously described.¹⁹ Respective primers are given in Table 1. 

Table 1

 

Primers and Probes

Target	Name of Primers*	Sequence of Primers
Viral LATs	LAT For	5′ CCC ACG TAC TCC AAG AAG GC 3′
	LAT Rev	5′ AGA CCC AAG CAT AGA GAG CCA G 3′
	LAT Taq	5′ FAM CCC ACC CCG CCT GTG TTT TTG TG Tamra 3′
Viral ICP4 transcripts	ICP4 Nfor	5′ TTT TCC CAC CCA AGC ATC G 3′
	ICP4 Nrev	5′ TCG TCG TCG GCT CGA AAG 3′
	ICP4 For	5′ AGG TGA CCT ACC GTG CTA C 3′
	ICP4 Rev	5′ CTT GTT CTC CGA CGC CAT C 3′
	ICP4 Taq	5′ FAM TCC GCC GTC GCA GCC GTA TCC Tamra 3′
Viral TK transcripts	TK Nfor	5′ TGA AAC TCC CGC ACC TCT TCG 3′
	TK Nrev	5′ ACA CCC GCC AGT AAG TCA TCG 3′
	TK For	5′ TTC TCG CGG CCA TAA CAA C 3′
	TK Rev	5′ ACC GTC TAT ATA AAC CCG CAG 3′
	TK Taq	5′ FAM AGC AAG AAG CCA CGG AAG TCC GCC CG Tamra 3′
Cellular GADPH	GADPH For	5′ CAA GGT CAT CCA TGA CAA CTT TG 3′
	GADPH Rev	5′ GGC CAT CCA CAG TCT TCT GG 3′
Viral DNA (gG gene)	gG For	5′ CTG TTC TCG TTC CTC ACT GCC T 3′
	gG Rev	5′ CAA AAA CGA TAA GGT GTG GAT GAC 3′
	gG Taq	5′ FAM CCC TGG ACA CCC TCT TCG TCG TCA G Tamra 3′
Viral UL18 transcripts	UL18 Nfor	5′ TGGATTCGCTGGACCTCACG 3′
	UL18 Nrev	5′ ATTTAGGTATGCTGGGGTTTCTCC 3′
	UL18 For	5′ AGACCTCAACGTGCTGTACTAC 3′
	UL18 Rev	5′ CGAGTAAACCATGTTAAGGACCAG 3′
	UL18 Taq	5′ FAM CGGGAGTCGCCTCTCGCTGCT Tamra 3′

*  Primer names containing a capital N correspond to the primers used in the nested PCR.

For 2-kb LATs, 2.5 μL extracted RNA were amplified with the Quantitect Probe RT-PCR kit (Qiagen), using 10 μM of each primers and 4 μM of LAT probe. 

For GADPH gene and transcripts, 1 μL extracted DNA or RNA was amplified with the Light Cycler DNA Master SYBR Green I kit or with the Light Cycler RNA Master SYBR Green I kit (Roche), using 6 μM primers. 

For gG gene, 2 μL extracted DNA were amplified with the Quantitect Probe PCR kit (Qiagen) using 10 μM primers and 4 μM gG probe. 

For quantification of ICP4, TK, and UL18 transcripts, we performed quantitative nested RT-PCR. One microliter of RNA was reverse transcribed and subsequently amplified using 10 μM of the external pair of primers using the AccessQuick RT-PCR system (Promega, Paris, France). Finally, 1 μL of the RT-PCR was further amplified using Quantitect Probe PCR kit (Qiagen) using 10 μM of the internal pair of primers and 4 μL of probe. 

Flow Cytometry Analysis (FACS)

TGs from mice killed at 28 dpi were incubated in PBS containing 40 units/mL papain + 4 mg/mL collagenase type II (Worthington Biochemical Corporation, Lakewood Township, NJ, USA) and 4,6 mg/mL dispase type II (MB), at 37°C for 20 minutes. Dissociated cells were pelleted (400 g for 4 minutes), washed twice in PBS, resuspended in 500 μL ice-cold PBS and fixed in 4% paraformaldehyde. Three aliquots of 2.5 × 10⁵ cells were incubated with antibodies specific for either HSV1 antigens (polyclonal rabbit anti-HSV1; Dako), macrophages (mouse anti CD11b; eBioscience, San Diego, CA, USA) or lymphocytes (mouse anti-CD8a; eBioscience), diluted at 1/250 in FACS buffer (PBS 1×, bovine serum albumin 1%). After three washes in PBS, cells were incubated for 1 hour with an anti-rat IgG, Texas red-conjugated antibody (Santa Cruz Biotechnology, Inc., Dallas, TX, USA) for anti-lymphocytes and anti-macrophages antibodies, or anti-rabbit IgG, fluorescein conjugated (Dako). Cells were analyzed by flow cytometry (Cytometer ELITE ESP; Beckman-Coulter, Villepinte, France). Thresholds were set up to include viable cells only, as assessed by forward scatter data, and to include a range of fluorescence representing less than 1% of the fluorescence of control cells. 

Results

Morbidity and Mortality According to Inoculation Methods

In the OO model, the injection of 10⁶ PFU of SC16-HSV1 strain into the lip resulted in a 60% rate of morbidity and a 45% rate of mortality within the first 15 dpi, as previously described.^14,15 In contrast, no sign of infection was observed with the KOS strain, even with an inoculum of 10⁶ PFU (Figs. 1, 2). 

Figure 1

 

Survival rates of mice infected by SC16 or KOS strains of HSV1 depending on infection site. (A, B) Survival rates of mice infected via CS with different titers of inoculums of the SC16 strain (A) or the KOS strain (B) of HSV1. (C, D) Survival rates of mice infected via lip inoculation with different titer of inoculums of the SC16 strain (C) or the KOS strain (D) of HSV1.

Figure 2

 

Rates of ocular inflammation in mice infected by SC16 or KOS strains of HSV1 depending on infection site. (A, B) Rates of ocular inflammation in mice infected via CS with different titer of inoculums of the SC16 strain (A) or the KOS strain (B) of HSV1. (C, D) Rates of ocular inflammation in mice infected via lip inoculation of different titer of inoculums of the SC16 strain (C) or the KOS strain (D) of HSV1.

In the CS model, the rates of survival and ocular inflammation following the inoculation of the SC16-HSV1 strain showed that an intermediate inoculum between 10 and 100 PFU would induce morbidity and mortality rates similar to those observed in the OO model with 10⁶ PFU. We hypothesized that 70 PFU would be the optimal inoculum to achieve this target, which was validated with 30 additional mice inoculated with 70 PFU of SC16 strain. In contrast, 10⁶ PFU of KOS-HSV1 strain were required to induce a comparable morbidity in the SC model (Figs. 1, 2). We observed persistent inflammation and numerous neo-vessels in the cornea at 28 dpi in the CS model (especially with the KOS strain), whereas in the OO model, dilation of vessels and corneal clouding dramatically reduced from 14 dpi, leading to almost normal cornea at 28 dpi (Fig. 3). 

Figure 3

 

Clinical patterns of the ocular disease. Photographs of eyes of mice inoculated with the SC16 strain of HSV1 into the lip: the corneal inflammation had vanished at 28 dpi (A), whereas persistent corneal inflammation was found following inoculation with the KOS strain of HSV1 using CS (B, C).

To compare viral strains and routes of inoculation with similar rates of morbidity and mortality, the following inoculums were chosen for the rest of our study: 10⁶ PFU of KOS or 70 PFU of SC16 in the CS model, and 10⁶ PFU of SC16 in the OO model. Since no comparison was possible with the KOS strain in the OO model (no clinical signs despite the maximal inoculum; i.e., 10⁶ PFU), this combination was not further analyzed. 

Comparison of Viral Propagation Pathways

The propagation pathways of the SC16 strain of HSV1 in the OO model are schematized in Figure 4A.¹⁴ A similar set of tracing experiments was made for KOS-HSV1 in the CS model, and results are schematized in Figure 4B. In both cases, TG and SCG were infected from the acute phase of the disease. 

Figure 4

 

Pathways of viral propagation according to inoculation site. (A) Spread of HSV1 in CS model. (B) Spread of HSV1 in OO model as detailed in references.¹⁴ CG, ciliary ganglion.

Detection of LAT-Expressing Neurons at 28 dpi

LAT-positive TG neurons were detected using ISH assays in 17 mice inoculated in the lip or the cornea.¹⁵ On the side of inoculation (left), despite a similar size of inoculum (10⁶ PFU), LAT-positive neurons were more abundant in the OO model (SC16 strain) than in the CS model with the KOS strain (38.8 vs. 9.2 in the left TG, respectively; Table 2), which likely reflect differences in neuroinvasiveness between the two strains. Following SC16 inoculation in the CS model, some few LAT-positive neurons were detected in TGs following SC16 inoculation in the CS model, but only when a tyramide-biotin system was used to amplify the ISH.¹⁹ This may be the consequence of the very low inoculum (70 PFU of SC16-HSV1), and/or may reflect low expression of LATs at the single neuron level with SC16 strain in the CS model. 

Table 2

 

Numbers of Cells Positive for LAT Detected by ISH in Trigeminal Ganglion Depending on Viral Strain and Inoculation Site

Strain, Viral Dose, Inoculation Site	(Mouse)	Number of Positive Neurons in TG
		Right TG	Left TG
SC16, 106 PFU, OO	(125)	24	44
	(126)	42	43
	(127)	10	25
	(128)	6	15
	(129)	29	68
	(130)	20	38
	Average	21.8	38.8
SC16, 70 PFU, CS	(150)	UTT	UTT
	(151)	UTT	UTT
	(152)	UTT	UTT
	(153)	UTT	UTT
	(154)	UTT	UTT
	Average	UTT	UTT
KOS, 106 PFU, CS	(162)	7	17
	(163)	15	21
	(180)	2	5
	(181)	0	6
	(182)	2	1
	(183)	0	5
	Average	4.3	9.2

 UTT, under the threshold with the usual ISH method (but some positive neurons when the signal was amplified using the tyramide-biotin system).

As controls, no positive hybridization signal was observed on sections from mock-infected mice, and ISH signal persisted on sections of infected animals after DNAse treatment. In addition, no antigen-positive neurons (immunoassay using polyclonal anti-HSV1 antibodies) was observed in histological sections adjacent to those positive for LATs at 28 dpi, suggesting that HSV1 did not reactivate in the ISH-positive tissues. 

Quantification of Viral Transcripts and Genomes

All quantitative PCR (or RT-PCR) assays were made twice for each tissues extract, and were normalized with the amounts of GADPH gene copies (or transcripts) in the corresponding tissue extracts. The average amounts of genome or transcripts copies, and rates of positive structures are represented in Figures 5 to 9 and raw data as Supplementary Tables S1 through S5. 

Figure 5

 

HSV1 DNA copies at different times of infection, depending on viral strain and injection site. Amounts of HSV1 DNA copies were standardized with the number of GADPH DNA copies in the same tissue extracts. Bars represent the average amounts at respectively 6, 28, and 100 dpi, in either the OO model with the SC16 strain (A), or the CS model with the SC16 strain (B), or the KOS strain (C). The evolution of the amounts of HSV1 DNA copies (after standardization) is represented as the ratio between the mean rates observed at 6 and 28 dpi (D) and between 28 and 100 dpi (E).

Figure 6

 

LAT copies at different times of infection, depending on viral strain and injection site and examples of detection by ISH. Amounts of LAT copies were standardized with the number of GADPH transcripts in the same tissue extracts. Bars represent the average amounts at respectively 6, 28, and 100 dpi, in either the OO model with the SC16 strain (A), or the CS model with the SC16 strain (B), or the KOS strain (C). The evolution of the amounts of LATs (after standardization) is represented as the ratio between the mean rates observed at 6 and 28 dpi (D) and between 28 and 100 dpi (E). (F, G) ISH for LATs on sections of left TGs, either from a mouse killed at 28 dpi in the lip with strain SC16 (F), or after inoculation by CS with strain KOS (G), or mock-inoculated (H). The neurons positive for LATs expression are shown by black arrows.

Figure 7

 

ICP4 transcripts copies at different times of infection, depending on viral strain and injection site. Amounts of ICP4 transcripts were standardized with the number of GADPH transcripts in the same tissue extracts. Bars represent the average amounts at respectively 6, 28, and 100 dpi, in either the OO model with the SC16 strain (A), or the CS model with the SC16 strain (B), or the KOS strain (C). The evolution of the amounts of ICP4 transcripts (after standardization) is represented as the ratio between the mean rates observed at 6 and 28 dpi (D) and between 28 and 100 dpi (E).

Figure 8

 

TK transcripts copies at different times of infection, depending on viral strain and injection site. Amounts of TK transcripts were standardized with the number of GADPH transcripts in the same tissue extracts. Bars represent the average amounts at respectively 6, 28, and 100 dpi, in either the OO model with the SC16 strain (A), or the CS model with the SC16 strain (B), or the KOS strain (C). The evolution of the amounts of TK transcripts (after standardization) is represented as the ratio between the mean rates observed at 6 and 28 dpi (D) and between 28 and 100 dpi (E).

Figure 9

 

UL18 transcripts copies at different times of infection, depending on viral strain and injection site. Amounts of UL18 transcripts were standardized with the number of GADPH transcripts in the same tissue extracts. Bars represent the average amounts at respectively 6, 28, and 100 dpi, in either the OO model with the SC16 strain (A), or the CS model with the SC16 strain (B), or the KOS strain (C). The evolution of the amounts of UL18 transcripts (after standardization) is represented as the ratio between the mean rates observed at 6 and 28 dpi (D) and between 28 and 100 dpi (E).

OO Model With SC16 Strain.

The amounts of viral products diverged markedly between neurological structures at 6 dpi (Supplementary Tables S1–S5). For example, viral genome varied from 10⁹ (SCGs and spinal cord) to 10⁷ copies (left TG and hypothalamus), and 2-kb LATs varied from 10⁸ (left SCG and spinal cord) to 10⁵ copies (right SCG). A similar order of magnitude was also found for ICP4, TK, and UL18 transcripts (Figs. 5A155215521552–9A). 

From 6 to 28 dpi, the 2-kb LATs amount diminished in a large proportion in all tissues with the exception of the TGs where they were stable (left TG) or slightly increased (right TG). In comparison, reduction of viral genome copies and ICP4 transcripts was more important, with SCGs as the site of maximal reduction. Indeed, only one-fifth of SCGs was finally positive at 28 dpi for ICP4 transcripts (Figs. 5A, 7A). At this time of infection, TK and UL18 transcripts were not detected in all tissues (Figs. 8A, 9A). 

From 28 to 100 dpi, the amounts of 2-kb LATs were almost stable in all structures, whereas a reduction was clearly observed for genome copies, especially in the TGs. Interestingly, no ICP4 transcripts were detected in all tested tissues at 100 dpi, neither were TK and UL18 transcripts (Figs. 5A155215521552–9A). 

CS Model With the SC16 Strain.

Despite the lower size of inoculum, the amounts of viral DNA and transcripts in the TG at 6 dpi were similar to those observed with the OO model, while SCG, hypothalamus, and spinal cord were less infected (Figs. 5B155215521552–9B; Supplementary Tables S1–S5). This could reflect a very effective infection of primary neurons in the CS model, due to the very high number of nerve endings into the cornea and/or a more efficient replication of the virus in the cornea compared to the lip (Figs. 5B155215521552–9B; Supplementary Tables S1–S5). 

From 6 to 28 dpi, the decrease in the number of viral genome copies and 2-kb LATs was much less important in the CS model than in the OO model, with the exception of the TGs (no dramatic reduction, as in the OO model). In three out of six tissue types, the average number of ICP4 transcripts was only slightly reduced and approximately one-half of tissue samples were still positive for TK transcripts. However, UL18 was no more detected at 28 dpi (Figs. 5D155215521552–9D). 

From 28 to 100 dpi, the average number of 2-kb LATs increased in the TGs, and the number of viral genome copies decreased, leading to a final viral load less important than in the OO model. Three out of six TGs and two out of six SCGs were still positive for TK and ICP4 transcripts at 100 dpi, respectively, but these viral transcripts were not found in the other tissues, and UL18 was not detected (Figs. 5155215521552–9; Supplementary Tables S1–S5). 

CS Model With the KOS Strain.

These experimental conditions induced less genome copies, and less ICP4, TK, or UL18 transcripts at 6 dpi than with the OO model (SC16 strain), despite a similar titer of inoculums (10⁶ PFU; Figs. 5C155215521552–9C, Supplementary Tables S1–S5). Moreover, the 2-kb LATs increased in all structures from 6 to 28 dpi and 100 dpi, in contrast with the other experimental conditions. Finally, a delay in the extinction of productive infection (and/or the onset of latent infection) was observed, as proved by the presence of UL18 transcripts in 3 out of 30 tissue samples at 28 dpi, and the presence of TK transcripts in 6 out of 18 tissue samples at 100 dpi (Figs. 5155215521552–9; Supplementary Tables S1–S5). 

FACS of TG Dissociated Cells

The prolonged clinical inflammation of the corneas in the CS model (especially with the KOS strain) and the results of RT-PCR assays were confirmed with the FACS analysis on dissociated cells from TGs at 28 dpi. Less HSV1-positive cells were observed in the OO model than following CS inoculation (with either KOS or SC16 strain, P = 0.006 and 0.002, respectively), and KOS strain in CS model induced more lymphocytes and less macrophages in TGs at 28 dpi than the SC16 strain inoculated in the cornea or the lip (P = 0.04 and 0.004, respectively; Fig. 10). 

Figure 10

 

Flow cytometry in TGs at 28 dpi, depending on viral strain and injection site. Aliquots of cells were immunostained for makers of either HSV1 infection (polyclonal anti-HSV1 antibodies), macrophages (anti-CD11b antibodies), and lymphocytes (anti-CD8a antibodies). The results are given as a percentage of total dissociated cells. The percentage of HSV1 cells was significantly lower following inoculation of SC16 in the OO model than in the CS model, with either the SC16 (*1) or the KOS (*2) strain of HSV1. The percentage of lymphocyte was higher following the inoculation of KOS compared to SC16 in the CS model (*3), whereas the opposite was observed for macrophages (*4).

Discussion

The expression of HSV1 genes during the latent phase of infection has been studied for years, but several questions remain unanswered, including the real state of gene expression. Early works showed that LATs were the only viral product abundantly expressed during HSV1 latency,^{5,10–12,28,29} which correlates with the fact that the LAT regions are transcriptionally active during latency unlike the rest of the viral genome.^9,13,30 Interestingly, genes encoding lytic infection-associated ICP0 and ICP4 proteins are also localized in the LAT locus, which could explain some basal level of transcription during the latent stage of infection. Indeed, a weak expression of transcripts encoding ICP0 and ICP4 was reported in the TG at 28 days after the inoculation of the KOS or the 17+ strains of HSV1 by the CS method.^20–22 In these studies, TK transcripts were also found, which was rather unexpected since the corresponding gene is located in a theoretically highly repressed part of the HSV1 genome during latency.^9,13,31 In contrast, in our previous experiments based on the inoculation of the SC16-HSV1 strain in the lip, we observed ICP0-encoding transcripts in the TGs at 28 dpi, but no TK or UL18 transcripts (respectively early and late transcripts of HSV1),¹⁹ which was suggestive of a strong repression of the replication process in the tested tissues. The present study combining two methods of inoculation and two viral strains provides further information on the impact of experimental conditions on the viral gene expression during latent HSV1 infection. 

The Expression of Viral Transcripts at 6 dpi Varies With the Viral Strain and the Route of Inoculation

To induce similar rates of ocular morbidity and mortality, 10⁶ PFU of SC16-HSV1 were required in the OO model versus only 70 PFU in the CS model (Figs. 1, 2). Similarly, 10⁶ PFU of the KOS strain induced an ocular disease in the CS model but not in the OO model. These differences in the sizes of inoculums mandatory to induce clinical infection according to the site of inoculation are likely related to the much higher number of nerve endings into the cornea compared to the lip.³² The differences between the four combinations of experimental conditions also reflect the importance of the neuroinvasiveness of the viral strains used. The SC16 strain is considered as a wild type and neurovirulent strain,^33–35 isolated from a human labial lesion.^24,25 In contrast, the KOS strain is known to have reduced virulence in vivo.^36–39 Combined with the low neuronal input in the lip, this explains the lack of morbidity of this strain in the OO model. The low virulence of the KOS strain was also highlighted with PCR and RT-PCR assays at 6 dpi (Figs. 5155215521552–9; Supplementary Tables S1–S5). These results are in keeping with the results reported in a rabbit ocular infection model,⁴⁰ or in the mouse after inoculation into the vitreous body where 90 PFU of SC16 strain induce 100% of bilateral retinitis²⁶ versus 19% to 57% with 1.5 × 10⁴ to 2 × 10⁵ PFU of the KOS strain.^41–43 The KOS strain is known to harbor a mutation within the gene encoding viral glycoprotein B (gB),^44–47 which could explain the reduced virulence since gB is involved in viral adsorption and penetration, and in the cell-to-cell viral spreading.⁴⁸ Additional mutations have also been described in the US9 gene,⁴⁶ involved in neuronal transport,^49,50 and in the US8A gene. This gene overlaps the US8 gene, encoding gE,⁴⁶ a protein that is important for anterograde transport of the virus⁵¹; that is, from the soma to the nerve endings. Together with the gB modifications, these mutations likely reduce the uptake of virus particles by second order cells and the anterograde transport of KOS. Indeed, our experiments clearly showed a reduced level of infection in second order neurons, that is, those infected either by transneuronal transport (via the synaptic junctions) or by local transport (Fig. 4). The reduced anterograde transport may also explain the lack of ocular infection when KOS was used in the OO model. Moreover, two mutations are also present in the ICP34.5 gene, which is known to have a major impact on neurovirulence.^37,52 All these modifications of the KOS strain, compared to the wild type SC16 strain, likely explicate the reduction of the expression of KOS in the second order neurological structures, like the spinal cord and the hypothalamus. 

Expression of Viral Transcripts at Late Stages of Infection: Is Viral Latency Achieved?

While no TK or UL18 transcripts were detected in the OO model at 28 dpi, TK transcripts were found in a significant proportion of tissue extracts at 28 dpi in the CS model with both viral strains, and even at 100 dpi with the KOS strain (¹/₆ of samples). Similarly, 10% of the tissue extracts at 28 dpi were also positive for UL18 transcripts. Flow cytometry showed a higher proportion of HSV1 antigen positive cells from dissociated TGs at 28 dpi with the CS model than with the OO model (Fig. 10). Furthermore, there were more LAT-expressing neurons in the TG and more amounts of LATs in the OO model (SC16 strain) than in the CS model (Fig. 6), suggesting that the accumulation of LATs in the TG neurons at 28 dpi, described as a marker of the balance from productive to latent HSV1 infection,^19,28 does not occur as efficiently in the two models. 

Taken together, these results suggest that the complete extinction of HSV1 replication cycle is not fully obtained in the CS model, especially with the KOS strain. 

Role of Inflammation Induced by CS

Corneal scarification is a major trauma that induces a persistent inflammatory response in the TG⁵³ and an up-regulation of transcriptional factors such oct-1, c-jun, and c-fos,⁵⁴ which have binding sites in the promoter regions of several IE and E genes of HSV1.^55–57 In the CS model, all animals still exhibited ocular inflammation at 28 dpi (which persisted over 3 months), with active corneal neovessels and clouding of the cornea, whereas eyes were not clinically inflamed in the OO model at 28 dpi and later (Fig. 3). Flow cytometry showed more lymphocytes at 28 dpi in the CS model, especially with the KOS strain, than in the OO model. Indeed, unrelenting infiltration of inflammatory cells has been previously described in the TG following CS,^58,59 and the modifications of gB in the KOS strain,^44,45 which is one of the targets of both protective antibodies and cytotoxic T cells⁶⁰ may enhance this phenomenon. Since corneal nerve injury is a well-known triggering factor for HSV1 reactivation,^61,62 the up-regulation of the viral gene expression in the CS model could be due to injury-like stimulation of the nerve ending by the persisting inflammation in the cornea. 

HSV1 Behavior and Type of Neurons

The OO model clearly showed a distinct evolution of HSV1 infection according to the type of neurological structures. From 6 to 28 dpi, there was a less important decrease of ICP4 transcripts and viral genome copies in the TG neurons than in other tested tissues, and LATs accumulated instead of decreasing. Together with our previous results,¹⁹ these results strengthen the observation that HSV1 exhibits a special behavior in the TG during the onset of latent infection, which could be defined as follows: few reduction (10¹ to 10³) of both viral genome copies and ICP4 transcripts, and accumulation of LATs and of nonspliced ICP0 transcripts. 

This typical pattern of HSV1 latency observed in the OO model was less clearly observed in the CS model, suggesting that the OO model using SC16 strain may be particularly relevant for studying both the cellular and viral events that occur during the latent stage of infection. It also appears as an excellent basic system for studying HSV1 reactivation, since any modification induced by a reactivation-triggering factor should be easily detected. Consequently, this animal model may also qualify as a well-adapted system for assessing emerging preventive antiviral strategies in the future, as the basal level of viral gene expression during latent infection is clearly defined. 

Finally, our results show how the viral gene expression differs upon the type of neurons, especially during the establishment of HSV1 latency, as previously suggested.^14,35 Trigeminal neurons, and even some subtypes of them, seem to be more susceptible than other neurons to experimental viral reactivation,^63,64 and this is supported in clinical practice by the frequency of herpetic in tissues innervated by TG neurons (lips and cornea), whereas upper body tissues that are not directly connected to the TG (e.g., the iris, the retina, or the brain) are rarely injured.⁶¹ The unique pattern of HSV1 gene expression observed in the TG neurons suggests that some special events occur in these cells, but the question on whether these peculiar features are the consequence or the source remains to be addressed. 

Acknowledgments

The authors thank Noura Ahmar-Erras and Christel Picard for technical assistance, and Magali Breckler for careful reading of the manuscript. 

Supported by two grants (ANR-005-MIIM-008-01 and ANR-13-BSV3-0001-02) from the Agence Nationale de la Recherche. 

Disclosure: S. Cavallero, None; N. Huot, None; L. Francelle, None; P. Lomonte, None; T. Naas, None; M. Labetoulle, None